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;
315 if (unlikely(lte->out_refcnt > MAX_OPEN_STREAMS))
316 return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
318 return (*ctx->saved_cbs->begin_stream)(lte, ctx->saved_cbs->begin_stream_ctx);
322 extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx)
324 struct apply_ctx *ctx = _ctx;
325 union wimlib_progress_info *progress = &ctx->progress;
328 if (likely(ctx->supported_features.hard_links)) {
329 progress->extract.completed_bytes +=
330 (u64)size * ctx->cur_stream->out_refcnt;
332 const struct stream_owner *owners = stream_owners(ctx->cur_stream);
333 for (u32 i = 0; i < ctx->cur_stream->out_refcnt; i++) {
334 const struct wim_inode *inode = owners[i].inode;
335 const struct wim_dentry *dentry;
337 list_for_each_entry(dentry,
338 &inode->i_extraction_aliases,
339 d_extraction_alias_node)
341 progress->extract.completed_bytes += size;
345 if (progress->extract.completed_bytes >= ctx->next_progress) {
347 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
351 if (progress->extract.completed_bytes >=
352 progress->extract.total_bytes)
354 ctx->next_progress = UINT64_MAX;
356 /* Send new message as soon as another 1/128 of the
357 * total has been extracted. (Arbitrary number.) */
359 progress->extract.completed_bytes +
360 progress->extract.total_bytes / 128;
362 /* ... Unless that would be more than 5000000 bytes, in
363 * which case send the next after the next 5000000
364 * bytes. (Another arbitrary number.) */
365 if (progress->extract.completed_bytes + 5000000 <
368 progress->extract.completed_bytes + 5000000;
370 /* ... But always send a message as soon as we're
371 * completely done. */
372 if (progress->extract.total_bytes < ctx->next_progress)
373 ctx->next_progress = progress->extract.total_bytes;
377 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
378 /* Just extracting to temporary file for now. */
379 ret = full_write(&ctx->tmpfile_fd, chunk, size);
381 ERROR_WITH_ERRNO("Error writing data to "
382 "temporary file \"%"TS"\"",
387 return (*ctx->saved_cbs->consume_chunk)(chunk, size,
388 ctx->saved_cbs->consume_chunk_ctx);
393 extract_from_tmpfile(const tchar *tmpfile_name, struct apply_ctx *ctx)
395 struct wim_lookup_table_entry tmpfile_lte;
396 struct wim_lookup_table_entry *orig_lte = ctx->cur_stream;
397 const struct stream_owner *owners = stream_owners(orig_lte);
398 const struct read_stream_list_callbacks *cbs = ctx->saved_cbs;
401 /* Copy the stream's data from the temporary file to each of its
404 * This is executed only in the very uncommon case that a
405 * single-instance stream is being extracted to more than
406 * MAX_OPEN_STREAMS locations! */
408 memcpy(&tmpfile_lte, orig_lte, sizeof(struct wim_lookup_table_entry));
409 tmpfile_lte.resource_location = RESOURCE_IN_FILE_ON_DISK;
410 tmpfile_lte.file_on_disk = ctx->tmpfile_name;
411 tmpfile_lte.out_refcnt = 1;
413 for (u32 i = 0; i < orig_lte->out_refcnt; i++) {
414 tmpfile_lte.inline_stream_owners[0] = owners[i];
416 /* Note: it usually doesn't matter whether we pass the original
417 * stream entry to callbacks provided by the extraction backend
418 * as opposed to the tmpfile stream entry, since they shouldn't
419 * actually read data from the stream other than through the
420 * read_stream_prefix() call below. But for
421 * WIMLIB_EXTRACT_FLAG_WIMBOOT mode on Windows it does matter
422 * because it needs the original stream location in order to
423 * create the external backing reference. */
425 ret = (*cbs->begin_stream)(orig_lte, cbs->begin_stream_ctx);
429 /* Extra SHA-1 isn't necessary here, but it shouldn't hurt as
430 * this case is very rare anyway. */
431 ret = extract_stream(&tmpfile_lte, tmpfile_lte.size,
433 cbs->consume_chunk_ctx);
435 return (*cbs->end_stream)(orig_lte, ret,
436 cbs->end_stream_ctx);
442 end_extract_stream_wrapper(struct wim_lookup_table_entry *stream,
443 int status, void *_ctx)
445 struct apply_ctx *ctx = _ctx;
447 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
448 filedes_close(&ctx->tmpfile_fd);
450 status = extract_from_tmpfile(ctx->tmpfile_name, ctx);
451 filedes_invalidate(&ctx->tmpfile_fd);
452 tunlink(ctx->tmpfile_name);
453 FREE(ctx->tmpfile_name);
456 return (*ctx->saved_cbs->end_stream)(stream, status,
457 ctx->saved_cbs->end_stream_ctx);
462 * Read the list of single-instance streams to extract and feed their data into
463 * the specified callback functions.
465 * This handles checksumming each stream.
467 * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
469 * This also works if the WIM is being read from a pipe, whereas attempting to
470 * read streams directly (e.g. with read_full_stream_into_buf()) will not.
472 * This also will split up streams that will need to be extracted to more than
473 * MAX_OPEN_STREAMS locations, as measured by the 'out_refcnt' of each stream.
474 * Therefore, the apply_operations implementation need not worry about running
475 * out of file descriptors, unless it might open more than one file descriptor
476 * per nominal destination (e.g. Win32 currently might because the destination
477 * file system might not support hard links).
480 extract_stream_list(struct apply_ctx *ctx,
481 const struct read_stream_list_callbacks *cbs)
483 struct read_stream_list_callbacks wrapper_cbs = {
484 .begin_stream = begin_extract_stream_wrapper,
485 .begin_stream_ctx = ctx,
486 .consume_chunk = extract_chunk_wrapper,
487 .consume_chunk_ctx = ctx,
488 .end_stream = end_extract_stream_wrapper,
489 .end_stream_ctx = ctx,
491 ctx->saved_cbs = cbs;
492 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
493 return load_streams_from_pipe(ctx, &wrapper_cbs);
495 return read_stream_list(&ctx->stream_list,
496 offsetof(struct wim_lookup_table_entry,
498 &wrapper_cbs, VERIFY_STREAM_HASHES);
502 /* Extract a WIM dentry to standard output.
504 * This obviously doesn't make sense in all cases. We return an error if the
505 * dentry does not correspond to a regular file. Otherwise we extract the
506 * unnamed data stream only. */
508 extract_dentry_to_stdout(struct wim_dentry *dentry,
509 const struct wim_lookup_table *lookup_table)
511 struct wim_inode *inode = dentry->d_inode;
512 struct wim_lookup_table_entry *lte;
513 struct filedes _stdout;
515 if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
516 FILE_ATTRIBUTE_DIRECTORY))
518 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
519 "extracted to standard output", dentry_full_path(dentry));
520 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
523 lte = inode_unnamed_lte(inode, lookup_table);
525 const u8 *hash = inode_unnamed_stream_hash(inode);
526 if (!is_zero_hash(hash))
527 return stream_not_found_error(inode, hash);
531 filedes_init(&_stdout, STDOUT_FILENO);
532 return extract_full_stream_to_fd(lte, &_stdout);
536 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
537 const struct wim_lookup_table *lookup_table)
539 for (size_t i = 0; i < num_dentries; i++) {
540 int ret = extract_dentry_to_stdout(dentries[i], lookup_table);
547 /**********************************************************************/
550 * Removes duplicate dentries from the array.
552 * Returns the new number of dentries, packed at the front of the array.
555 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
558 for (i = 0; i < num_trees; i++) {
559 if (!trees[i]->tmp_flag) {
560 /* Found distinct dentry. */
561 trees[i]->tmp_flag = 1;
562 trees[j++] = trees[i];
565 for (i = 0; i < j; i++)
566 trees[i]->tmp_flag = 0;
571 * Remove dentries that are descendants of other dentries in the array.
573 * Returns the new number of dentries, packed at the front of the array.
576 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
579 for (i = 0; i < num_trees; i++)
580 trees[i]->tmp_flag = 1;
581 for (i = 0; i < num_trees; i++) {
582 struct wim_dentry *d = trees[i];
583 while (!dentry_is_root(d)) {
588 trees[j++] = trees[i];
592 trees[i]->tmp_flag = 0;
595 for (i = 0; i < j; i++)
596 trees[i]->tmp_flag = 0;
601 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
603 struct list_head *dentry_list = _dentry_list;
604 list_add_tail(&dentry->d_extraction_list_node, dentry_list);
609 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
611 dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
615 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
617 list_del(&dentry->d_extraction_list_node);
618 dentry_reset_extraction_list_node(dentry);
623 * Build the preliminary list of dentries to be extracted.
625 * The list maintains the invariant that if d1 and d2 are in the list and d1 is
626 * an ancestor of d2, then d1 appears before d2 in the list.
629 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
630 size_t num_trees, bool add_ancestors)
632 INIT_LIST_HEAD(dentry_list);
634 /* Add the trees recursively. */
635 for (size_t i = 0; i < num_trees; i++)
636 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
638 /* If requested, add ancestors of the trees. */
640 for (size_t i = 0; i < num_trees; i++) {
641 struct wim_dentry *dentry = trees[i];
642 struct wim_dentry *ancestor;
643 struct list_head *place_after;
645 if (dentry_is_root(dentry))
648 place_after = dentry_list;
651 ancestor = ancestor->d_parent;
652 if (will_extract_dentry(ancestor)) {
653 place_after = &ancestor->d_extraction_list_node;
656 } while (!dentry_is_root(ancestor));
660 ancestor = ancestor->d_parent;
661 if (will_extract_dentry(ancestor))
663 list_add(&ancestor->d_extraction_list_node, place_after);
664 } while (!dentry_is_root(ancestor));
670 destroy_dentry_list(struct list_head *dentry_list)
672 struct wim_dentry *dentry, *tmp;
673 struct wim_inode *inode;
675 list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
676 inode = dentry->d_inode;
677 dentry_reset_extraction_list_node(dentry);
678 inode->i_visited = 0;
679 if ((void *)dentry->d_extraction_name != (void *)dentry->file_name)
680 FREE(dentry->d_extraction_name);
681 dentry->d_extraction_name = NULL;
682 dentry->d_extraction_name_nchars = 0;
687 destroy_stream_list(struct list_head *stream_list)
689 struct wim_lookup_table_entry *lte;
691 list_for_each_entry(lte, stream_list, extraction_list)
692 if (lte->out_refcnt > ARRAY_LEN(lte->inline_stream_owners))
693 FREE(lte->stream_owners);
697 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
699 static const utf16lechar replacement_char = cpu_to_le16('?');
703 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
709 for (i = 0; i < num_chars; i++) {
712 case cpu_to_le16('\\'):
713 case cpu_to_le16(':'):
714 case cpu_to_le16('*'):
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('|'):
721 case cpu_to_le16('/'):
722 case cpu_to_le16('\0'):
724 name[i] = replacement_char;
731 if (name[num_chars - 1] == cpu_to_le16(' ') ||
732 name[num_chars - 1] == cpu_to_le16('.'))
735 name[num_chars - 1] = replacement_char;
744 dentry_calculate_extraction_name(struct wim_dentry *dentry,
745 struct apply_ctx *ctx)
749 if (!dentry_is_supported(dentry, &ctx->supported_features))
752 if (dentry_is_root(dentry))
756 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
757 dentry->d_extraction_name = dentry->file_name;
758 dentry->d_extraction_name_nchars = dentry->file_name_nbytes /
764 if (!ctx->supported_features.case_sensitive_filenames) {
765 struct wim_dentry *other;
766 list_for_each_entry(other, &dentry->d_ci_conflict_list,
769 if (will_extract_dentry(other)) {
770 if (ctx->extract_flags &
771 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
772 WARNING("\"%"TS"\" has the same "
773 "case-insensitive name as "
774 "\"%"TS"\"; extracting "
775 "dummy name instead",
776 dentry_full_path(dentry),
777 dentry_full_path(other));
780 WARNING("Not extracting \"%"TS"\": "
781 "has same case-insensitive "
783 dentry_full_path(dentry),
784 dentry_full_path(other));
791 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
792 ret = utf16le_get_tstr(dentry->file_name,
793 dentry->file_name_nbytes,
794 (const tchar **)&dentry->d_extraction_name,
795 &dentry->d_extraction_name_nchars);
796 dentry->d_extraction_name_nchars /= sizeof(tchar);
799 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
801 WARNING("\"%"TS"\" has an invalid filename "
802 "that is not supported on this platform; "
803 "extracting dummy name instead",
804 dentry_full_path(dentry));
807 WARNING("Not extracting \"%"TS"\": has an invalid filename "
808 "that is not supported on this platform",
809 dentry_full_path(dentry));
816 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
818 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
819 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
821 const tchar *tchar_name;
824 ret = utf16le_get_tstr(utf16_name_copy,
825 dentry->file_name_nbytes,
826 &tchar_name, &tchar_nchars);
830 tchar_nchars /= sizeof(tchar);
832 size_t fixed_name_num_chars = tchar_nchars;
833 tchar fixed_name[tchar_nchars + 50];
835 tmemcpy(fixed_name, tchar_name, tchar_nchars);
836 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
837 T(" (invalid filename #%lu)"),
838 ++ctx->invalid_sequence);
840 utf16le_put_tstr(tchar_name);
842 dentry->d_extraction_name = memdup(fixed_name,
843 2 * fixed_name_num_chars + 2);
844 if (!dentry->d_extraction_name)
845 return WIMLIB_ERR_NOMEM;
846 dentry->d_extraction_name_nchars = fixed_name_num_chars;
851 for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
856 * Calculate the actual filename component at which each WIM dentry will be
857 * extracted, with special handling for dentries that are unsupported by the
858 * extraction backend or have invalid names.
860 * ctx->supported_features must be filled in.
862 * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
865 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
866 struct apply_ctx *ctx)
868 struct list_head *prev, *cur;
870 /* Can't use list_for_each_entry() because a call to
871 * dentry_calculate_extraction_name() may delete the current dentry and
872 * its children from the list. */
876 struct wim_dentry *dentry;
880 if (cur == dentry_list)
883 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
885 ret = dentry_calculate_extraction_name(dentry, ctx);
889 if (prev->next == cur)
892 ; /* Current dentry and its children (which follow in
893 the list) were deleted. prev stays the same. */
899 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
900 struct wim_lookup_table *lookup_table)
902 struct wim_inode *inode = dentry->d_inode;
903 struct wim_lookup_table_entry *lte;
907 /* Special case: when extracting from a pipe, the WIM lookup table is
908 * initially empty, so "resolving" an inode's streams is initially not
909 * possible. However, we still need to keep track of which streams,
910 * identified by SHA1 message digests, need to be extracted, so we
911 * "resolve" the inode's streams anyway by allocating new entries. */
912 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
914 ret = inode_resolve_streams(inode, lookup_table, force);
917 for (u32 i = 0; i <= inode->i_num_ads; i++) {
918 lte = inode_stream_lte_resolved(inode, i);
926 * For each dentry to be extracted, resolve all streams in the corresponding
927 * inode and set 'out_refcnt' in each to 0.
929 * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
932 dentry_list_resolve_streams(struct list_head *dentry_list,
933 struct apply_ctx *ctx)
935 struct wim_dentry *dentry;
938 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
939 ret = dentry_resolve_streams(dentry,
941 ctx->wim->lookup_table);
949 ref_stream(struct wim_lookup_table_entry *lte, u32 stream_idx,
950 struct wim_dentry *dentry, struct apply_ctx *ctx)
952 struct wim_inode *inode = dentry->d_inode;
953 struct stream_owner *stream_owners;
958 /* Tally the size only for each extraction of the stream (not hard
960 if (inode->i_visited && ctx->supported_features.hard_links)
963 ctx->progress.extract.total_bytes += lte->size;
964 ctx->progress.extract.num_streams++;
966 if (inode->i_visited)
969 /* Add stream to the dentry_list only one time, even if it's going
970 * to be extracted to multiple inodes. */
971 if (lte->out_refcnt == 0) {
972 list_add_tail(<e->extraction_list, &ctx->stream_list);
973 ctx->num_streams_remaining++;
976 /* If inode not yet been visited, append it to the stream_owners array. */
977 if (lte->out_refcnt < ARRAY_LEN(lte->inline_stream_owners)) {
978 stream_owners = lte->inline_stream_owners;
980 struct stream_owner *prev_stream_owners;
981 size_t alloc_stream_owners;
983 if (lte->out_refcnt == ARRAY_LEN(lte->inline_stream_owners)) {
984 prev_stream_owners = NULL;
985 alloc_stream_owners = ARRAY_LEN(lte->inline_stream_owners);
987 prev_stream_owners = lte->stream_owners;
988 alloc_stream_owners = lte->alloc_stream_owners;
991 if (lte->out_refcnt == alloc_stream_owners) {
992 alloc_stream_owners *= 2;
993 stream_owners = REALLOC(prev_stream_owners,
994 alloc_stream_owners *
995 sizeof(stream_owners[0]));
997 return WIMLIB_ERR_NOMEM;
998 if (!prev_stream_owners) {
999 memcpy(stream_owners,
1000 lte->inline_stream_owners,
1001 sizeof(lte->inline_stream_owners));
1003 lte->stream_owners = stream_owners;
1004 lte->alloc_stream_owners = alloc_stream_owners;
1006 stream_owners = lte->stream_owners;
1008 stream_owners[lte->out_refcnt].inode = inode;
1009 if (stream_idx == 0) {
1010 stream_owners[lte->out_refcnt].stream_name = NULL;
1012 stream_owners[lte->out_refcnt].stream_name =
1013 inode->i_ads_entries[stream_idx - 1].stream_name;
1020 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1022 struct wim_inode *inode = dentry->d_inode;
1025 /* The unnamed data stream will always be extracted, except in an
1027 if (!inode_is_encrypted_directory(inode)) {
1029 struct wim_lookup_table_entry *stream;
1031 stream = inode_unnamed_stream_resolved(inode, &stream_idx);
1032 ret = ref_stream(stream, stream_idx, dentry, ctx);
1037 /* Named data streams will be extracted only if supported in the current
1038 * extraction mode and volume, and to avoid complications, if not doing
1039 * a linked extraction. */
1040 if (ctx->supported_features.named_data_streams) {
1041 for (u16 i = 0; i < inode->i_num_ads; i++) {
1042 if (!ads_entry_is_named_stream(&inode->i_ads_entries[i]))
1044 ret = ref_stream(inode->i_ads_entries[i].lte, i + 1,
1050 inode->i_visited = 1;
1055 * For each dentry to be extracted, iterate through the data streams of the
1056 * corresponding inode. For each such stream that is not to be ignored due to
1057 * the supported features or extraction flags, add it to the list of streams to
1058 * be extracted (ctx->stream_list) if not already done so.
1060 * Also builds a mapping from each stream to the inodes referencing it.
1062 * This also initializes the extract progress info with byte and stream
1065 * ctx->supported_features must be filled in.
1067 * Possible error codes: WIMLIB_ERR_NOMEM.
1070 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1072 struct wim_dentry *dentry;
1075 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1076 ret = dentry_ref_streams(dentry, ctx);
1080 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1081 dentry->d_inode->i_visited = 0;
1086 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1088 struct wim_dentry *dentry;
1089 struct wim_inode *inode;
1091 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1092 inode = dentry->d_inode;
1093 if (!inode->i_visited)
1094 INIT_LIST_HEAD(&inode->i_extraction_aliases);
1095 list_add_tail(&dentry->d_extraction_alias_node,
1096 &inode->i_extraction_aliases);
1097 inode->i_visited = 1;
1099 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1100 dentry->d_inode->i_visited = 0;
1104 inode_tally_features(const struct wim_inode *inode,
1105 struct wim_features *features)
1107 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1108 features->archive_files++;
1109 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1110 features->hidden_files++;
1111 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1112 features->system_files++;
1113 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1114 features->compressed_files++;
1115 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1116 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1117 features->encrypted_directories++;
1119 features->encrypted_files++;
1121 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1122 features->not_context_indexed_files++;
1123 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1124 features->sparse_files++;
1125 if (inode_has_named_stream(inode))
1126 features->named_data_streams++;
1127 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1128 features->reparse_points++;
1129 if (inode_is_symlink(inode))
1130 features->symlink_reparse_points++;
1132 features->other_reparse_points++;
1134 if (inode->i_security_id != -1)
1135 features->security_descriptors++;
1136 if (inode_has_unix_data(inode))
1137 features->unix_data++;
1140 /* Tally features necessary to extract a dentry and the corresponding inode. */
1142 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1144 struct wim_inode *inode = dentry->d_inode;
1146 if (dentry_has_short_name(dentry))
1147 features->short_names++;
1149 if (inode->i_visited) {
1150 features->hard_links++;
1152 inode_tally_features(inode, features);
1153 inode->i_visited = 1;
1157 /* Tally the features necessary to extract the specified dentries. */
1159 dentry_list_get_features(struct list_head *dentry_list,
1160 struct wim_features *features)
1162 struct wim_dentry *dentry;
1164 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1165 dentry_tally_features(dentry, features);
1167 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1168 dentry->d_inode->i_visited = 0;
1172 do_feature_check(const struct wim_features *required_features,
1173 const struct wim_features *supported_features,
1176 /* File attributes. */
1177 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1178 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1179 * We're an archive program, so theoretically we can do what we
1182 if (required_features->hidden_files &&
1183 !supported_features->hidden_files)
1184 WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1185 required_features->hidden_files);
1187 if (required_features->system_files &&
1188 !supported_features->system_files)
1189 WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1190 required_features->system_files);
1192 if (required_features->compressed_files &&
1193 !supported_features->compressed_files)
1194 WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1195 required_features->compressed_files);
1197 if (required_features->not_context_indexed_files &&
1198 !supported_features->not_context_indexed_files)
1199 WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1200 required_features->not_context_indexed_files);
1202 if (required_features->sparse_files &&
1203 !supported_features->sparse_files)
1204 WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1205 required_features->sparse_files);
1207 if (required_features->encrypted_directories &&
1208 !supported_features->encrypted_directories)
1209 WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1210 required_features->encrypted_directories);
1213 /* Encrypted files. */
1214 if (required_features->encrypted_files &&
1215 !supported_features->encrypted_files)
1216 WARNING("Ignoring %lu encrypted files",
1217 required_features->encrypted_files);
1219 /* Named data streams. */
1220 if (required_features->named_data_streams &&
1221 (!supported_features->named_data_streams))
1222 WARNING("Ignoring named data streams of %lu files",
1223 required_features->named_data_streams);
1226 if (required_features->hard_links && !supported_features->hard_links)
1227 WARNING("Extracting %lu hard links as independent files",
1228 required_features->hard_links);
1230 /* Symbolic links and reparse points. */
1231 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1232 required_features->symlink_reparse_points &&
1233 !supported_features->symlink_reparse_points &&
1234 !supported_features->reparse_points)
1236 ERROR("Extraction backend does not support symbolic links!");
1237 return WIMLIB_ERR_UNSUPPORTED;
1239 if (required_features->reparse_points &&
1240 !supported_features->reparse_points)
1242 if (supported_features->symlink_reparse_points) {
1243 if (required_features->other_reparse_points) {
1244 WARNING("Ignoring %lu non-symlink/junction "
1245 "reparse point files",
1246 required_features->other_reparse_points);
1249 WARNING("Ignoring %lu reparse point files",
1250 required_features->reparse_points);
1254 /* Security descriptors. */
1255 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1256 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1257 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1258 required_features->security_descriptors &&
1259 !supported_features->security_descriptors)
1261 ERROR("Extraction backend does not support security descriptors!");
1262 return WIMLIB_ERR_UNSUPPORTED;
1264 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1265 required_features->security_descriptors &&
1266 !supported_features->security_descriptors)
1267 WARNING("Ignoring Windows NT security descriptors of %lu files",
1268 required_features->security_descriptors);
1271 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1272 required_features->unix_data && !supported_features->unix_data)
1274 ERROR("Extraction backend does not support UNIX data!");
1275 return WIMLIB_ERR_UNSUPPORTED;
1278 if (required_features->unix_data &&
1279 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1281 WARNING("Ignoring UNIX metadata of %lu files",
1282 required_features->unix_data);
1286 if (required_features->short_names &&
1287 !supported_features->short_names)
1289 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1290 ERROR("Extraction backend does not support DOS names!");
1291 return WIMLIB_ERR_UNSUPPORTED;
1293 WARNING("Ignoring DOS names of %lu files",
1294 required_features->short_names);
1298 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1299 !supported_features->timestamps)
1301 ERROR("Extraction backend does not support timestamps!");
1302 return WIMLIB_ERR_UNSUPPORTED;
1308 static const struct apply_operations *
1309 select_apply_operations(int extract_flags)
1312 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1313 return &ntfs_3g_apply_ops;
1316 return &win32_apply_ops;
1318 return &unix_apply_ops;
1323 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1324 const tchar *target, int extract_flags)
1326 const struct apply_operations *ops;
1327 struct apply_ctx *ctx;
1329 LIST_HEAD(dentry_list);
1331 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1332 ret = extract_dentries_to_stdout(trees, num_trees,
1337 num_trees = remove_duplicate_trees(trees, num_trees);
1338 num_trees = remove_contained_trees(trees, num_trees);
1340 ops = select_apply_operations(extract_flags);
1342 if (num_trees > 1 && ops->single_tree_only) {
1343 ERROR("Extracting multiple directory trees "
1344 "at once is not supported in %s extraction mode!",
1346 ret = WIMLIB_ERR_UNSUPPORTED;
1350 ctx = CALLOC(1, ops->context_size);
1352 ret = WIMLIB_ERR_NOMEM;
1357 ctx->target = target;
1358 ctx->target_nchars = tstrlen(target);
1359 ctx->extract_flags = extract_flags;
1360 if (ctx->wim->progfunc) {
1361 ctx->progfunc = ctx->wim->progfunc;
1362 ctx->progctx = ctx->wim->progctx;
1363 ctx->progress.extract.image = wim->current_image;
1364 ctx->progress.extract.extract_flags = (extract_flags &
1365 WIMLIB_EXTRACT_MASK_PUBLIC);
1366 ctx->progress.extract.wimfile_name = wim->filename;
1367 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1368 wim->current_image);
1369 ctx->progress.extract.target = target;
1371 INIT_LIST_HEAD(&ctx->stream_list);
1372 filedes_invalidate(&ctx->tmpfile_fd);
1374 ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1378 build_dentry_list(&dentry_list, trees, num_trees,
1380 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1382 dentry_list_get_features(&dentry_list, &ctx->required_features);
1384 ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1385 ctx->extract_flags);
1389 ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1393 ret = dentry_list_resolve_streams(&dentry_list, ctx);
1397 ret = dentry_list_ref_streams(&dentry_list, ctx);
1401 dentry_list_build_inode_alias_lists(&dentry_list);
1403 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1404 /* When extracting from a pipe, the number of bytes of data to
1405 * extract can't be determined in the normal way (examining the
1406 * lookup table), since at this point all we have is a set of
1407 * SHA1 message digests of streams that need to be extracted.
1408 * However, we can get a reasonably accurate estimate by taking
1409 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1410 * data. This does assume that a full image is being extracted,
1411 * but currently there is no API for doing otherwise. (Also,
1412 * subtract <HARDLINKBYTES> from this if hard links are
1413 * supported by the extraction mode.) */
1414 ctx->progress.extract.total_bytes =
1415 wim_info_get_image_total_bytes(wim->wim_info,
1416 wim->current_image);
1417 if (ctx->supported_features.hard_links) {
1418 ctx->progress.extract.total_bytes -=
1419 wim_info_get_image_hard_link_bytes(wim->wim_info,
1420 wim->current_image);
1424 ret = extract_progress(ctx,
1425 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1426 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1427 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1431 ret = (*ops->extract)(&dentry_list, ctx);
1435 if (ctx->progress.extract.completed_bytes <
1436 ctx->progress.extract.total_bytes)
1438 ctx->progress.extract.completed_bytes =
1439 ctx->progress.extract.total_bytes;
1440 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1445 ret = extract_progress(ctx,
1446 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1447 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1448 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1450 destroy_stream_list(&ctx->stream_list);
1451 destroy_dentry_list(&dentry_list);
1458 mkdir_if_needed(const tchar *target)
1461 if (tstat(target, &stbuf)) {
1462 if (errno == ENOENT) {
1463 if (tmkdir(target, 0755)) {
1464 ERROR_WITH_ERRNO("Failed to create directory "
1465 "\"%"TS"\"", target);
1466 return WIMLIB_ERR_MKDIR;
1469 ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target);
1470 return WIMLIB_ERR_STAT;
1472 } else if (!S_ISDIR(stbuf.st_mode)) {
1473 ERROR("\"%"TS"\" is not a directory", target);
1474 return WIMLIB_ERR_NOTDIR;
1479 /* Make sure the extraction flags make sense, and update them if needed. */
1481 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1483 int extract_flags = *extract_flags_p;
1485 /* Check for invalid flag combinations */
1487 if ((extract_flags &
1488 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1489 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1490 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1491 return WIMLIB_ERR_INVALID_PARAM;
1493 if ((extract_flags &
1494 (WIMLIB_EXTRACT_FLAG_RPFIX |
1495 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1496 WIMLIB_EXTRACT_FLAG_NORPFIX))
1497 return WIMLIB_ERR_INVALID_PARAM;
1499 #ifndef WITH_NTFS_3G
1500 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1501 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
1502 " it cannot apply a WIM image directly to an NTFS volume.");
1503 return WIMLIB_ERR_UNSUPPORTED;
1508 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1509 ERROR("WIMBoot extraction is only supported on Windows!");
1510 return WIMLIB_ERR_UNSUPPORTED;
1514 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1515 WIMLIB_EXTRACT_FLAG_NORPFIX |
1516 WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1517 WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1519 /* For full-image extraction, do reparse point fixups by default
1520 * if the WIM header says they are enabled. */
1521 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1522 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1525 *extract_flags_p = extract_flags;
1530 get_wildcard_flags(int extract_flags)
1532 u32 wildcard_flags = 0;
1534 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB)
1535 wildcard_flags |= WILDCARD_FLAG_ERROR_IF_NO_MATCH;
1537 wildcard_flags |= WILDCARD_FLAG_WARN_IF_NO_MATCH;
1539 if (default_ignore_case)
1540 wildcard_flags |= WILDCARD_FLAG_CASE_INSENSITIVE;
1542 return wildcard_flags;
1545 struct append_dentry_ctx {
1546 struct wim_dentry **dentries;
1547 size_t num_dentries;
1548 size_t num_alloc_dentries;
1552 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1554 struct append_dentry_ctx *ctx = _ctx;
1556 if (ctx->num_dentries == ctx->num_alloc_dentries) {
1557 struct wim_dentry **new_dentries;
1560 new_length = max(ctx->num_alloc_dentries + 8,
1561 ctx->num_alloc_dentries * 3 / 2);
1562 new_dentries = REALLOC(ctx->dentries,
1563 new_length * sizeof(ctx->dentries[0]));
1564 if (new_dentries == NULL)
1565 return WIMLIB_ERR_NOMEM;
1566 ctx->dentries = new_dentries;
1567 ctx->num_alloc_dentries = new_length;
1569 ctx->dentries[ctx->num_dentries++] = dentry;
1574 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1575 const tchar * const *paths, size_t num_paths,
1579 struct wim_dentry **trees;
1582 if (wim == NULL || target == NULL || target[0] == T('\0') ||
1583 (num_paths != 0 && paths == NULL))
1584 return WIMLIB_ERR_INVALID_PARAM;
1586 ret = check_extract_flags(wim, &extract_flags);
1590 ret = select_wim_image(wim, image);
1594 ret = wim_checksum_unhashed_streams(wim);
1598 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1599 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1600 (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1602 ret = mkdir_if_needed(target);
1607 if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1609 struct append_dentry_ctx append_dentry_ctx = {
1612 .num_alloc_dentries = 0,
1615 u32 wildcard_flags = get_wildcard_flags(extract_flags);
1617 for (size_t i = 0; i < num_paths; i++) {
1618 tchar *path = canonicalize_wim_path(paths[i]);
1620 ret = WIMLIB_ERR_NOMEM;
1621 trees = append_dentry_ctx.dentries;
1622 goto out_free_trees;
1624 ret = expand_wildcard(wim, path,
1630 trees = append_dentry_ctx.dentries;
1631 goto out_free_trees;
1634 trees = append_dentry_ctx.dentries;
1635 num_trees = append_dentry_ctx.num_dentries;
1637 trees = MALLOC(num_paths * sizeof(trees[0]));
1639 return WIMLIB_ERR_NOMEM;
1641 for (size_t i = 0; i < num_paths; i++) {
1643 tchar *path = canonicalize_wim_path(paths[i]);
1645 ret = WIMLIB_ERR_NOMEM;
1646 goto out_free_trees;
1649 trees[i] = get_dentry(wim, path,
1650 WIMLIB_CASE_PLATFORM_DEFAULT);
1652 if (trees[i] == NULL) {
1653 ERROR("Path \"%"TS"\" does not exist "
1655 paths[i], wim->current_image);
1656 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1657 goto out_free_trees;
1660 num_trees = num_paths;
1663 if (num_trees == 0) {
1665 goto out_free_trees;
1668 ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1675 extract_single_image(WIMStruct *wim, int image,
1676 const tchar *target, int extract_flags)
1678 const tchar *path = WIMLIB_WIM_ROOT_PATH;
1679 extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1680 return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1683 static const tchar * const filename_forbidden_chars =
1692 /* This function checks if it is okay to use a WIM image's name as a directory
1695 image_name_ok_as_dir(const tchar *image_name)
1697 return image_name && *image_name &&
1698 !tstrpbrk(image_name, filename_forbidden_chars) &&
1699 tstrcmp(image_name, T(".")) &&
1700 tstrcmp(image_name, T(".."));
1703 /* Extracts all images from the WIM to the directory @target, with the images
1704 * placed in subdirectories named by their image names. */
1706 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1708 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
1709 size_t output_path_len = tstrlen(target);
1710 tchar buf[output_path_len + 1 + image_name_max_len + 1];
1713 const tchar *image_name;
1715 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1716 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1717 return WIMLIB_ERR_INVALID_PARAM;
1720 ret = mkdir_if_needed(target);
1723 tmemcpy(buf, target, output_path_len);
1724 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1725 for (image = 1; image <= wim->hdr.image_count; image++) {
1726 image_name = wimlib_get_image_name(wim, image);
1727 if (image_name_ok_as_dir(image_name)) {
1728 tstrcpy(buf + output_path_len + 1, image_name);
1730 /* Image name is empty or contains forbidden characters.
1731 * Use image number instead. */
1732 tsprintf(buf + output_path_len + 1, T("%d"), image);
1734 ret = extract_single_image(wim, image, buf, extract_flags);
1742 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1745 if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1746 WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1747 WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1748 return WIMLIB_ERR_INVALID_PARAM;
1750 if (image == WIMLIB_ALL_IMAGES)
1751 return extract_all_images(wim, target, extract_flags);
1753 return extract_single_image(wim, image, target, extract_flags);
1757 /****************************************************************************
1759 ****************************************************************************/
1762 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1763 const tchar * const *paths, size_t num_paths,
1766 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1767 return WIMLIB_ERR_INVALID_PARAM;
1769 return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1774 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1775 const tchar *path_list_file, int extract_flags)
1782 ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1784 ERROR("Failed to read path list file \"%"TS"\"",
1789 ret = wimlib_extract_paths(wim, image, target,
1790 (const tchar * const *)paths, num_paths,
1798 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1799 const tchar *image_num_or_name,
1800 const tchar *target,
1802 wimlib_progress_func_t progfunc,
1807 struct filedes *in_fd;
1811 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1812 return WIMLIB_ERR_INVALID_PARAM;
1814 /* Read the WIM header from the pipe and get a WIMStruct to represent
1815 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
1816 * wimlib_open_wim(), getting a WIMStruct in this way will result in
1817 * an empty lookup table, no XML data read, and no filename set. */
1818 ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1823 /* Sanity check to make sure this is a pipable WIM. */
1824 if (pwm->hdr.magic != PWM_MAGIC) {
1825 ERROR("The WIM being read from file descriptor %d "
1826 "is not pipable!", pipe_fd);
1827 ret = WIMLIB_ERR_NOT_PIPABLE;
1828 goto out_wimlib_free;
1831 /* Sanity check to make sure the first part of a pipable split WIM is
1832 * sent over the pipe first. */
1833 if (pwm->hdr.part_number != 1) {
1834 ERROR("The first part of the split WIM must be "
1835 "sent over the pipe first.");
1836 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1837 goto out_wimlib_free;
1840 in_fd = &pwm->in_fd;
1841 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1843 /* As mentioned, the WIMStruct we created from the pipe does not have
1844 * XML data yet. Fix this by reading the extra copy of the XML data
1845 * that directly follows the header in pipable WIMs. (Note: see
1846 * write_pipable_wim() for more details about the format of pipable
1849 struct wim_lookup_table_entry xml_lte;
1850 struct wim_resource_spec xml_rspec;
1851 ret = read_pwm_stream_header(pwm, &xml_lte, &xml_rspec, 0, NULL);
1853 goto out_wimlib_free;
1855 if (!(xml_lte.flags & WIM_RESHDR_FLAG_METADATA))
1857 ERROR("Expected XML data, but found non-metadata "
1859 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1860 goto out_wimlib_free;
1863 wim_res_spec_to_hdr(&xml_rspec, &pwm->hdr.xml_data_reshdr);
1865 ret = read_wim_xml_data(pwm);
1867 goto out_wimlib_free;
1869 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
1870 ERROR("Image count in XML data is not the same as in WIM header.");
1871 ret = WIMLIB_ERR_IMAGE_COUNT;
1872 goto out_wimlib_free;
1876 /* Get image index (this may use the XML data that was just read to
1877 * resolve an image name). */
1878 if (image_num_or_name) {
1879 image = wimlib_resolve_image(pwm, image_num_or_name);
1880 if (image == WIMLIB_NO_IMAGE) {
1881 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
1883 ret = WIMLIB_ERR_INVALID_IMAGE;
1884 goto out_wimlib_free;
1885 } else if (image == WIMLIB_ALL_IMAGES) {
1886 ERROR("Applying all images from a pipe is not supported!");
1887 ret = WIMLIB_ERR_INVALID_IMAGE;
1888 goto out_wimlib_free;
1891 if (pwm->hdr.image_count != 1) {
1892 ERROR("No image was specified, but the pipable WIM "
1893 "did not contain exactly 1 image");
1894 ret = WIMLIB_ERR_INVALID_IMAGE;
1895 goto out_wimlib_free;
1900 /* Load the needed metadata resource. */
1901 for (i = 1; i <= pwm->hdr.image_count; i++) {
1902 struct wim_lookup_table_entry *metadata_lte;
1903 struct wim_image_metadata *imd;
1904 struct wim_resource_spec *metadata_rspec;
1906 metadata_lte = new_lookup_table_entry();
1907 if (metadata_lte == NULL) {
1908 ret = WIMLIB_ERR_NOMEM;
1909 goto out_wimlib_free;
1911 metadata_rspec = MALLOC(sizeof(struct wim_resource_spec));
1912 if (metadata_rspec == NULL) {
1913 ret = WIMLIB_ERR_NOMEM;
1914 free_lookup_table_entry(metadata_lte);
1915 goto out_wimlib_free;
1918 ret = read_pwm_stream_header(pwm, metadata_lte, metadata_rspec, 0, NULL);
1919 imd = pwm->image_metadata[i - 1];
1920 imd->metadata_lte = metadata_lte;
1922 FREE(metadata_rspec);
1923 goto out_wimlib_free;
1926 if (!(metadata_lte->flags & WIM_RESHDR_FLAG_METADATA)) {
1927 ERROR("Expected metadata resource, but found "
1928 "non-metadata stream.");
1929 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1930 goto out_wimlib_free;
1934 /* Metadata resource is for the image being extracted.
1935 * Parse it and save the metadata in memory. */
1936 ret = read_metadata_resource(pwm, imd);
1938 goto out_wimlib_free;
1941 /* Metadata resource is not for the image being
1942 * extracted. Skip over it. */
1943 ret = skip_wim_stream(metadata_lte);
1945 goto out_wimlib_free;
1948 /* Extract the image. */
1949 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
1950 ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
1951 /* Clean up and return. */
1959 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
1960 const tchar *target, int extract_flags)
1962 return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
1971 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1974 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1975 return WIMLIB_ERR_INVALID_PARAM;
1976 return do_wimlib_extract_image(wim, image, target, extract_flags);