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, 0,
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,
310 u32 flags, void *_ctx)
312 struct apply_ctx *ctx = _ctx;
314 ctx->cur_stream = lte;
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, flags,
320 ctx->saved_cbs->begin_stream_ctx);
324 extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx)
326 struct apply_ctx *ctx = _ctx;
327 union wimlib_progress_info *progress = &ctx->progress;
330 if (likely(ctx->supported_features.hard_links)) {
331 progress->extract.completed_bytes +=
332 (u64)size * ctx->cur_stream->out_refcnt;
334 const struct stream_owner *owners = stream_owners(ctx->cur_stream);
335 for (u32 i = 0; i < ctx->cur_stream->out_refcnt; i++) {
336 const struct wim_inode *inode = owners[i].inode;
337 const struct wim_dentry *dentry;
339 list_for_each_entry(dentry,
340 &inode->i_extraction_aliases,
341 d_extraction_alias_node)
343 progress->extract.completed_bytes += size;
347 if (progress->extract.completed_bytes >= ctx->next_progress) {
349 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
353 if (progress->extract.completed_bytes >=
354 progress->extract.total_bytes)
356 ctx->next_progress = UINT64_MAX;
358 /* Send new message as soon as another 1/128 of the
359 * total has been extracted. (Arbitrary number.) */
361 progress->extract.completed_bytes +
362 progress->extract.total_bytes / 128;
364 /* ... Unless that would be more than 5000000 bytes, in
365 * which case send the next after the next 5000000
366 * bytes. (Another arbitrary number.) */
367 if (progress->extract.completed_bytes + 5000000 <
370 progress->extract.completed_bytes + 5000000;
372 /* ... But always send a message as soon as we're
373 * completely done. */
374 if (progress->extract.total_bytes < ctx->next_progress)
375 ctx->next_progress = progress->extract.total_bytes;
379 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
380 /* Just extracting to temporary file for now. */
381 ret = full_write(&ctx->tmpfile_fd, chunk, size);
383 ERROR_WITH_ERRNO("Error writing data to "
384 "temporary file \"%"TS"\"",
389 return (*ctx->saved_cbs->consume_chunk)(chunk, size,
390 ctx->saved_cbs->consume_chunk_ctx);
395 extract_from_tmpfile(const tchar *tmpfile_name, struct apply_ctx *ctx)
397 struct wim_lookup_table_entry tmpfile_lte;
398 struct wim_lookup_table_entry *orig_lte = ctx->cur_stream;
399 const struct stream_owner *owners = stream_owners(orig_lte);
400 const struct read_stream_list_callbacks *cbs = ctx->saved_cbs;
403 /* Copy the stream's data from the temporary file to each of its
406 * This is executed only in the very uncommon case that a
407 * single-instance stream is being extracted to more than
408 * MAX_OPEN_STREAMS locations! */
410 memcpy(&tmpfile_lte, orig_lte, sizeof(struct wim_lookup_table_entry));
411 tmpfile_lte.resource_location = RESOURCE_IN_FILE_ON_DISK;
412 tmpfile_lte.file_on_disk = ctx->tmpfile_name;
413 tmpfile_lte.out_refcnt = 1;
415 for (u32 i = 0; i < orig_lte->out_refcnt; i++) {
416 tmpfile_lte.inline_stream_owners[0] = owners[i];
418 /* Note: it usually doesn't matter whether we pass the original
419 * stream entry to callbacks provided by the extraction backend
420 * as opposed to the tmpfile stream entry, since they shouldn't
421 * actually read data from the stream other than through the
422 * read_stream_prefix() call below. But for
423 * WIMLIB_EXTRACT_FLAG_WIMBOOT mode on Windows it does matter
424 * because it needs the original stream location in order to
425 * create the external backing reference. */
427 ret = (*cbs->begin_stream)(orig_lte, 0,
428 cbs->begin_stream_ctx);
432 /* Extra SHA-1 isn't necessary here, but it shouldn't hurt as
433 * this case is very rare anyway. */
434 ret = extract_stream(&tmpfile_lte, tmpfile_lte.size,
436 cbs->consume_chunk_ctx);
438 return (*cbs->end_stream)(orig_lte, ret,
439 cbs->end_stream_ctx);
445 end_extract_stream_wrapper(struct wim_lookup_table_entry *stream,
446 int status, void *_ctx)
448 struct apply_ctx *ctx = _ctx;
450 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
451 filedes_close(&ctx->tmpfile_fd);
453 status = extract_from_tmpfile(ctx->tmpfile_name, ctx);
454 filedes_invalidate(&ctx->tmpfile_fd);
455 tunlink(ctx->tmpfile_name);
456 FREE(ctx->tmpfile_name);
459 return (*ctx->saved_cbs->end_stream)(stream, status,
460 ctx->saved_cbs->end_stream_ctx);
465 * Read the list of single-instance streams to extract and feed their data into
466 * the specified callback functions.
468 * This handles checksumming each stream.
470 * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
472 * This also works if the WIM is being read from a pipe, whereas attempting to
473 * read streams directly (e.g. with read_full_stream_into_buf()) will not.
475 * This also will split up streams that will need to be extracted to more than
476 * MAX_OPEN_STREAMS locations, as measured by the 'out_refcnt' of each stream.
477 * Therefore, the apply_operations implementation need not worry about running
478 * out of file descriptors, unless it might open more than one file descriptor
479 * per nominal destination (e.g. Win32 currently might because the destination
480 * file system might not support hard links).
483 extract_stream_list(struct apply_ctx *ctx,
484 const struct read_stream_list_callbacks *cbs)
486 struct read_stream_list_callbacks wrapper_cbs = {
487 .begin_stream = begin_extract_stream_wrapper,
488 .begin_stream_ctx = ctx,
489 .consume_chunk = extract_chunk_wrapper,
490 .consume_chunk_ctx = ctx,
491 .end_stream = end_extract_stream_wrapper,
492 .end_stream_ctx = ctx,
494 ctx->saved_cbs = cbs;
495 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
496 return load_streams_from_pipe(ctx, &wrapper_cbs);
498 return read_stream_list(&ctx->stream_list,
499 offsetof(struct wim_lookup_table_entry,
501 &wrapper_cbs, VERIFY_STREAM_HASHES);
505 /* Extract a WIM dentry to standard output.
507 * This obviously doesn't make sense in all cases. We return an error if the
508 * dentry does not correspond to a regular file. Otherwise we extract the
509 * unnamed data stream only. */
511 extract_dentry_to_stdout(struct wim_dentry *dentry,
512 const struct wim_lookup_table *lookup_table)
514 struct wim_inode *inode = dentry->d_inode;
515 struct wim_lookup_table_entry *lte;
516 struct filedes _stdout;
518 if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
519 FILE_ATTRIBUTE_DIRECTORY))
521 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
522 "extracted to standard output", dentry_full_path(dentry));
523 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
526 lte = inode_unnamed_lte(inode, lookup_table);
528 const u8 *hash = inode_unnamed_stream_hash(inode);
529 if (!is_zero_hash(hash))
530 return stream_not_found_error(inode, hash);
534 filedes_init(&_stdout, STDOUT_FILENO);
535 return extract_full_stream_to_fd(lte, &_stdout);
539 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
540 const struct wim_lookup_table *lookup_table)
542 for (size_t i = 0; i < num_dentries; i++) {
543 int ret = extract_dentry_to_stdout(dentries[i], lookup_table);
550 /**********************************************************************/
553 * Removes duplicate dentries from the array.
555 * Returns the new number of dentries, packed at the front of the array.
558 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
561 for (i = 0; i < num_trees; i++) {
562 if (!trees[i]->tmp_flag) {
563 /* Found distinct dentry. */
564 trees[i]->tmp_flag = 1;
565 trees[j++] = trees[i];
568 for (i = 0; i < j; i++)
569 trees[i]->tmp_flag = 0;
574 * Remove dentries that are descendants of other dentries in the array.
576 * Returns the new number of dentries, packed at the front of the array.
579 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
582 for (i = 0; i < num_trees; i++)
583 trees[i]->tmp_flag = 1;
584 for (i = 0; i < num_trees; i++) {
585 struct wim_dentry *d = trees[i];
586 while (!dentry_is_root(d)) {
591 trees[j++] = trees[i];
595 trees[i]->tmp_flag = 0;
598 for (i = 0; i < j; i++)
599 trees[i]->tmp_flag = 0;
604 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
606 struct list_head *dentry_list = _dentry_list;
607 list_add_tail(&dentry->d_extraction_list_node, dentry_list);
612 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
614 dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
618 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
620 list_del(&dentry->d_extraction_list_node);
621 dentry_reset_extraction_list_node(dentry);
626 * Build the preliminary list of dentries to be extracted.
628 * The list maintains the invariant that if d1 and d2 are in the list and d1 is
629 * an ancestor of d2, then d1 appears before d2 in the list.
632 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
633 size_t num_trees, bool add_ancestors)
635 INIT_LIST_HEAD(dentry_list);
637 /* Add the trees recursively. */
638 for (size_t i = 0; i < num_trees; i++)
639 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
641 /* If requested, add ancestors of the trees. */
643 for (size_t i = 0; i < num_trees; i++) {
644 struct wim_dentry *dentry = trees[i];
645 struct wim_dentry *ancestor;
646 struct list_head *place_after;
648 if (dentry_is_root(dentry))
651 place_after = dentry_list;
654 ancestor = ancestor->d_parent;
655 if (will_extract_dentry(ancestor)) {
656 place_after = &ancestor->d_extraction_list_node;
659 } while (!dentry_is_root(ancestor));
663 ancestor = ancestor->d_parent;
664 if (will_extract_dentry(ancestor))
666 list_add(&ancestor->d_extraction_list_node, place_after);
667 } while (!dentry_is_root(ancestor));
673 destroy_dentry_list(struct list_head *dentry_list)
675 struct wim_dentry *dentry, *tmp;
676 struct wim_inode *inode;
678 list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
679 inode = dentry->d_inode;
680 dentry_reset_extraction_list_node(dentry);
681 inode->i_visited = 0;
682 if ((void *)dentry->d_extraction_name != (void *)dentry->file_name)
683 FREE(dentry->d_extraction_name);
684 dentry->d_extraction_name = NULL;
685 dentry->d_extraction_name_nchars = 0;
690 destroy_stream_list(struct list_head *stream_list)
692 struct wim_lookup_table_entry *lte;
694 list_for_each_entry(lte, stream_list, extraction_list)
695 if (lte->out_refcnt > ARRAY_LEN(lte->inline_stream_owners))
696 FREE(lte->stream_owners);
700 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
702 static const utf16lechar replacement_char = cpu_to_le16('?');
706 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
712 for (i = 0; i < num_chars; i++) {
715 case cpu_to_le16('\\'):
716 case cpu_to_le16(':'):
717 case cpu_to_le16('*'):
718 case cpu_to_le16('?'):
719 case cpu_to_le16('"'):
720 case cpu_to_le16('<'):
721 case cpu_to_le16('>'):
722 case cpu_to_le16('|'):
724 case cpu_to_le16('/'):
725 case cpu_to_le16('\0'):
727 name[i] = replacement_char;
734 if (name[num_chars - 1] == cpu_to_le16(' ') ||
735 name[num_chars - 1] == cpu_to_le16('.'))
738 name[num_chars - 1] = replacement_char;
747 dentry_calculate_extraction_name(struct wim_dentry *dentry,
748 struct apply_ctx *ctx)
752 if (!dentry_is_supported(dentry, &ctx->supported_features))
755 if (dentry_is_root(dentry))
759 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
760 dentry->d_extraction_name = dentry->file_name;
761 dentry->d_extraction_name_nchars = dentry->file_name_nbytes /
767 if (!ctx->supported_features.case_sensitive_filenames) {
768 struct wim_dentry *other;
769 list_for_each_entry(other, &dentry->d_ci_conflict_list,
772 if (will_extract_dentry(other)) {
773 if (ctx->extract_flags &
774 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
775 WARNING("\"%"TS"\" has the same "
776 "case-insensitive name as "
777 "\"%"TS"\"; extracting "
778 "dummy name instead",
779 dentry_full_path(dentry),
780 dentry_full_path(other));
783 WARNING("Not extracting \"%"TS"\": "
784 "has same case-insensitive "
786 dentry_full_path(dentry),
787 dentry_full_path(other));
794 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
795 ret = utf16le_get_tstr(dentry->file_name,
796 dentry->file_name_nbytes,
797 (const tchar **)&dentry->d_extraction_name,
798 &dentry->d_extraction_name_nchars);
799 dentry->d_extraction_name_nchars /= sizeof(tchar);
802 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
804 WARNING("\"%"TS"\" has an invalid filename "
805 "that is not supported on this platform; "
806 "extracting dummy name instead",
807 dentry_full_path(dentry));
810 WARNING("Not extracting \"%"TS"\": has an invalid filename "
811 "that is not supported on this platform",
812 dentry_full_path(dentry));
819 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
821 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
822 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
824 const tchar *tchar_name;
827 ret = utf16le_get_tstr(utf16_name_copy,
828 dentry->file_name_nbytes,
829 &tchar_name, &tchar_nchars);
833 tchar_nchars /= sizeof(tchar);
835 size_t fixed_name_num_chars = tchar_nchars;
836 tchar fixed_name[tchar_nchars + 50];
838 tmemcpy(fixed_name, tchar_name, tchar_nchars);
839 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
840 T(" (invalid filename #%lu)"),
841 ++ctx->invalid_sequence);
843 utf16le_put_tstr(tchar_name);
845 dentry->d_extraction_name = memdup(fixed_name,
846 2 * fixed_name_num_chars + 2);
847 if (!dentry->d_extraction_name)
848 return WIMLIB_ERR_NOMEM;
849 dentry->d_extraction_name_nchars = fixed_name_num_chars;
854 for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
859 * Calculate the actual filename component at which each WIM dentry will be
860 * extracted, with special handling for dentries that are unsupported by the
861 * extraction backend or have invalid names.
863 * ctx->supported_features must be filled in.
865 * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
868 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
869 struct apply_ctx *ctx)
871 struct list_head *prev, *cur;
873 /* Can't use list_for_each_entry() because a call to
874 * dentry_calculate_extraction_name() may delete the current dentry and
875 * its children from the list. */
879 struct wim_dentry *dentry;
883 if (cur == dentry_list)
886 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
888 ret = dentry_calculate_extraction_name(dentry, ctx);
892 if (prev->next == cur)
895 ; /* Current dentry and its children (which follow in
896 the list) were deleted. prev stays the same. */
902 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
903 struct wim_lookup_table *lookup_table)
905 struct wim_inode *inode = dentry->d_inode;
906 struct wim_lookup_table_entry *lte;
910 /* Special case: when extracting from a pipe, the WIM lookup table is
911 * initially empty, so "resolving" an inode's streams is initially not
912 * possible. However, we still need to keep track of which streams,
913 * identified by SHA1 message digests, need to be extracted, so we
914 * "resolve" the inode's streams anyway by allocating new entries. */
915 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
917 ret = inode_resolve_streams(inode, lookup_table, force);
920 for (u32 i = 0; i <= inode->i_num_ads; i++) {
921 lte = inode_stream_lte_resolved(inode, i);
929 * For each dentry to be extracted, resolve all streams in the corresponding
930 * inode and set 'out_refcnt' in each to 0.
932 * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
935 dentry_list_resolve_streams(struct list_head *dentry_list,
936 struct apply_ctx *ctx)
938 struct wim_dentry *dentry;
941 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
942 ret = dentry_resolve_streams(dentry,
944 ctx->wim->lookup_table);
952 ref_stream(struct wim_lookup_table_entry *lte, u32 stream_idx,
953 struct wim_dentry *dentry, struct apply_ctx *ctx)
955 struct wim_inode *inode = dentry->d_inode;
956 struct stream_owner *stream_owners;
961 /* Tally the size only for each extraction of the stream (not hard
963 if (inode->i_visited && ctx->supported_features.hard_links)
966 ctx->progress.extract.total_bytes += lte->size;
967 ctx->progress.extract.num_streams++;
969 if (inode->i_visited)
972 /* Add stream to the dentry_list only one time, even if it's going
973 * to be extracted to multiple inodes. */
974 if (lte->out_refcnt == 0) {
975 list_add_tail(<e->extraction_list, &ctx->stream_list);
976 ctx->num_streams_remaining++;
979 /* If inode not yet been visited, append it to the stream_owners array. */
980 if (lte->out_refcnt < ARRAY_LEN(lte->inline_stream_owners)) {
981 stream_owners = lte->inline_stream_owners;
983 struct stream_owner *prev_stream_owners;
984 size_t alloc_stream_owners;
986 if (lte->out_refcnt == ARRAY_LEN(lte->inline_stream_owners)) {
987 prev_stream_owners = NULL;
988 alloc_stream_owners = ARRAY_LEN(lte->inline_stream_owners);
990 prev_stream_owners = lte->stream_owners;
991 alloc_stream_owners = lte->alloc_stream_owners;
994 if (lte->out_refcnt == alloc_stream_owners) {
995 alloc_stream_owners *= 2;
996 stream_owners = REALLOC(prev_stream_owners,
997 alloc_stream_owners *
998 sizeof(stream_owners[0]));
1000 return WIMLIB_ERR_NOMEM;
1001 if (!prev_stream_owners) {
1002 memcpy(stream_owners,
1003 lte->inline_stream_owners,
1004 sizeof(lte->inline_stream_owners));
1006 lte->stream_owners = stream_owners;
1007 lte->alloc_stream_owners = alloc_stream_owners;
1009 stream_owners = lte->stream_owners;
1011 stream_owners[lte->out_refcnt].inode = inode;
1012 if (stream_idx == 0) {
1013 stream_owners[lte->out_refcnt].stream_name = NULL;
1015 stream_owners[lte->out_refcnt].stream_name =
1016 inode->i_ads_entries[stream_idx - 1].stream_name;
1023 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1025 struct wim_inode *inode = dentry->d_inode;
1028 /* The unnamed data stream will always be extracted, except in an
1030 if (!inode_is_encrypted_directory(inode)) {
1032 struct wim_lookup_table_entry *stream;
1034 stream = inode_unnamed_stream_resolved(inode, &stream_idx);
1035 ret = ref_stream(stream, stream_idx, dentry, ctx);
1040 /* Named data streams will be extracted only if supported in the current
1041 * extraction mode and volume, and to avoid complications, if not doing
1042 * a linked extraction. */
1043 if (ctx->supported_features.named_data_streams) {
1044 for (u16 i = 0; i < inode->i_num_ads; i++) {
1045 if (!ads_entry_is_named_stream(&inode->i_ads_entries[i]))
1047 ret = ref_stream(inode->i_ads_entries[i].lte, i + 1,
1053 inode->i_visited = 1;
1058 * For each dentry to be extracted, iterate through the data streams of the
1059 * corresponding inode. For each such stream that is not to be ignored due to
1060 * the supported features or extraction flags, add it to the list of streams to
1061 * be extracted (ctx->stream_list) if not already done so.
1063 * Also builds a mapping from each stream to the inodes referencing it.
1065 * This also initializes the extract progress info with byte and stream
1068 * ctx->supported_features must be filled in.
1070 * Possible error codes: WIMLIB_ERR_NOMEM.
1073 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1075 struct wim_dentry *dentry;
1078 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1079 ret = dentry_ref_streams(dentry, ctx);
1083 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1084 dentry->d_inode->i_visited = 0;
1089 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1091 struct wim_dentry *dentry;
1092 struct wim_inode *inode;
1094 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1095 inode = dentry->d_inode;
1096 if (!inode->i_visited)
1097 INIT_LIST_HEAD(&inode->i_extraction_aliases);
1098 list_add_tail(&dentry->d_extraction_alias_node,
1099 &inode->i_extraction_aliases);
1100 inode->i_visited = 1;
1102 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1103 dentry->d_inode->i_visited = 0;
1107 inode_tally_features(const struct wim_inode *inode,
1108 struct wim_features *features)
1110 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1111 features->archive_files++;
1112 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1113 features->hidden_files++;
1114 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1115 features->system_files++;
1116 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1117 features->compressed_files++;
1118 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1119 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1120 features->encrypted_directories++;
1122 features->encrypted_files++;
1124 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1125 features->not_context_indexed_files++;
1126 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1127 features->sparse_files++;
1128 if (inode_has_named_stream(inode))
1129 features->named_data_streams++;
1130 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1131 features->reparse_points++;
1132 if (inode_is_symlink(inode))
1133 features->symlink_reparse_points++;
1135 features->other_reparse_points++;
1137 if (inode->i_security_id != -1)
1138 features->security_descriptors++;
1139 if (inode_has_unix_data(inode))
1140 features->unix_data++;
1143 /* Tally features necessary to extract a dentry and the corresponding inode. */
1145 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1147 struct wim_inode *inode = dentry->d_inode;
1149 if (dentry_has_short_name(dentry))
1150 features->short_names++;
1152 if (inode->i_visited) {
1153 features->hard_links++;
1155 inode_tally_features(inode, features);
1156 inode->i_visited = 1;
1160 /* Tally the features necessary to extract the specified dentries. */
1162 dentry_list_get_features(struct list_head *dentry_list,
1163 struct wim_features *features)
1165 struct wim_dentry *dentry;
1167 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1168 dentry_tally_features(dentry, features);
1170 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1171 dentry->d_inode->i_visited = 0;
1175 do_feature_check(const struct wim_features *required_features,
1176 const struct wim_features *supported_features,
1179 /* File attributes. */
1180 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1181 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1182 * We're an archive program, so theoretically we can do what we
1185 if (required_features->hidden_files &&
1186 !supported_features->hidden_files)
1187 WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1188 required_features->hidden_files);
1190 if (required_features->system_files &&
1191 !supported_features->system_files)
1192 WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1193 required_features->system_files);
1195 if (required_features->compressed_files &&
1196 !supported_features->compressed_files)
1197 WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1198 required_features->compressed_files);
1200 if (required_features->not_context_indexed_files &&
1201 !supported_features->not_context_indexed_files)
1202 WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1203 required_features->not_context_indexed_files);
1205 if (required_features->sparse_files &&
1206 !supported_features->sparse_files)
1207 WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1208 required_features->sparse_files);
1210 if (required_features->encrypted_directories &&
1211 !supported_features->encrypted_directories)
1212 WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1213 required_features->encrypted_directories);
1216 /* Encrypted files. */
1217 if (required_features->encrypted_files &&
1218 !supported_features->encrypted_files)
1219 WARNING("Ignoring %lu encrypted files",
1220 required_features->encrypted_files);
1222 /* Named data streams. */
1223 if (required_features->named_data_streams &&
1224 (!supported_features->named_data_streams))
1225 WARNING("Ignoring named data streams of %lu files",
1226 required_features->named_data_streams);
1229 if (required_features->hard_links && !supported_features->hard_links)
1230 WARNING("Extracting %lu hard links as independent files",
1231 required_features->hard_links);
1233 /* Symbolic links and reparse points. */
1234 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1235 required_features->symlink_reparse_points &&
1236 !supported_features->symlink_reparse_points &&
1237 !supported_features->reparse_points)
1239 ERROR("Extraction backend does not support symbolic links!");
1240 return WIMLIB_ERR_UNSUPPORTED;
1242 if (required_features->reparse_points &&
1243 !supported_features->reparse_points)
1245 if (supported_features->symlink_reparse_points) {
1246 if (required_features->other_reparse_points) {
1247 WARNING("Ignoring %lu non-symlink/junction "
1248 "reparse point files",
1249 required_features->other_reparse_points);
1252 WARNING("Ignoring %lu reparse point files",
1253 required_features->reparse_points);
1257 /* Security descriptors. */
1258 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1259 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1260 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1261 required_features->security_descriptors &&
1262 !supported_features->security_descriptors)
1264 ERROR("Extraction backend does not support security descriptors!");
1265 return WIMLIB_ERR_UNSUPPORTED;
1267 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1268 required_features->security_descriptors &&
1269 !supported_features->security_descriptors)
1270 WARNING("Ignoring Windows NT security descriptors of %lu files",
1271 required_features->security_descriptors);
1274 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1275 required_features->unix_data && !supported_features->unix_data)
1277 ERROR("Extraction backend does not support UNIX data!");
1278 return WIMLIB_ERR_UNSUPPORTED;
1281 if (required_features->unix_data &&
1282 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1284 WARNING("Ignoring UNIX metadata of %lu files",
1285 required_features->unix_data);
1289 if (required_features->short_names &&
1290 !supported_features->short_names)
1292 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1293 ERROR("Extraction backend does not support DOS names!");
1294 return WIMLIB_ERR_UNSUPPORTED;
1296 WARNING("Ignoring DOS names of %lu files",
1297 required_features->short_names);
1301 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1302 !supported_features->timestamps)
1304 ERROR("Extraction backend does not support timestamps!");
1305 return WIMLIB_ERR_UNSUPPORTED;
1311 static const struct apply_operations *
1312 select_apply_operations(int extract_flags)
1315 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1316 return &ntfs_3g_apply_ops;
1319 return &win32_apply_ops;
1321 return &unix_apply_ops;
1326 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1327 const tchar *target, int extract_flags)
1329 const struct apply_operations *ops;
1330 struct apply_ctx *ctx;
1332 LIST_HEAD(dentry_list);
1334 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1335 ret = extract_dentries_to_stdout(trees, num_trees,
1340 num_trees = remove_duplicate_trees(trees, num_trees);
1341 num_trees = remove_contained_trees(trees, num_trees);
1343 ops = select_apply_operations(extract_flags);
1345 if (num_trees > 1 && ops->single_tree_only) {
1346 ERROR("Extracting multiple directory trees "
1347 "at once is not supported in %s extraction mode!",
1349 ret = WIMLIB_ERR_UNSUPPORTED;
1353 ctx = CALLOC(1, ops->context_size);
1355 ret = WIMLIB_ERR_NOMEM;
1360 ctx->target = target;
1361 ctx->target_nchars = tstrlen(target);
1362 ctx->extract_flags = extract_flags;
1363 if (ctx->wim->progfunc) {
1364 ctx->progfunc = ctx->wim->progfunc;
1365 ctx->progctx = ctx->wim->progctx;
1366 ctx->progress.extract.image = wim->current_image;
1367 ctx->progress.extract.extract_flags = (extract_flags &
1368 WIMLIB_EXTRACT_MASK_PUBLIC);
1369 ctx->progress.extract.wimfile_name = wim->filename;
1370 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1371 wim->current_image);
1372 ctx->progress.extract.target = target;
1374 INIT_LIST_HEAD(&ctx->stream_list);
1375 filedes_invalidate(&ctx->tmpfile_fd);
1377 ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1381 build_dentry_list(&dentry_list, trees, num_trees,
1383 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1385 dentry_list_get_features(&dentry_list, &ctx->required_features);
1387 ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1388 ctx->extract_flags);
1392 ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1396 ret = dentry_list_resolve_streams(&dentry_list, ctx);
1400 ret = dentry_list_ref_streams(&dentry_list, ctx);
1404 dentry_list_build_inode_alias_lists(&dentry_list);
1406 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1407 /* When extracting from a pipe, the number of bytes of data to
1408 * extract can't be determined in the normal way (examining the
1409 * lookup table), since at this point all we have is a set of
1410 * SHA1 message digests of streams that need to be extracted.
1411 * However, we can get a reasonably accurate estimate by taking
1412 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1413 * data. This does assume that a full image is being extracted,
1414 * but currently there is no API for doing otherwise. (Also,
1415 * subtract <HARDLINKBYTES> from this if hard links are
1416 * supported by the extraction mode.) */
1417 ctx->progress.extract.total_bytes =
1418 wim_info_get_image_total_bytes(wim->wim_info,
1419 wim->current_image);
1420 if (ctx->supported_features.hard_links) {
1421 ctx->progress.extract.total_bytes -=
1422 wim_info_get_image_hard_link_bytes(wim->wim_info,
1423 wim->current_image);
1427 ret = extract_progress(ctx,
1428 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1429 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1430 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1434 ret = (*ops->extract)(&dentry_list, ctx);
1438 if (ctx->progress.extract.completed_bytes <
1439 ctx->progress.extract.total_bytes)
1441 ctx->progress.extract.completed_bytes =
1442 ctx->progress.extract.total_bytes;
1443 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1448 ret = extract_progress(ctx,
1449 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1450 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1451 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1453 destroy_stream_list(&ctx->stream_list);
1454 destroy_dentry_list(&dentry_list);
1461 mkdir_if_needed(const tchar *target)
1464 if (tstat(target, &stbuf)) {
1465 if (errno == ENOENT) {
1466 if (tmkdir(target, 0755)) {
1467 ERROR_WITH_ERRNO("Failed to create directory "
1468 "\"%"TS"\"", target);
1469 return WIMLIB_ERR_MKDIR;
1472 ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target);
1473 return WIMLIB_ERR_STAT;
1475 } else if (!S_ISDIR(stbuf.st_mode)) {
1476 ERROR("\"%"TS"\" is not a directory", target);
1477 return WIMLIB_ERR_NOTDIR;
1482 /* Make sure the extraction flags make sense, and update them if needed. */
1484 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1486 int extract_flags = *extract_flags_p;
1488 /* Check for invalid flag combinations */
1490 if ((extract_flags &
1491 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1492 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1493 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1494 return WIMLIB_ERR_INVALID_PARAM;
1496 if ((extract_flags &
1497 (WIMLIB_EXTRACT_FLAG_RPFIX |
1498 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1499 WIMLIB_EXTRACT_FLAG_NORPFIX))
1500 return WIMLIB_ERR_INVALID_PARAM;
1502 #ifndef WITH_NTFS_3G
1503 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1504 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
1505 " it cannot apply a WIM image directly to an NTFS volume.");
1506 return WIMLIB_ERR_UNSUPPORTED;
1511 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1512 ERROR("WIMBoot extraction is only supported on Windows!");
1513 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_streams(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
1820 * an empty lookup 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 struct wim_lookup_table_entry xml_lte;
1853 struct wim_resource_spec xml_rspec;
1854 ret = read_pwm_stream_header(pwm, &xml_lte, &xml_rspec, 0, NULL);
1856 goto out_wimlib_free;
1858 if (!(xml_lte.flags & WIM_RESHDR_FLAG_METADATA))
1860 ERROR("Expected XML data, but found non-metadata "
1862 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1863 goto out_wimlib_free;
1866 wim_res_spec_to_hdr(&xml_rspec, &pwm->hdr.xml_data_reshdr);
1868 ret = read_wim_xml_data(pwm);
1870 goto out_wimlib_free;
1872 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
1873 ERROR("Image count in XML data is not the same as in WIM header.");
1874 ret = WIMLIB_ERR_IMAGE_COUNT;
1875 goto out_wimlib_free;
1879 /* Get image index (this may use the XML data that was just read to
1880 * resolve an image name). */
1881 if (image_num_or_name) {
1882 image = wimlib_resolve_image(pwm, image_num_or_name);
1883 if (image == WIMLIB_NO_IMAGE) {
1884 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
1886 ret = WIMLIB_ERR_INVALID_IMAGE;
1887 goto out_wimlib_free;
1888 } else if (image == WIMLIB_ALL_IMAGES) {
1889 ERROR("Applying all images from a pipe is not supported!");
1890 ret = WIMLIB_ERR_INVALID_IMAGE;
1891 goto out_wimlib_free;
1894 if (pwm->hdr.image_count != 1) {
1895 ERROR("No image was specified, but the pipable WIM "
1896 "did not contain exactly 1 image");
1897 ret = WIMLIB_ERR_INVALID_IMAGE;
1898 goto out_wimlib_free;
1903 /* Load the needed metadata resource. */
1904 for (i = 1; i <= pwm->hdr.image_count; i++) {
1905 struct wim_lookup_table_entry *metadata_lte;
1906 struct wim_image_metadata *imd;
1907 struct wim_resource_spec *metadata_rspec;
1909 metadata_lte = new_lookup_table_entry();
1910 if (metadata_lte == NULL) {
1911 ret = WIMLIB_ERR_NOMEM;
1912 goto out_wimlib_free;
1914 metadata_rspec = MALLOC(sizeof(struct wim_resource_spec));
1915 if (metadata_rspec == NULL) {
1916 ret = WIMLIB_ERR_NOMEM;
1917 free_lookup_table_entry(metadata_lte);
1918 goto out_wimlib_free;
1921 ret = read_pwm_stream_header(pwm, metadata_lte, metadata_rspec, 0, NULL);
1922 imd = pwm->image_metadata[i - 1];
1923 imd->metadata_lte = metadata_lte;
1925 FREE(metadata_rspec);
1926 goto out_wimlib_free;
1929 if (!(metadata_lte->flags & WIM_RESHDR_FLAG_METADATA)) {
1930 ERROR("Expected metadata resource, but found "
1931 "non-metadata stream.");
1932 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1933 goto out_wimlib_free;
1937 /* Metadata resource is for the image being extracted.
1938 * Parse it and save the metadata in memory. */
1939 ret = read_metadata_resource(pwm, imd);
1941 goto out_wimlib_free;
1944 /* Metadata resource is not for the image being
1945 * extracted. Skip over it. */
1946 ret = skip_wim_stream(metadata_lte);
1948 goto out_wimlib_free;
1951 /* Extract the image. */
1952 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
1953 ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
1954 /* Clean up and return. */
1962 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
1963 const tchar *target, int extract_flags)
1965 return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
1974 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1977 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1978 return WIMLIB_ERR_INVALID_PARAM;
1979 return do_wimlib_extract_image(wim, image, target, extract_flags);