4 * Support for extracting WIM images, or files or directories contained in a WIM
9 * Copyright (C) 2012, 2013, 2014 Eric Biggers
11 * This file is part of wimlib, a library for working with WIM files.
13 * wimlib is free software; you can redistribute it and/or modify it under the
14 * terms of the GNU General Public License as published by the Free
15 * Software Foundation; either version 3 of the License, or (at your option)
18 * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
19 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
20 * A PARTICULAR PURPOSE. See the GNU General Public License for more
23 * You should have received a copy of the GNU General Public License
24 * along with wimlib; if not, see http://www.gnu.org/licenses/.
28 * This file provides the API functions wimlib_extract_image(),
29 * wimlib_extract_image_from_pipe(), wimlib_extract_paths(), and
30 * wimlib_extract_pathlist(). Internally, all end up calling
31 * do_wimlib_extract_paths() and extract_trees().
33 * Although wimlib supports multiple extraction modes/backends (NTFS-3g, UNIX,
34 * Win32), this file does not itself have code to extract files or directories
35 * to any specific target; instead, it handles generic functionality and relies
36 * on lower-level callback functions declared in `struct apply_operations' to do
37 * the actual extraction.
44 #include "wimlib/apply.h"
45 #include "wimlib/dentry.h"
46 #include "wimlib/encoding.h"
47 #include "wimlib/endianness.h"
48 #include "wimlib/error.h"
49 #include "wimlib/lookup_table.h"
50 #include "wimlib/metadata.h"
51 #include "wimlib/pathlist.h"
52 #include "wimlib/paths.h"
53 #include "wimlib/reparse.h"
54 #include "wimlib/resource.h"
55 #include "wimlib/security.h"
56 #include "wimlib/unix_data.h"
58 # include "wimlib/win32.h" /* for realpath() equivalent */
60 #include "wimlib/xml.h"
61 #include "wimlib/wildcard.h"
62 #include "wimlib/wim.h"
70 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x80000000
71 #define WIMLIB_EXTRACT_FLAG_IMAGEMODE 0x40000000
73 /* Keep in sync with wimlib.h */
74 #define WIMLIB_EXTRACT_MASK_PUBLIC \
75 (WIMLIB_EXTRACT_FLAG_NTFS | \
76 WIMLIB_EXTRACT_FLAG_UNIX_DATA | \
77 WIMLIB_EXTRACT_FLAG_NO_ACLS | \
78 WIMLIB_EXTRACT_FLAG_STRICT_ACLS | \
79 WIMLIB_EXTRACT_FLAG_RPFIX | \
80 WIMLIB_EXTRACT_FLAG_NORPFIX | \
81 WIMLIB_EXTRACT_FLAG_TO_STDOUT | \
82 WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES | \
83 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS | \
84 WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS | \
85 WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES | \
86 WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS | \
87 WIMLIB_EXTRACT_FLAG_GLOB_PATHS | \
88 WIMLIB_EXTRACT_FLAG_STRICT_GLOB | \
89 WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES | \
90 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE | \
91 WIMLIB_EXTRACT_FLAG_WIMBOOT)
93 /* Check whether the extraction of a dentry should be skipped completely. */
95 dentry_is_supported(struct wim_dentry *dentry,
96 const struct wim_features *supported_features)
98 struct wim_inode *inode = dentry->d_inode;
100 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
101 return supported_features->reparse_points ||
102 (inode_is_symlink(inode) &&
103 supported_features->symlink_reparse_points);
105 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
106 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
107 return supported_features->encrypted_directories != 0;
109 return supported_features->encrypted_files != 0;
115 #define PWM_ALLOW_WIM_HDR 0x00001
117 /* Read the header from a stream in a pipable WIM. */
119 read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte,
120 struct wim_resource_spec *rspec,
121 int flags, struct wim_header_disk *hdr_ret)
124 struct pwm_stream_hdr stream_hdr;
125 struct wim_header_disk pwm_hdr;
127 struct wim_reshdr reshdr;
130 ret = full_read(&pwm->in_fd, &buf.stream_hdr, sizeof(buf.stream_hdr));
134 if ((flags & PWM_ALLOW_WIM_HDR) &&
135 le64_to_cpu(buf.stream_hdr.magic) == PWM_MAGIC)
137 BUILD_BUG_ON(sizeof(buf.pwm_hdr) < sizeof(buf.stream_hdr));
138 ret = full_read(&pwm->in_fd, &buf.stream_hdr + 1,
139 sizeof(buf.pwm_hdr) - sizeof(buf.stream_hdr));
143 lte->resource_location = RESOURCE_NONEXISTENT;
144 memcpy(hdr_ret, &buf.pwm_hdr, sizeof(buf.pwm_hdr));
148 if (le64_to_cpu(buf.stream_hdr.magic) != PWM_STREAM_MAGIC) {
149 ERROR("Data read on pipe is invalid (expected stream header).");
150 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
153 copy_hash(lte->hash, buf.stream_hdr.hash);
155 reshdr.size_in_wim = 0;
156 reshdr.flags = le32_to_cpu(buf.stream_hdr.flags);
157 reshdr.offset_in_wim = pwm->in_fd.offset;
158 reshdr.uncompressed_size = le64_to_cpu(buf.stream_hdr.uncompressed_size);
159 wim_res_hdr_to_spec(&reshdr, pwm, rspec);
160 lte_bind_wim_resource_spec(lte, rspec);
161 lte->flags = rspec->flags;
162 lte->size = rspec->uncompressed_size;
163 lte->offset_in_res = 0;
167 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
172 load_streams_from_pipe(struct apply_ctx *ctx,
173 const struct read_stream_list_callbacks *cbs)
175 struct wim_lookup_table_entry *found_lte = NULL;
176 struct wim_resource_spec *rspec = NULL;
177 struct wim_lookup_table *lookup_table;
180 ret = WIMLIB_ERR_NOMEM;
181 found_lte = new_lookup_table_entry();
185 rspec = MALLOC(sizeof(struct wim_resource_spec));
189 lookup_table = ctx->wim->lookup_table;
190 memcpy(ctx->progress.extract.guid, ctx->wim->hdr.guid, WIM_GUID_LEN);
191 ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
192 ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
193 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
197 while (ctx->num_streams_remaining) {
198 struct wim_header_disk pwm_hdr;
199 struct wim_lookup_table_entry *needed_lte;
201 if (found_lte->resource_location != RESOURCE_NONEXISTENT)
202 lte_unbind_wim_resource_spec(found_lte);
203 ret = read_pwm_stream_header(ctx->wim, found_lte, rspec,
204 PWM_ALLOW_WIM_HDR, &pwm_hdr);
208 if ((found_lte->resource_location != RESOURCE_NONEXISTENT)
209 && !(found_lte->flags & WIM_RESHDR_FLAG_METADATA)
210 && (needed_lte = lookup_stream(lookup_table, found_lte->hash))
211 && (needed_lte->out_refcnt))
213 needed_lte->offset_in_res = found_lte->offset_in_res;
214 needed_lte->flags = found_lte->flags;
215 needed_lte->size = found_lte->size;
217 lte_unbind_wim_resource_spec(found_lte);
218 lte_bind_wim_resource_spec(needed_lte, rspec);
220 ret = (*cbs->begin_stream)(needed_lte,
221 cbs->begin_stream_ctx);
223 lte_unbind_wim_resource_spec(needed_lte);
227 ret = extract_stream(needed_lte, needed_lte->size,
229 cbs->consume_chunk_ctx);
231 ret = (*cbs->end_stream)(needed_lte, ret,
232 cbs->end_stream_ctx);
233 lte_unbind_wim_resource_spec(needed_lte);
236 ctx->num_streams_remaining--;
237 } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) {
238 ret = skip_wim_stream(found_lte);
242 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
243 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
245 if (part_number != ctx->progress.extract.part_number ||
246 total_parts != ctx->progress.extract.total_parts ||
247 memcmp(pwm_hdr.guid, ctx->progress.extract.guid,
250 ctx->progress.extract.part_number = part_number;
251 ctx->progress.extract.total_parts = total_parts;
252 memcpy(ctx->progress.extract.guid,
253 pwm_hdr.guid, WIM_GUID_LEN);
254 ret = extract_progress(ctx,
255 WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
263 if (found_lte->resource_location != RESOURCE_IN_WIM)
265 free_lookup_table_entry(found_lte);
269 /* Creates a temporary file opened for writing. The open file descriptor is
270 * returned in @fd_ret and its name is returned in @name_ret (dynamically
273 create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
280 name = ttempnam(NULL, T("wimlib"));
282 ERROR_WITH_ERRNO("Failed to create temporary filename");
283 return WIMLIB_ERR_NOMEM;
286 open_flags = O_WRONLY | O_CREAT | O_EXCL | O_BINARY;
288 open_flags |= _O_SHORT_LIVED;
290 raw_fd = topen(name, open_flags, 0600);
293 if (errno == EEXIST) {
297 ERROR_WITH_ERRNO("Failed to create temporary file "
300 return WIMLIB_ERR_OPEN;
303 filedes_init(fd_ret, raw_fd);
309 begin_extract_stream_wrapper(struct wim_lookup_table_entry *lte, void *_ctx)
311 struct apply_ctx *ctx = _ctx;
313 ctx->cur_stream = lte;
314 ctx->cur_stream_offset = 0;
316 if (unlikely(lte->out_refcnt > MAX_OPEN_STREAMS))
317 return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
319 return (*ctx->saved_cbs->begin_stream)(lte, ctx->saved_cbs->begin_stream_ctx);
323 extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx)
325 struct apply_ctx *ctx = _ctx;
326 union wimlib_progress_info *progress = &ctx->progress;
329 ctx->cur_stream_offset += size;
331 if (likely(ctx->supported_features.hard_links)) {
332 progress->extract.completed_bytes +=
333 (u64)size * ctx->cur_stream->out_refcnt;
334 if (ctx->cur_stream_offset == ctx->cur_stream->size)
335 progress->extract.completed_streams += ctx->cur_stream->out_refcnt;
337 const struct stream_owner *owners = stream_owners(ctx->cur_stream);
338 for (u32 i = 0; i < ctx->cur_stream->out_refcnt; i++) {
339 const struct wim_inode *inode = owners[i].inode;
340 const struct wim_dentry *dentry;
342 list_for_each_entry(dentry,
343 &inode->i_extraction_aliases,
344 d_extraction_alias_node)
346 progress->extract.completed_bytes += size;
347 if (ctx->cur_stream_offset == ctx->cur_stream->size)
348 progress->extract.completed_streams++;
352 if (progress->extract.completed_bytes >= ctx->next_progress) {
354 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
358 if (progress->extract.completed_bytes >=
359 progress->extract.total_bytes)
361 ctx->next_progress = UINT64_MAX;
363 /* Send new message as soon as another 1/128 of the
364 * total has been extracted. (Arbitrary number.) */
366 progress->extract.completed_bytes +
367 progress->extract.total_bytes / 128;
369 /* ... Unless that would be more than 5000000 bytes, in
370 * which case send the next after the next 5000000
371 * bytes. (Another arbitrary number.) */
372 if (progress->extract.completed_bytes + 5000000 <
375 progress->extract.completed_bytes + 5000000;
377 /* ... But always send a message as soon as we're
378 * completely done. */
379 if (progress->extract.total_bytes < ctx->next_progress)
380 ctx->next_progress = progress->extract.total_bytes;
384 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
385 /* Just extracting to temporary file for now. */
386 ret = full_write(&ctx->tmpfile_fd, chunk, size);
388 ERROR_WITH_ERRNO("Error writing data to "
389 "temporary file \"%"TS"\"",
394 return (*ctx->saved_cbs->consume_chunk)(chunk, size,
395 ctx->saved_cbs->consume_chunk_ctx);
400 extract_from_tmpfile(const tchar *tmpfile_name, struct apply_ctx *ctx)
402 struct wim_lookup_table_entry tmpfile_lte;
403 struct wim_lookup_table_entry *orig_lte = ctx->cur_stream;
404 const struct stream_owner *owners = stream_owners(orig_lte);
405 const struct read_stream_list_callbacks *cbs = ctx->saved_cbs;
408 /* Copy the stream's data from the temporary file to each of its
411 * This is executed only in the very uncommon case that a
412 * single-instance stream is being extracted to more than
413 * MAX_OPEN_STREAMS locations! */
415 memcpy(&tmpfile_lte, orig_lte, sizeof(struct wim_lookup_table_entry));
416 tmpfile_lte.resource_location = RESOURCE_IN_FILE_ON_DISK;
417 tmpfile_lte.file_on_disk = ctx->tmpfile_name;
418 tmpfile_lte.out_refcnt = 1;
420 for (u32 i = 0; i < orig_lte->out_refcnt; i++) {
421 tmpfile_lte.inline_stream_owners[0] = owners[i];
423 /* Note: it usually doesn't matter whether we pass the original
424 * stream entry to callbacks provided by the extraction backend
425 * as opposed to the tmpfile stream entry, since they shouldn't
426 * actually read data from the stream other than through the
427 * read_stream_prefix() call below. But for
428 * WIMLIB_EXTRACT_FLAG_WIMBOOT mode on Windows it does matter
429 * because it needs the original stream location in order to
430 * create the external backing reference. */
432 ret = (*cbs->begin_stream)(orig_lte, cbs->begin_stream_ctx);
436 /* Extra SHA-1 isn't necessary here, but it shouldn't hurt as
437 * this case is very rare anyway. */
438 ret = extract_stream(&tmpfile_lte, tmpfile_lte.size,
440 cbs->consume_chunk_ctx);
442 return (*cbs->end_stream)(orig_lte, ret,
443 cbs->end_stream_ctx);
449 end_extract_stream_wrapper(struct wim_lookup_table_entry *stream,
450 int status, void *_ctx)
452 struct apply_ctx *ctx = _ctx;
454 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
455 filedes_close(&ctx->tmpfile_fd);
457 status = extract_from_tmpfile(ctx->tmpfile_name, ctx);
458 filedes_invalidate(&ctx->tmpfile_fd);
459 tunlink(ctx->tmpfile_name);
460 FREE(ctx->tmpfile_name);
463 return (*ctx->saved_cbs->end_stream)(stream, status,
464 ctx->saved_cbs->end_stream_ctx);
469 * Read the list of single-instance streams to extract and feed their data into
470 * the specified callback functions.
472 * This handles checksumming each stream.
474 * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
476 * This also works if the WIM is being read from a pipe, whereas attempting to
477 * read streams directly (e.g. with read_full_stream_into_buf()) will not.
479 * This also will split up streams that will need to be extracted to more than
480 * MAX_OPEN_STREAMS locations, as measured by the 'out_refcnt' of each stream.
481 * Therefore, the apply_operations implementation need not worry about running
482 * out of file descriptors, unless it might open more than one file descriptor
483 * per nominal destination (e.g. Win32 currently might because the destination
484 * file system might not support hard links).
487 extract_stream_list(struct apply_ctx *ctx,
488 const struct read_stream_list_callbacks *cbs)
490 struct read_stream_list_callbacks wrapper_cbs = {
491 .begin_stream = begin_extract_stream_wrapper,
492 .begin_stream_ctx = ctx,
493 .consume_chunk = extract_chunk_wrapper,
494 .consume_chunk_ctx = ctx,
495 .end_stream = end_extract_stream_wrapper,
496 .end_stream_ctx = ctx,
498 ctx->saved_cbs = cbs;
499 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
500 return load_streams_from_pipe(ctx, &wrapper_cbs);
502 return read_stream_list(&ctx->stream_list,
503 offsetof(struct wim_lookup_table_entry,
505 &wrapper_cbs, VERIFY_STREAM_HASHES);
509 /* Extract a WIM dentry to standard output.
511 * This obviously doesn't make sense in all cases. We return an error if the
512 * dentry does not correspond to a regular file. Otherwise we extract the
513 * unnamed data stream only. */
515 extract_dentry_to_stdout(struct wim_dentry *dentry,
516 const struct wim_lookup_table *lookup_table)
518 struct wim_inode *inode = dentry->d_inode;
519 struct wim_lookup_table_entry *lte;
520 struct filedes _stdout;
522 if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
523 FILE_ATTRIBUTE_DIRECTORY))
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 lte = inode_unnamed_lte(inode, lookup_table);
532 const u8 *hash = inode_unnamed_stream_hash(inode);
533 if (!is_zero_hash(hash))
534 return stream_not_found_error(inode, hash);
538 filedes_init(&_stdout, STDOUT_FILENO);
539 return extract_full_stream_to_fd(lte, &_stdout);
543 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
544 const struct wim_lookup_table *lookup_table)
546 for (size_t i = 0; i < num_dentries; i++) {
547 int ret = extract_dentry_to_stdout(dentries[i], lookup_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]->tmp_flag) {
567 /* Found distinct dentry. */
568 trees[i]->tmp_flag = 1;
569 trees[j++] = trees[i];
572 for (i = 0; i < j; i++)
573 trees[i]->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]->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]->tmp_flag = 0;
602 for (i = 0; i < j; i++)
603 trees[i]->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 list_del(&dentry->d_extraction_list_node);
625 dentry_reset_extraction_list_node(dentry);
630 * Build the preliminary list of dentries to be extracted.
632 * The list maintains the invariant that if d1 and d2 are in the list and d1 is
633 * an ancestor of d2, then d1 appears before d2 in the list.
636 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
637 size_t num_trees, bool add_ancestors)
639 INIT_LIST_HEAD(dentry_list);
641 /* Add the trees recursively. */
642 for (size_t i = 0; i < num_trees; i++)
643 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
645 /* If requested, add ancestors of the trees. */
647 for (size_t i = 0; i < num_trees; i++) {
648 struct wim_dentry *dentry = trees[i];
649 struct wim_dentry *ancestor;
650 struct list_head *place_after;
652 if (dentry_is_root(dentry))
655 place_after = dentry_list;
658 ancestor = ancestor->d_parent;
659 if (will_extract_dentry(ancestor)) {
660 place_after = &ancestor->d_extraction_list_node;
663 } while (!dentry_is_root(ancestor));
667 ancestor = ancestor->d_parent;
668 if (will_extract_dentry(ancestor))
670 list_add(&ancestor->d_extraction_list_node, place_after);
671 } while (!dentry_is_root(ancestor));
677 destroy_dentry_list(struct list_head *dentry_list)
679 struct wim_dentry *dentry, *tmp;
680 struct wim_inode *inode;
682 list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
683 inode = dentry->d_inode;
684 dentry_reset_extraction_list_node(dentry);
685 inode->i_visited = 0;
686 if ((void *)dentry->d_extraction_name != (void *)dentry->file_name)
687 FREE(dentry->d_extraction_name);
688 dentry->d_extraction_name = NULL;
689 dentry->d_extraction_name_nchars = 0;
694 destroy_stream_list(struct list_head *stream_list)
696 struct wim_lookup_table_entry *lte;
698 list_for_each_entry(lte, stream_list, extraction_list)
699 if (lte->out_refcnt > ARRAY_LEN(lte->inline_stream_owners))
700 FREE(lte->stream_owners);
704 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
706 static const utf16lechar replacement_char = cpu_to_le16('?');
710 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
716 for (i = 0; i < num_chars; i++) {
719 case cpu_to_le16('\\'):
720 case cpu_to_le16(':'):
721 case cpu_to_le16('*'):
722 case cpu_to_le16('?'):
723 case cpu_to_le16('"'):
724 case cpu_to_le16('<'):
725 case cpu_to_le16('>'):
726 case cpu_to_le16('|'):
728 case cpu_to_le16('/'):
729 case cpu_to_le16('\0'):
731 name[i] = replacement_char;
738 if (name[num_chars - 1] == cpu_to_le16(' ') ||
739 name[num_chars - 1] == cpu_to_le16('.'))
742 name[num_chars - 1] = replacement_char;
751 dentry_calculate_extraction_name(struct wim_dentry *dentry,
752 struct apply_ctx *ctx)
756 if (!dentry_is_supported(dentry, &ctx->supported_features))
759 if (dentry_is_root(dentry))
763 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
764 dentry->d_extraction_name = dentry->file_name;
765 dentry->d_extraction_name_nchars = dentry->file_name_nbytes /
771 if (!ctx->supported_features.case_sensitive_filenames) {
772 struct wim_dentry *other;
773 list_for_each_entry(other, &dentry->d_ci_conflict_list,
776 if (will_extract_dentry(other)) {
777 if (ctx->extract_flags &
778 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
779 WARNING("\"%"TS"\" has the same "
780 "case-insensitive name as "
781 "\"%"TS"\"; extracting "
782 "dummy name instead",
783 dentry_full_path(dentry),
784 dentry_full_path(other));
787 WARNING("Not extracting \"%"TS"\": "
788 "has same case-insensitive "
790 dentry_full_path(dentry),
791 dentry_full_path(other));
798 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
799 ret = utf16le_get_tstr(dentry->file_name,
800 dentry->file_name_nbytes,
801 (const tchar **)&dentry->d_extraction_name,
802 &dentry->d_extraction_name_nchars);
803 dentry->d_extraction_name_nchars /= sizeof(tchar);
806 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
808 WARNING("\"%"TS"\" has an invalid filename "
809 "that is not supported on this platform; "
810 "extracting dummy name instead",
811 dentry_full_path(dentry));
814 WARNING("Not extracting \"%"TS"\": has an invalid filename "
815 "that is not supported on this platform",
816 dentry_full_path(dentry));
823 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
825 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
826 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
828 const tchar *tchar_name;
831 ret = utf16le_get_tstr(utf16_name_copy,
832 dentry->file_name_nbytes,
833 &tchar_name, &tchar_nchars);
837 tchar_nchars /= sizeof(tchar);
839 size_t fixed_name_num_chars = tchar_nchars;
840 tchar fixed_name[tchar_nchars + 50];
842 tmemcpy(fixed_name, tchar_name, tchar_nchars);
843 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
844 T(" (invalid filename #%lu)"),
845 ++ctx->invalid_sequence);
847 utf16le_put_tstr(tchar_name);
849 dentry->d_extraction_name = memdup(fixed_name,
850 2 * fixed_name_num_chars + 2);
851 if (!dentry->d_extraction_name)
852 return WIMLIB_ERR_NOMEM;
853 dentry->d_extraction_name_nchars = fixed_name_num_chars;
858 for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
863 * Calculate the actual filename component at which each WIM dentry will be
864 * extracted, with special handling for dentries that are unsupported by the
865 * extraction backend or have invalid names.
867 * ctx->supported_features must be filled in.
869 * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
872 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
873 struct apply_ctx *ctx)
875 struct list_head *prev, *cur;
877 /* Can't use list_for_each_entry() because a call to
878 * dentry_calculate_extraction_name() may delete the current dentry and
879 * its children from the list. */
883 struct wim_dentry *dentry;
887 if (cur == dentry_list)
890 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
892 ret = dentry_calculate_extraction_name(dentry, ctx);
896 if (prev->next == cur)
899 ; /* Current dentry and its children (which follow in
900 the list) were deleted. prev stays the same. */
906 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
907 struct wim_lookup_table *lookup_table)
909 struct wim_inode *inode = dentry->d_inode;
910 struct wim_lookup_table_entry *lte;
914 /* Special case: when extracting from a pipe, the WIM lookup table is
915 * initially empty, so "resolving" an inode's streams is initially not
916 * possible. However, we still need to keep track of which streams,
917 * identified by SHA1 message digests, need to be extracted, so we
918 * "resolve" the inode's streams anyway by allocating new entries. */
919 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
921 ret = inode_resolve_streams(inode, lookup_table, force);
924 for (u32 i = 0; i <= inode->i_num_ads; i++) {
925 lte = inode_stream_lte_resolved(inode, i);
933 * For each dentry to be extracted, resolve all streams in the corresponding
934 * inode and set 'out_refcnt' in each to 0.
936 * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
939 dentry_list_resolve_streams(struct list_head *dentry_list,
940 struct apply_ctx *ctx)
942 struct wim_dentry *dentry;
945 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
946 ret = dentry_resolve_streams(dentry,
948 ctx->wim->lookup_table);
956 ref_stream(struct wim_lookup_table_entry *lte, u32 stream_idx,
957 struct wim_dentry *dentry, struct apply_ctx *ctx)
959 struct wim_inode *inode = dentry->d_inode;
960 struct stream_owner *stream_owners;
965 /* Tally the size only for each extraction of the stream (not hard
967 if (inode->i_visited && ctx->supported_features.hard_links)
970 ctx->progress.extract.total_bytes += lte->size;
971 ctx->progress.extract.total_streams++;
973 if (inode->i_visited)
976 /* Add stream to the dentry_list only one time, even if it's going
977 * to be extracted to multiple inodes. */
978 if (lte->out_refcnt == 0) {
979 list_add_tail(<e->extraction_list, &ctx->stream_list);
980 ctx->num_streams_remaining++;
983 /* If inode not yet been visited, append it to the stream_owners array. */
984 if (lte->out_refcnt < ARRAY_LEN(lte->inline_stream_owners)) {
985 stream_owners = lte->inline_stream_owners;
987 struct stream_owner *prev_stream_owners;
988 size_t alloc_stream_owners;
990 if (lte->out_refcnt == ARRAY_LEN(lte->inline_stream_owners)) {
991 prev_stream_owners = NULL;
992 alloc_stream_owners = ARRAY_LEN(lte->inline_stream_owners);
994 prev_stream_owners = lte->stream_owners;
995 alloc_stream_owners = lte->alloc_stream_owners;
998 if (lte->out_refcnt == alloc_stream_owners) {
999 alloc_stream_owners *= 2;
1000 stream_owners = REALLOC(prev_stream_owners,
1001 alloc_stream_owners *
1002 sizeof(stream_owners[0]));
1004 return WIMLIB_ERR_NOMEM;
1005 if (!prev_stream_owners) {
1006 memcpy(stream_owners,
1007 lte->inline_stream_owners,
1008 sizeof(lte->inline_stream_owners));
1010 lte->stream_owners = stream_owners;
1011 lte->alloc_stream_owners = alloc_stream_owners;
1013 stream_owners = lte->stream_owners;
1015 stream_owners[lte->out_refcnt].inode = inode;
1016 if (stream_idx == 0) {
1017 stream_owners[lte->out_refcnt].stream_name = NULL;
1019 stream_owners[lte->out_refcnt].stream_name =
1020 inode->i_ads_entries[stream_idx - 1].stream_name;
1027 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1029 struct wim_inode *inode = dentry->d_inode;
1032 /* The unnamed data stream will always be extracted, except in an
1034 if (!inode_is_encrypted_directory(inode)) {
1036 struct wim_lookup_table_entry *stream;
1038 stream = inode_unnamed_stream_resolved(inode, &stream_idx);
1039 ret = ref_stream(stream, stream_idx, dentry, ctx);
1044 /* Named data streams will be extracted only if supported in the current
1045 * extraction mode and volume, and to avoid complications, if not doing
1046 * a linked extraction. */
1047 if (ctx->supported_features.named_data_streams) {
1048 for (u16 i = 0; i < inode->i_num_ads; i++) {
1049 if (!ads_entry_is_named_stream(&inode->i_ads_entries[i]))
1051 ret = ref_stream(inode->i_ads_entries[i].lte, i + 1,
1057 inode->i_visited = 1;
1062 * For each dentry to be extracted, iterate through the data streams of the
1063 * corresponding inode. For each such stream that is not to be ignored due to
1064 * the supported features or extraction flags, add it to the list of streams to
1065 * be extracted (ctx->stream_list) if not already done so.
1067 * Also builds a mapping from each stream to the inodes referencing it.
1069 * This also initializes the extract progress info with byte and stream
1072 * ctx->supported_features must be filled in.
1074 * Possible error codes: WIMLIB_ERR_NOMEM.
1077 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1079 struct wim_dentry *dentry;
1082 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1083 ret = dentry_ref_streams(dentry, ctx);
1087 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1088 dentry->d_inode->i_visited = 0;
1093 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1095 struct wim_dentry *dentry;
1096 struct wim_inode *inode;
1098 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1099 inode = dentry->d_inode;
1100 if (!inode->i_visited)
1101 INIT_LIST_HEAD(&inode->i_extraction_aliases);
1102 list_add_tail(&dentry->d_extraction_alias_node,
1103 &inode->i_extraction_aliases);
1104 inode->i_visited = 1;
1106 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1107 dentry->d_inode->i_visited = 0;
1111 inode_tally_features(const struct wim_inode *inode,
1112 struct wim_features *features)
1114 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1115 features->archive_files++;
1116 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1117 features->hidden_files++;
1118 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1119 features->system_files++;
1120 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1121 features->compressed_files++;
1122 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1123 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1124 features->encrypted_directories++;
1126 features->encrypted_files++;
1128 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1129 features->not_context_indexed_files++;
1130 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1131 features->sparse_files++;
1132 if (inode_has_named_stream(inode))
1133 features->named_data_streams++;
1134 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1135 features->reparse_points++;
1136 if (inode_is_symlink(inode))
1137 features->symlink_reparse_points++;
1139 features->other_reparse_points++;
1141 if (inode->i_security_id != -1)
1142 features->security_descriptors++;
1143 if (inode_has_unix_data(inode))
1144 features->unix_data++;
1147 /* Tally features necessary to extract a dentry and the corresponding inode. */
1149 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1151 struct wim_inode *inode = dentry->d_inode;
1153 if (dentry_has_short_name(dentry))
1154 features->short_names++;
1156 if (inode->i_visited) {
1157 features->hard_links++;
1159 inode_tally_features(inode, features);
1160 inode->i_visited = 1;
1164 /* Tally the features necessary to extract the specified dentries. */
1166 dentry_list_get_features(struct list_head *dentry_list,
1167 struct wim_features *features)
1169 struct wim_dentry *dentry;
1171 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1172 dentry_tally_features(dentry, features);
1174 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1175 dentry->d_inode->i_visited = 0;
1179 do_feature_check(const struct wim_features *required_features,
1180 const struct wim_features *supported_features,
1183 /* File attributes. */
1184 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1185 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1186 * We're an archive program, so theoretically we can do what we
1189 if (required_features->hidden_files &&
1190 !supported_features->hidden_files)
1191 WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1192 required_features->hidden_files);
1194 if (required_features->system_files &&
1195 !supported_features->system_files)
1196 WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1197 required_features->system_files);
1199 if (required_features->compressed_files &&
1200 !supported_features->compressed_files)
1201 WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1202 required_features->compressed_files);
1204 if (required_features->not_context_indexed_files &&
1205 !supported_features->not_context_indexed_files)
1206 WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1207 required_features->not_context_indexed_files);
1209 if (required_features->sparse_files &&
1210 !supported_features->sparse_files)
1211 WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1212 required_features->sparse_files);
1214 if (required_features->encrypted_directories &&
1215 !supported_features->encrypted_directories)
1216 WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1217 required_features->encrypted_directories);
1220 /* Encrypted files. */
1221 if (required_features->encrypted_files &&
1222 !supported_features->encrypted_files)
1223 WARNING("Ignoring %lu encrypted files",
1224 required_features->encrypted_files);
1226 /* Named data streams. */
1227 if (required_features->named_data_streams &&
1228 (!supported_features->named_data_streams))
1229 WARNING("Ignoring named data streams of %lu files",
1230 required_features->named_data_streams);
1233 if (required_features->hard_links && !supported_features->hard_links)
1234 WARNING("Extracting %lu hard links as independent files",
1235 required_features->hard_links);
1237 /* Symbolic links and reparse points. */
1238 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1239 required_features->symlink_reparse_points &&
1240 !supported_features->symlink_reparse_points &&
1241 !supported_features->reparse_points)
1243 ERROR("Extraction backend does not support symbolic links!");
1244 return WIMLIB_ERR_UNSUPPORTED;
1246 if (required_features->reparse_points &&
1247 !supported_features->reparse_points)
1249 if (supported_features->symlink_reparse_points) {
1250 if (required_features->other_reparse_points) {
1251 WARNING("Ignoring %lu non-symlink/junction "
1252 "reparse point files",
1253 required_features->other_reparse_points);
1256 WARNING("Ignoring %lu reparse point files",
1257 required_features->reparse_points);
1261 /* Security descriptors. */
1262 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1263 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1264 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1265 required_features->security_descriptors &&
1266 !supported_features->security_descriptors)
1268 ERROR("Extraction backend does not support security descriptors!");
1269 return WIMLIB_ERR_UNSUPPORTED;
1271 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1272 required_features->security_descriptors &&
1273 !supported_features->security_descriptors)
1274 WARNING("Ignoring Windows NT security descriptors of %lu files",
1275 required_features->security_descriptors);
1278 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1279 required_features->unix_data && !supported_features->unix_data)
1281 ERROR("Extraction backend does not support UNIX data!");
1282 return WIMLIB_ERR_UNSUPPORTED;
1285 if (required_features->unix_data &&
1286 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1288 WARNING("Ignoring UNIX metadata of %lu files",
1289 required_features->unix_data);
1293 if (required_features->short_names &&
1294 !supported_features->short_names)
1296 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1297 ERROR("Extraction backend does not support DOS names!");
1298 return WIMLIB_ERR_UNSUPPORTED;
1300 WARNING("Ignoring DOS names of %lu files",
1301 required_features->short_names);
1305 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1306 !supported_features->timestamps)
1308 ERROR("Extraction backend does not support timestamps!");
1309 return WIMLIB_ERR_UNSUPPORTED;
1315 static const struct apply_operations *
1316 select_apply_operations(int extract_flags)
1319 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1320 return &ntfs_3g_apply_ops;
1323 return &win32_apply_ops;
1325 return &unix_apply_ops;
1330 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1331 const tchar *target, int extract_flags)
1333 const struct apply_operations *ops;
1334 struct apply_ctx *ctx;
1336 LIST_HEAD(dentry_list);
1338 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1339 ret = extract_dentries_to_stdout(trees, num_trees,
1344 num_trees = remove_duplicate_trees(trees, num_trees);
1345 num_trees = remove_contained_trees(trees, num_trees);
1347 ops = select_apply_operations(extract_flags);
1349 if (num_trees > 1 && ops->single_tree_only) {
1350 ERROR("Extracting multiple directory trees "
1351 "at once is not supported in %s extraction mode!",
1353 ret = WIMLIB_ERR_UNSUPPORTED;
1357 ctx = CALLOC(1, ops->context_size);
1359 ret = WIMLIB_ERR_NOMEM;
1364 ctx->target = target;
1365 ctx->target_nchars = tstrlen(target);
1366 ctx->extract_flags = extract_flags;
1367 if (ctx->wim->progfunc) {
1368 ctx->progfunc = ctx->wim->progfunc;
1369 ctx->progctx = ctx->wim->progctx;
1370 ctx->progress.extract.image = wim->current_image;
1371 ctx->progress.extract.extract_flags = (extract_flags &
1372 WIMLIB_EXTRACT_MASK_PUBLIC);
1373 ctx->progress.extract.wimfile_name = wim->filename;
1374 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1375 wim->current_image);
1376 ctx->progress.extract.target = target;
1378 INIT_LIST_HEAD(&ctx->stream_list);
1379 filedes_invalidate(&ctx->tmpfile_fd);
1381 ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1385 build_dentry_list(&dentry_list, trees, num_trees,
1387 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1389 dentry_list_get_features(&dentry_list, &ctx->required_features);
1391 ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1392 ctx->extract_flags);
1396 ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1400 ret = dentry_list_resolve_streams(&dentry_list, ctx);
1404 ret = dentry_list_ref_streams(&dentry_list, ctx);
1408 dentry_list_build_inode_alias_lists(&dentry_list);
1410 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1411 /* When extracting from a pipe, the number of bytes of data to
1412 * extract can't be determined in the normal way (examining the
1413 * lookup table), since at this point all we have is a set of
1414 * SHA1 message digests of streams that need to be extracted.
1415 * However, we can get a reasonably accurate estimate by taking
1416 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1417 * data. This does assume that a full image is being extracted,
1418 * but currently there is no API for doing otherwise. (Also,
1419 * subtract <HARDLINKBYTES> from this if hard links are
1420 * supported by the extraction mode.) */
1421 ctx->progress.extract.total_bytes =
1422 wim_info_get_image_total_bytes(wim->wim_info,
1423 wim->current_image);
1424 if (ctx->supported_features.hard_links) {
1425 ctx->progress.extract.total_bytes -=
1426 wim_info_get_image_hard_link_bytes(wim->wim_info,
1427 wim->current_image);
1431 ret = extract_progress(ctx,
1432 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1433 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1434 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1438 ret = (*ops->extract)(&dentry_list, ctx);
1442 if (ctx->progress.extract.completed_bytes <
1443 ctx->progress.extract.total_bytes)
1445 ctx->progress.extract.completed_bytes =
1446 ctx->progress.extract.total_bytes;
1447 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1452 ret = extract_progress(ctx,
1453 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1454 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1455 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1457 destroy_stream_list(&ctx->stream_list);
1458 destroy_dentry_list(&dentry_list);
1465 mkdir_if_needed(const tchar *target)
1468 if (tstat(target, &stbuf)) {
1469 if (errno == ENOENT) {
1470 if (tmkdir(target, 0755)) {
1471 ERROR_WITH_ERRNO("Failed to create directory "
1472 "\"%"TS"\"", target);
1473 return WIMLIB_ERR_MKDIR;
1476 ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target);
1477 return WIMLIB_ERR_STAT;
1479 } else if (!S_ISDIR(stbuf.st_mode)) {
1480 ERROR("\"%"TS"\" is not a directory", target);
1481 return WIMLIB_ERR_NOTDIR;
1486 /* Make sure the extraction flags make sense, and update them if needed. */
1488 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1490 int extract_flags = *extract_flags_p;
1492 /* Check for invalid flag combinations */
1494 if ((extract_flags &
1495 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1496 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1497 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1498 return WIMLIB_ERR_INVALID_PARAM;
1500 if ((extract_flags &
1501 (WIMLIB_EXTRACT_FLAG_RPFIX |
1502 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1503 WIMLIB_EXTRACT_FLAG_NORPFIX))
1504 return WIMLIB_ERR_INVALID_PARAM;
1506 #ifndef WITH_NTFS_3G
1507 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1508 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
1509 " it cannot apply a WIM image directly to an NTFS volume.");
1510 return WIMLIB_ERR_UNSUPPORTED;
1515 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1516 ERROR("WIMBoot extraction is only supported on Windows!");
1517 return WIMLIB_ERR_UNSUPPORTED;
1521 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1522 WIMLIB_EXTRACT_FLAG_NORPFIX |
1523 WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1524 WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1526 /* For full-image extraction, do reparse point fixups by default
1527 * if the WIM header says they are enabled. */
1528 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1529 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1532 *extract_flags_p = extract_flags;
1537 get_wildcard_flags(int extract_flags)
1539 u32 wildcard_flags = 0;
1541 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB)
1542 wildcard_flags |= WILDCARD_FLAG_ERROR_IF_NO_MATCH;
1544 wildcard_flags |= WILDCARD_FLAG_WARN_IF_NO_MATCH;
1546 if (default_ignore_case)
1547 wildcard_flags |= WILDCARD_FLAG_CASE_INSENSITIVE;
1549 return wildcard_flags;
1552 struct append_dentry_ctx {
1553 struct wim_dentry **dentries;
1554 size_t num_dentries;
1555 size_t num_alloc_dentries;
1559 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1561 struct append_dentry_ctx *ctx = _ctx;
1563 if (ctx->num_dentries == ctx->num_alloc_dentries) {
1564 struct wim_dentry **new_dentries;
1567 new_length = max(ctx->num_alloc_dentries + 8,
1568 ctx->num_alloc_dentries * 3 / 2);
1569 new_dentries = REALLOC(ctx->dentries,
1570 new_length * sizeof(ctx->dentries[0]));
1571 if (new_dentries == NULL)
1572 return WIMLIB_ERR_NOMEM;
1573 ctx->dentries = new_dentries;
1574 ctx->num_alloc_dentries = new_length;
1576 ctx->dentries[ctx->num_dentries++] = dentry;
1581 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1582 const tchar * const *paths, size_t num_paths,
1586 struct wim_dentry **trees;
1589 if (wim == NULL || target == NULL || target[0] == T('\0') ||
1590 (num_paths != 0 && paths == NULL))
1591 return WIMLIB_ERR_INVALID_PARAM;
1593 ret = check_extract_flags(wim, &extract_flags);
1597 ret = select_wim_image(wim, image);
1601 ret = wim_checksum_unhashed_streams(wim);
1605 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1606 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1607 (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1609 ret = mkdir_if_needed(target);
1614 if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1616 struct append_dentry_ctx append_dentry_ctx = {
1619 .num_alloc_dentries = 0,
1622 u32 wildcard_flags = get_wildcard_flags(extract_flags);
1624 for (size_t i = 0; i < num_paths; i++) {
1625 tchar *path = canonicalize_wim_path(paths[i]);
1627 ret = WIMLIB_ERR_NOMEM;
1628 trees = append_dentry_ctx.dentries;
1629 goto out_free_trees;
1631 ret = expand_wildcard(wim, path,
1637 trees = append_dentry_ctx.dentries;
1638 goto out_free_trees;
1641 trees = append_dentry_ctx.dentries;
1642 num_trees = append_dentry_ctx.num_dentries;
1644 trees = MALLOC(num_paths * sizeof(trees[0]));
1646 return WIMLIB_ERR_NOMEM;
1648 for (size_t i = 0; i < num_paths; i++) {
1650 tchar *path = canonicalize_wim_path(paths[i]);
1652 ret = WIMLIB_ERR_NOMEM;
1653 goto out_free_trees;
1656 trees[i] = get_dentry(wim, path,
1657 WIMLIB_CASE_PLATFORM_DEFAULT);
1659 if (trees[i] == NULL) {
1660 ERROR("Path \"%"TS"\" does not exist "
1662 paths[i], wim->current_image);
1663 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1664 goto out_free_trees;
1667 num_trees = num_paths;
1670 if (num_trees == 0) {
1672 goto out_free_trees;
1675 ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1682 extract_single_image(WIMStruct *wim, int image,
1683 const tchar *target, int extract_flags)
1685 const tchar *path = WIMLIB_WIM_ROOT_PATH;
1686 extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1687 return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1690 static const tchar * const filename_forbidden_chars =
1699 /* This function checks if it is okay to use a WIM image's name as a directory
1702 image_name_ok_as_dir(const tchar *image_name)
1704 return image_name && *image_name &&
1705 !tstrpbrk(image_name, filename_forbidden_chars) &&
1706 tstrcmp(image_name, T(".")) &&
1707 tstrcmp(image_name, T(".."));
1710 /* Extracts all images from the WIM to the directory @target, with the images
1711 * placed in subdirectories named by their image names. */
1713 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1715 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
1716 size_t output_path_len = tstrlen(target);
1717 tchar buf[output_path_len + 1 + image_name_max_len + 1];
1720 const tchar *image_name;
1722 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1723 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1724 return WIMLIB_ERR_INVALID_PARAM;
1727 ret = mkdir_if_needed(target);
1730 tmemcpy(buf, target, output_path_len);
1731 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1732 for (image = 1; image <= wim->hdr.image_count; image++) {
1733 image_name = wimlib_get_image_name(wim, image);
1734 if (image_name_ok_as_dir(image_name)) {
1735 tstrcpy(buf + output_path_len + 1, image_name);
1737 /* Image name is empty or contains forbidden characters.
1738 * Use image number instead. */
1739 tsprintf(buf + output_path_len + 1, T("%d"), image);
1741 ret = extract_single_image(wim, image, buf, extract_flags);
1749 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1752 if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1753 WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1754 WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1755 return WIMLIB_ERR_INVALID_PARAM;
1757 if (image == WIMLIB_ALL_IMAGES)
1758 return extract_all_images(wim, target, extract_flags);
1760 return extract_single_image(wim, image, target, extract_flags);
1764 /****************************************************************************
1766 ****************************************************************************/
1769 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1770 const tchar * const *paths, size_t num_paths,
1773 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1774 return WIMLIB_ERR_INVALID_PARAM;
1776 return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1781 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1782 const tchar *path_list_file, int extract_flags)
1789 ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1791 ERROR("Failed to read path list file \"%"TS"\"",
1796 ret = wimlib_extract_paths(wim, image, target,
1797 (const tchar * const *)paths, num_paths,
1805 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1806 const tchar *image_num_or_name,
1807 const tchar *target,
1809 wimlib_progress_func_t progfunc,
1814 struct filedes *in_fd;
1818 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1819 return WIMLIB_ERR_INVALID_PARAM;
1821 /* Read the WIM header from the pipe and get a WIMStruct to represent
1822 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
1823 * wimlib_open_wim(), getting a WIMStruct in this way will result in
1824 * an empty lookup table, no XML data read, and no filename set. */
1825 ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1830 /* Sanity check to make sure this is a pipable WIM. */
1831 if (pwm->hdr.magic != PWM_MAGIC) {
1832 ERROR("The WIM being read from file descriptor %d "
1833 "is not pipable!", pipe_fd);
1834 ret = WIMLIB_ERR_NOT_PIPABLE;
1835 goto out_wimlib_free;
1838 /* Sanity check to make sure the first part of a pipable split WIM is
1839 * sent over the pipe first. */
1840 if (pwm->hdr.part_number != 1) {
1841 ERROR("The first part of the split WIM must be "
1842 "sent over the pipe first.");
1843 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1844 goto out_wimlib_free;
1847 in_fd = &pwm->in_fd;
1848 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1850 /* As mentioned, the WIMStruct we created from the pipe does not have
1851 * XML data yet. Fix this by reading the extra copy of the XML data
1852 * that directly follows the header in pipable WIMs. (Note: see
1853 * write_pipable_wim() for more details about the format of pipable
1856 struct wim_lookup_table_entry xml_lte;
1857 struct wim_resource_spec xml_rspec;
1858 ret = read_pwm_stream_header(pwm, &xml_lte, &xml_rspec, 0, NULL);
1860 goto out_wimlib_free;
1862 if (!(xml_lte.flags & WIM_RESHDR_FLAG_METADATA))
1864 ERROR("Expected XML data, but found non-metadata "
1866 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1867 goto out_wimlib_free;
1870 wim_res_spec_to_hdr(&xml_rspec, &pwm->hdr.xml_data_reshdr);
1872 ret = read_wim_xml_data(pwm);
1874 goto out_wimlib_free;
1876 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
1877 ERROR("Image count in XML data is not the same as in WIM header.");
1878 ret = WIMLIB_ERR_IMAGE_COUNT;
1879 goto out_wimlib_free;
1883 /* Get image index (this may use the XML data that was just read to
1884 * resolve an image name). */
1885 if (image_num_or_name) {
1886 image = wimlib_resolve_image(pwm, image_num_or_name);
1887 if (image == WIMLIB_NO_IMAGE) {
1888 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
1890 ret = WIMLIB_ERR_INVALID_IMAGE;
1891 goto out_wimlib_free;
1892 } else if (image == WIMLIB_ALL_IMAGES) {
1893 ERROR("Applying all images from a pipe is not supported!");
1894 ret = WIMLIB_ERR_INVALID_IMAGE;
1895 goto out_wimlib_free;
1898 if (pwm->hdr.image_count != 1) {
1899 ERROR("No image was specified, but the pipable WIM "
1900 "did not contain exactly 1 image");
1901 ret = WIMLIB_ERR_INVALID_IMAGE;
1902 goto out_wimlib_free;
1907 /* Load the needed metadata resource. */
1908 for (i = 1; i <= pwm->hdr.image_count; i++) {
1909 struct wim_lookup_table_entry *metadata_lte;
1910 struct wim_image_metadata *imd;
1911 struct wim_resource_spec *metadata_rspec;
1913 metadata_lte = new_lookup_table_entry();
1914 if (metadata_lte == NULL) {
1915 ret = WIMLIB_ERR_NOMEM;
1916 goto out_wimlib_free;
1918 metadata_rspec = MALLOC(sizeof(struct wim_resource_spec));
1919 if (metadata_rspec == NULL) {
1920 ret = WIMLIB_ERR_NOMEM;
1921 free_lookup_table_entry(metadata_lte);
1922 goto out_wimlib_free;
1925 ret = read_pwm_stream_header(pwm, metadata_lte, metadata_rspec, 0, NULL);
1926 imd = pwm->image_metadata[i - 1];
1927 imd->metadata_lte = metadata_lte;
1929 FREE(metadata_rspec);
1930 goto out_wimlib_free;
1933 if (!(metadata_lte->flags & WIM_RESHDR_FLAG_METADATA)) {
1934 ERROR("Expected metadata resource, but found "
1935 "non-metadata stream.");
1936 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1937 goto out_wimlib_free;
1941 /* Metadata resource is for the image being extracted.
1942 * Parse it and save the metadata in memory. */
1943 ret = read_metadata_resource(pwm, imd);
1945 goto out_wimlib_free;
1948 /* Metadata resource is not for the image being
1949 * extracted. Skip over it. */
1950 ret = skip_wim_stream(metadata_lte);
1952 goto out_wimlib_free;
1955 /* Extract the image. */
1956 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
1957 ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
1958 /* Clean up and return. */
1966 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
1967 const tchar *target, int extract_flags)
1969 return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
1978 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1981 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1982 return WIMLIB_ERR_INVALID_PARAM;
1983 return do_wimlib_extract_image(wim, image, target, extract_flags);