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 free software; you can redistribute it and/or modify it under
12 * the terms of the GNU Lesser General Public License as published by the Free
13 * Software Foundation; either version 3 of the License, or (at your option) any
16 * This file is distributed in the hope that it will be useful, but WITHOUT
17 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
18 * FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
21 * You should have received a copy of the GNU Lesser General Public License
22 * along with this file; if not, see http://www.gnu.org/licenses/.
26 * This file provides the API functions wimlib_extract_image(),
27 * wimlib_extract_image_from_pipe(), wimlib_extract_paths(), and
28 * wimlib_extract_pathlist(). Internally, all end up calling
29 * do_wimlib_extract_paths() and extract_trees().
31 * Although wimlib supports multiple extraction modes/backends (NTFS-3g, UNIX,
32 * Win32), this file does not itself have code to extract files or directories
33 * to any specific target; instead, it handles generic functionality and relies
34 * on lower-level callback functions declared in `struct apply_operations' to do
35 * the actual extraction.
42 #include "wimlib/apply.h"
43 #include "wimlib/assert.h"
44 #include "wimlib/dentry.h"
45 #include "wimlib/encoding.h"
46 #include "wimlib/endianness.h"
47 #include "wimlib/error.h"
48 #include "wimlib/lookup_table.h"
49 #include "wimlib/metadata.h"
50 #include "wimlib/pathlist.h"
51 #include "wimlib/paths.h"
52 #include "wimlib/reparse.h"
53 #include "wimlib/resource.h"
54 #include "wimlib/security.h"
55 #include "wimlib/unix_data.h"
57 # include "wimlib/win32.h" /* for realpath() equivalent */
59 #include "wimlib/xml.h"
60 #include "wimlib/wildcard.h"
61 #include "wimlib/wim.h"
69 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x80000000
70 #define WIMLIB_EXTRACT_FLAG_IMAGEMODE 0x40000000
72 /* Keep in sync with wimlib.h */
73 #define WIMLIB_EXTRACT_MASK_PUBLIC \
74 (WIMLIB_EXTRACT_FLAG_NTFS | \
75 WIMLIB_EXTRACT_FLAG_UNIX_DATA | \
76 WIMLIB_EXTRACT_FLAG_NO_ACLS | \
77 WIMLIB_EXTRACT_FLAG_STRICT_ACLS | \
78 WIMLIB_EXTRACT_FLAG_RPFIX | \
79 WIMLIB_EXTRACT_FLAG_NORPFIX | \
80 WIMLIB_EXTRACT_FLAG_TO_STDOUT | \
81 WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES | \
82 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS | \
83 WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS | \
84 WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES | \
85 WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS | \
86 WIMLIB_EXTRACT_FLAG_GLOB_PATHS | \
87 WIMLIB_EXTRACT_FLAG_STRICT_GLOB | \
88 WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES | \
89 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE | \
90 WIMLIB_EXTRACT_FLAG_WIMBOOT)
92 /* Send WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE or
93 * WIMLIB_PROGRESS_MSG_EXTRACT_METADATA. */
95 do_file_extract_progress(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
97 ctx->count_until_file_progress = 500; /* Arbitrary value to limit calls */
98 return extract_progress(ctx, msg);
102 start_file_phase(struct apply_ctx *ctx, uint64_t end_file_count, enum wimlib_progress_msg msg)
104 ctx->progress.extract.current_file_count = 0;
105 ctx->progress.extract.end_file_count = end_file_count;
106 return do_file_extract_progress(ctx, msg);
110 start_file_structure_phase(struct apply_ctx *ctx, uint64_t end_file_count)
112 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
116 start_file_metadata_phase(struct apply_ctx *ctx, uint64_t end_file_count)
118 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
122 end_file_phase(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
124 ctx->progress.extract.current_file_count = ctx->progress.extract.end_file_count;
125 return do_file_extract_progress(ctx, msg);
129 end_file_structure_phase(struct apply_ctx *ctx)
131 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
135 end_file_metadata_phase(struct apply_ctx *ctx)
137 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
140 /* Check whether the extraction of a dentry should be skipped completely. */
142 dentry_is_supported(struct wim_dentry *dentry,
143 const struct wim_features *supported_features)
145 struct wim_inode *inode = dentry->d_inode;
147 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
148 return supported_features->reparse_points ||
149 (inode_is_symlink(inode) &&
150 supported_features->symlink_reparse_points);
152 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
153 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
154 return supported_features->encrypted_directories != 0;
156 return supported_features->encrypted_files != 0;
162 #define PWM_ALLOW_WIM_HDR 0x00001
164 /* Read the header from a stream in a pipable WIM. */
166 read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte,
167 struct wim_resource_spec *rspec,
168 int flags, struct wim_header_disk *hdr_ret)
171 struct pwm_stream_hdr stream_hdr;
172 struct wim_header_disk pwm_hdr;
174 struct wim_reshdr reshdr;
177 ret = full_read(&pwm->in_fd, &buf.stream_hdr, sizeof(buf.stream_hdr));
181 if ((flags & PWM_ALLOW_WIM_HDR) &&
182 le64_to_cpu(buf.stream_hdr.magic) == PWM_MAGIC)
184 BUILD_BUG_ON(sizeof(buf.pwm_hdr) < sizeof(buf.stream_hdr));
185 ret = full_read(&pwm->in_fd, &buf.stream_hdr + 1,
186 sizeof(buf.pwm_hdr) - sizeof(buf.stream_hdr));
190 lte->resource_location = RESOURCE_NONEXISTENT;
191 memcpy(hdr_ret, &buf.pwm_hdr, sizeof(buf.pwm_hdr));
195 if (le64_to_cpu(buf.stream_hdr.magic) != PWM_STREAM_MAGIC) {
196 ERROR("Data read on pipe is invalid (expected stream header).");
197 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
200 copy_hash(lte->hash, buf.stream_hdr.hash);
202 reshdr.size_in_wim = 0;
203 reshdr.flags = le32_to_cpu(buf.stream_hdr.flags);
204 reshdr.offset_in_wim = pwm->in_fd.offset;
205 reshdr.uncompressed_size = le64_to_cpu(buf.stream_hdr.uncompressed_size);
206 wim_res_hdr_to_spec(&reshdr, pwm, rspec);
207 lte_bind_wim_resource_spec(lte, rspec);
208 lte->flags = rspec->flags;
209 lte->size = rspec->uncompressed_size;
210 lte->offset_in_res = 0;
214 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
219 load_streams_from_pipe(struct apply_ctx *ctx,
220 const struct read_stream_list_callbacks *cbs)
222 struct wim_lookup_table_entry *found_lte = NULL;
223 struct wim_resource_spec *rspec = NULL;
224 struct wim_lookup_table *lookup_table;
227 ret = WIMLIB_ERR_NOMEM;
228 found_lte = new_lookup_table_entry();
232 rspec = MALLOC(sizeof(struct wim_resource_spec));
236 lookup_table = ctx->wim->lookup_table;
237 memcpy(ctx->progress.extract.guid, ctx->wim->hdr.guid, WIM_GUID_LEN);
238 ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
239 ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
240 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
244 while (ctx->num_streams_remaining) {
245 struct wim_header_disk pwm_hdr;
246 struct wim_lookup_table_entry *needed_lte;
248 if (found_lte->resource_location != RESOURCE_NONEXISTENT)
249 lte_unbind_wim_resource_spec(found_lte);
250 ret = read_pwm_stream_header(ctx->wim, found_lte, rspec,
251 PWM_ALLOW_WIM_HDR, &pwm_hdr);
255 if ((found_lte->resource_location != RESOURCE_NONEXISTENT)
256 && !(found_lte->flags & WIM_RESHDR_FLAG_METADATA)
257 && (needed_lte = lookup_stream(lookup_table, found_lte->hash))
258 && (needed_lte->out_refcnt))
260 needed_lte->offset_in_res = found_lte->offset_in_res;
261 needed_lte->flags = found_lte->flags;
262 needed_lte->size = found_lte->size;
264 lte_unbind_wim_resource_spec(found_lte);
265 lte_bind_wim_resource_spec(needed_lte, rspec);
267 ret = (*cbs->begin_stream)(needed_lte,
268 cbs->begin_stream_ctx);
270 lte_unbind_wim_resource_spec(needed_lte);
274 ret = extract_stream(needed_lte, needed_lte->size,
276 cbs->consume_chunk_ctx);
278 ret = (*cbs->end_stream)(needed_lte, ret,
279 cbs->end_stream_ctx);
280 lte_unbind_wim_resource_spec(needed_lte);
283 ctx->num_streams_remaining--;
284 } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) {
285 ret = skip_wim_stream(found_lte);
289 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
290 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
292 if (part_number != ctx->progress.extract.part_number ||
293 total_parts != ctx->progress.extract.total_parts ||
294 memcmp(pwm_hdr.guid, ctx->progress.extract.guid,
297 ctx->progress.extract.part_number = part_number;
298 ctx->progress.extract.total_parts = total_parts;
299 memcpy(ctx->progress.extract.guid,
300 pwm_hdr.guid, WIM_GUID_LEN);
301 ret = extract_progress(ctx,
302 WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
310 if (found_lte && found_lte->resource_location != RESOURCE_IN_WIM)
312 free_lookup_table_entry(found_lte);
316 /* Creates a temporary file opened for writing. The open file descriptor is
317 * returned in @fd_ret and its name is returned in @name_ret (dynamically
320 create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
327 name = ttempnam(NULL, T("wimlib"));
329 ERROR_WITH_ERRNO("Failed to create temporary filename");
330 return WIMLIB_ERR_NOMEM;
333 open_flags = O_WRONLY | O_CREAT | O_EXCL | O_BINARY;
335 open_flags |= _O_SHORT_LIVED;
337 raw_fd = topen(name, open_flags, 0600);
340 if (errno == EEXIST) {
344 ERROR_WITH_ERRNO("Failed to create temporary file "
347 return WIMLIB_ERR_OPEN;
350 filedes_init(fd_ret, raw_fd);
356 begin_extract_stream_wrapper(struct wim_lookup_table_entry *lte, void *_ctx)
358 struct apply_ctx *ctx = _ctx;
360 ctx->cur_stream = lte;
361 ctx->cur_stream_offset = 0;
363 if (unlikely(lte->out_refcnt > MAX_OPEN_STREAMS))
364 return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
366 return (*ctx->saved_cbs->begin_stream)(lte, ctx->saved_cbs->begin_stream_ctx);
370 extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx)
372 struct apply_ctx *ctx = _ctx;
373 union wimlib_progress_info *progress = &ctx->progress;
376 ctx->cur_stream_offset += size;
378 if (likely(ctx->supported_features.hard_links)) {
379 progress->extract.completed_bytes +=
380 (u64)size * ctx->cur_stream->out_refcnt;
381 if (ctx->cur_stream_offset == ctx->cur_stream->size)
382 progress->extract.completed_streams += ctx->cur_stream->out_refcnt;
384 const struct stream_owner *owners = stream_owners(ctx->cur_stream);
385 for (u32 i = 0; i < ctx->cur_stream->out_refcnt; i++) {
386 const struct wim_inode *inode = owners[i].inode;
387 const struct wim_dentry *dentry;
389 list_for_each_entry(dentry,
390 &inode->i_extraction_aliases,
391 d_extraction_alias_node)
393 progress->extract.completed_bytes += size;
394 if (ctx->cur_stream_offset == ctx->cur_stream->size)
395 progress->extract.completed_streams++;
399 if (progress->extract.completed_bytes >= ctx->next_progress) {
401 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
405 if (progress->extract.completed_bytes >=
406 progress->extract.total_bytes)
408 ctx->next_progress = UINT64_MAX;
410 /* Send new message as soon as another 1/128 of the
411 * total has been extracted. (Arbitrary number.) */
413 progress->extract.completed_bytes +
414 progress->extract.total_bytes / 128;
416 /* ... Unless that would be more than 5000000 bytes, in
417 * which case send the next after the next 5000000
418 * bytes. (Another arbitrary number.) */
419 if (progress->extract.completed_bytes + 5000000 <
422 progress->extract.completed_bytes + 5000000;
424 /* ... But always send a message as soon as we're
425 * completely done. */
426 if (progress->extract.total_bytes < ctx->next_progress)
427 ctx->next_progress = progress->extract.total_bytes;
431 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
432 /* Just extracting to temporary file for now. */
433 ret = full_write(&ctx->tmpfile_fd, chunk, size);
435 ERROR_WITH_ERRNO("Error writing data to "
436 "temporary file \"%"TS"\"",
441 return (*ctx->saved_cbs->consume_chunk)(chunk, size,
442 ctx->saved_cbs->consume_chunk_ctx);
447 extract_from_tmpfile(const tchar *tmpfile_name, struct apply_ctx *ctx)
449 struct wim_lookup_table_entry tmpfile_lte;
450 struct wim_lookup_table_entry *orig_lte = ctx->cur_stream;
451 const struct read_stream_list_callbacks *cbs = ctx->saved_cbs;
453 const u32 orig_refcnt = orig_lte->out_refcnt;
455 BUILD_BUG_ON(MAX_OPEN_STREAMS < ARRAY_LEN(orig_lte->inline_stream_owners));
457 struct stream_owner *owners = orig_lte->stream_owners;
459 /* Copy the stream's data from the temporary file to each of its
462 * This is executed only in the very uncommon case that a
463 * single-instance stream is being extracted to more than
464 * MAX_OPEN_STREAMS locations! */
466 memcpy(&tmpfile_lte, orig_lte, sizeof(struct wim_lookup_table_entry));
467 tmpfile_lte.resource_location = RESOURCE_IN_FILE_ON_DISK;
468 tmpfile_lte.file_on_disk = ctx->tmpfile_name;
470 for (u32 i = 0; i < orig_refcnt; i++) {
472 /* Note: it usually doesn't matter whether we pass the original
473 * stream entry to callbacks provided by the extraction backend
474 * as opposed to the tmpfile stream entry, since they shouldn't
475 * actually read data from the stream other than through the
476 * read_stream_prefix() call below. But for
477 * WIMLIB_EXTRACT_FLAG_WIMBOOT mode on Windows it does matter
478 * because it needs the original stream location in order to
479 * create the external backing reference. */
481 orig_lte->out_refcnt = 1;
482 orig_lte->inline_stream_owners[0] = owners[i];
484 ret = (*cbs->begin_stream)(orig_lte, cbs->begin_stream_ctx);
488 /* Extra SHA-1 isn't necessary here, but it shouldn't hurt as
489 * this case is very rare anyway. */
490 ret = extract_stream(&tmpfile_lte, tmpfile_lte.size,
492 cbs->consume_chunk_ctx);
494 ret = (*cbs->end_stream)(orig_lte, ret, cbs->end_stream_ctx);
499 orig_lte->out_refcnt = 0;
504 end_extract_stream_wrapper(struct wim_lookup_table_entry *stream,
505 int status, void *_ctx)
507 struct apply_ctx *ctx = _ctx;
509 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
510 filedes_close(&ctx->tmpfile_fd);
512 status = extract_from_tmpfile(ctx->tmpfile_name, ctx);
513 filedes_invalidate(&ctx->tmpfile_fd);
514 tunlink(ctx->tmpfile_name);
515 FREE(ctx->tmpfile_name);
518 return (*ctx->saved_cbs->end_stream)(stream, status,
519 ctx->saved_cbs->end_stream_ctx);
524 * Read the list of single-instance streams to extract and feed their data into
525 * the specified callback functions.
527 * This handles checksumming each stream.
529 * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
531 * This also works if the WIM is being read from a pipe, whereas attempting to
532 * read streams directly (e.g. with read_full_stream_into_buf()) will not.
534 * This also will split up streams that will need to be extracted to more than
535 * MAX_OPEN_STREAMS locations, as measured by the 'out_refcnt' of each stream.
536 * Therefore, the apply_operations implementation need not worry about running
537 * out of file descriptors, unless it might open more than one file descriptor
538 * per nominal destination (e.g. Win32 currently might because the destination
539 * file system might not support hard links).
542 extract_stream_list(struct apply_ctx *ctx,
543 const struct read_stream_list_callbacks *cbs)
545 struct read_stream_list_callbacks wrapper_cbs = {
546 .begin_stream = begin_extract_stream_wrapper,
547 .begin_stream_ctx = ctx,
548 .consume_chunk = extract_chunk_wrapper,
549 .consume_chunk_ctx = ctx,
550 .end_stream = end_extract_stream_wrapper,
551 .end_stream_ctx = ctx,
553 ctx->saved_cbs = cbs;
554 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
555 return load_streams_from_pipe(ctx, &wrapper_cbs);
557 return read_stream_list(&ctx->stream_list,
558 offsetof(struct wim_lookup_table_entry,
560 &wrapper_cbs, VERIFY_STREAM_HASHES);
564 /* Extract a WIM dentry to standard output.
566 * This obviously doesn't make sense in all cases. We return an error if the
567 * dentry does not correspond to a regular file. Otherwise we extract the
568 * unnamed data stream only. */
570 extract_dentry_to_stdout(struct wim_dentry *dentry,
571 const struct wim_lookup_table *lookup_table)
573 struct wim_inode *inode = dentry->d_inode;
574 struct wim_lookup_table_entry *lte;
575 struct filedes _stdout;
577 if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
578 FILE_ATTRIBUTE_DIRECTORY))
580 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
581 "extracted to standard output", dentry_full_path(dentry));
582 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
585 lte = inode_unnamed_lte(inode, lookup_table);
587 const u8 *hash = inode_unnamed_stream_hash(inode);
588 if (!is_zero_hash(hash))
589 return stream_not_found_error(inode, hash);
593 filedes_init(&_stdout, STDOUT_FILENO);
594 return extract_full_stream_to_fd(lte, &_stdout);
598 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
599 const struct wim_lookup_table *lookup_table)
601 for (size_t i = 0; i < num_dentries; i++) {
602 int ret = extract_dentry_to_stdout(dentries[i], lookup_table);
609 /**********************************************************************/
612 * Removes duplicate dentries from the array.
614 * Returns the new number of dentries, packed at the front of the array.
617 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
620 for (i = 0; i < num_trees; i++) {
621 if (!trees[i]->tmp_flag) {
622 /* Found distinct dentry. */
623 trees[i]->tmp_flag = 1;
624 trees[j++] = trees[i];
627 for (i = 0; i < j; i++)
628 trees[i]->tmp_flag = 0;
633 * Remove dentries that are descendants of other dentries in the array.
635 * Returns the new number of dentries, packed at the front of the array.
638 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
641 for (i = 0; i < num_trees; i++)
642 trees[i]->tmp_flag = 1;
643 for (i = 0; i < num_trees; i++) {
644 struct wim_dentry *d = trees[i];
645 while (!dentry_is_root(d)) {
650 trees[j++] = trees[i];
654 trees[i]->tmp_flag = 0;
657 for (i = 0; i < j; i++)
658 trees[i]->tmp_flag = 0;
663 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
665 struct list_head *dentry_list = _dentry_list;
666 list_add_tail(&dentry->d_extraction_list_node, dentry_list);
671 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
673 dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
677 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
679 list_del(&dentry->d_extraction_list_node);
680 dentry_reset_extraction_list_node(dentry);
685 * Build the preliminary list of dentries to be extracted.
687 * The list maintains the invariant that if d1 and d2 are in the list and d1 is
688 * an ancestor of d2, then d1 appears before d2 in the list.
691 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
692 size_t num_trees, bool add_ancestors)
694 INIT_LIST_HEAD(dentry_list);
696 /* Add the trees recursively. */
697 for (size_t i = 0; i < num_trees; i++)
698 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
700 /* If requested, add ancestors of the trees. */
702 for (size_t i = 0; i < num_trees; i++) {
703 struct wim_dentry *dentry = trees[i];
704 struct wim_dentry *ancestor;
705 struct list_head *place_after;
707 if (dentry_is_root(dentry))
710 place_after = dentry_list;
713 ancestor = ancestor->d_parent;
714 if (will_extract_dentry(ancestor)) {
715 place_after = &ancestor->d_extraction_list_node;
718 } while (!dentry_is_root(ancestor));
722 ancestor = ancestor->d_parent;
723 if (will_extract_dentry(ancestor))
725 list_add(&ancestor->d_extraction_list_node, place_after);
726 } while (!dentry_is_root(ancestor));
732 destroy_dentry_list(struct list_head *dentry_list)
734 struct wim_dentry *dentry, *tmp;
735 struct wim_inode *inode;
737 list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
738 inode = dentry->d_inode;
739 dentry_reset_extraction_list_node(dentry);
740 inode->i_visited = 0;
741 inode->i_can_externally_back = 0;
742 if ((void *)dentry->d_extraction_name != (void *)dentry->file_name)
743 FREE(dentry->d_extraction_name);
744 dentry->d_extraction_name = NULL;
745 dentry->d_extraction_name_nchars = 0;
750 destroy_stream_list(struct list_head *stream_list)
752 struct wim_lookup_table_entry *lte;
754 list_for_each_entry(lte, stream_list, extraction_list)
755 if (lte->out_refcnt > ARRAY_LEN(lte->inline_stream_owners))
756 FREE(lte->stream_owners);
760 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
762 static const utf16lechar replacement_char = cpu_to_le16('?');
766 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
772 for (i = 0; i < num_chars; i++) {
775 case cpu_to_le16('\\'):
776 case cpu_to_le16(':'):
777 case cpu_to_le16('*'):
778 case cpu_to_le16('?'):
779 case cpu_to_le16('"'):
780 case cpu_to_le16('<'):
781 case cpu_to_le16('>'):
782 case cpu_to_le16('|'):
784 case cpu_to_le16('/'):
785 case cpu_to_le16('\0'):
787 name[i] = replacement_char;
794 if (name[num_chars - 1] == cpu_to_le16(' ') ||
795 name[num_chars - 1] == cpu_to_le16('.'))
798 name[num_chars - 1] = replacement_char;
807 dentry_calculate_extraction_name(struct wim_dentry *dentry,
808 struct apply_ctx *ctx)
812 if (!dentry_is_supported(dentry, &ctx->supported_features))
815 if (dentry_is_root(dentry))
819 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
820 dentry->d_extraction_name = dentry->file_name;
821 dentry->d_extraction_name_nchars = dentry->file_name_nbytes /
827 if (!ctx->supported_features.case_sensitive_filenames) {
828 struct wim_dentry *other;
829 list_for_each_entry(other, &dentry->d_ci_conflict_list,
832 if (will_extract_dentry(other)) {
833 if (ctx->extract_flags &
834 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
835 WARNING("\"%"TS"\" has the same "
836 "case-insensitive name as "
837 "\"%"TS"\"; extracting "
838 "dummy name instead",
839 dentry_full_path(dentry),
840 dentry_full_path(other));
843 WARNING("Not extracting \"%"TS"\": "
844 "has same case-insensitive "
846 dentry_full_path(dentry),
847 dentry_full_path(other));
854 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
855 ret = utf16le_get_tstr(dentry->file_name,
856 dentry->file_name_nbytes,
857 (const tchar **)&dentry->d_extraction_name,
858 &dentry->d_extraction_name_nchars);
859 dentry->d_extraction_name_nchars /= sizeof(tchar);
862 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
864 WARNING("\"%"TS"\" has an invalid filename "
865 "that is not supported on this platform; "
866 "extracting dummy name instead",
867 dentry_full_path(dentry));
870 WARNING("Not extracting \"%"TS"\": has an invalid filename "
871 "that is not supported on this platform",
872 dentry_full_path(dentry));
879 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
881 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
882 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
884 const tchar *tchar_name;
887 ret = utf16le_get_tstr(utf16_name_copy,
888 dentry->file_name_nbytes,
889 &tchar_name, &tchar_nchars);
893 tchar_nchars /= sizeof(tchar);
895 size_t fixed_name_num_chars = tchar_nchars;
896 tchar fixed_name[tchar_nchars + 50];
898 tmemcpy(fixed_name, tchar_name, tchar_nchars);
899 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
900 T(" (invalid filename #%lu)"),
901 ++ctx->invalid_sequence);
903 utf16le_put_tstr(tchar_name);
905 dentry->d_extraction_name = memdup(fixed_name,
906 2 * fixed_name_num_chars + 2);
907 if (!dentry->d_extraction_name)
908 return WIMLIB_ERR_NOMEM;
909 dentry->d_extraction_name_nchars = fixed_name_num_chars;
914 for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
919 * Calculate the actual filename component at which each WIM dentry will be
920 * extracted, with special handling for dentries that are unsupported by the
921 * extraction backend or have invalid names.
923 * ctx->supported_features must be filled in.
925 * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
928 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
929 struct apply_ctx *ctx)
931 struct list_head *prev, *cur;
933 /* Can't use list_for_each_entry() because a call to
934 * dentry_calculate_extraction_name() may delete the current dentry and
935 * its children from the list. */
939 struct wim_dentry *dentry;
943 if (cur == dentry_list)
946 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
948 ret = dentry_calculate_extraction_name(dentry, ctx);
952 if (prev->next == cur)
955 ; /* Current dentry and its children (which follow in
956 the list) were deleted. prev stays the same. */
962 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
963 struct wim_lookup_table *lookup_table)
965 struct wim_inode *inode = dentry->d_inode;
966 struct wim_lookup_table_entry *lte;
970 /* Special case: when extracting from a pipe, the WIM lookup table is
971 * initially empty, so "resolving" an inode's streams is initially not
972 * possible. However, we still need to keep track of which streams,
973 * identified by SHA1 message digests, need to be extracted, so we
974 * "resolve" the inode's streams anyway by allocating new entries. */
975 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
977 ret = inode_resolve_streams(inode, lookup_table, force);
980 for (u32 i = 0; i <= inode->i_num_ads; i++) {
981 lte = inode_stream_lte_resolved(inode, i);
989 * For each dentry to be extracted, resolve all streams in the corresponding
990 * inode and set 'out_refcnt' in each to 0.
992 * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
995 dentry_list_resolve_streams(struct list_head *dentry_list,
996 struct apply_ctx *ctx)
998 struct wim_dentry *dentry;
1001 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1002 ret = dentry_resolve_streams(dentry,
1004 ctx->wim->lookup_table);
1012 ref_stream(struct wim_lookup_table_entry *lte, unsigned stream_idx,
1013 struct wim_dentry *dentry, struct apply_ctx *ctx)
1015 struct wim_inode *inode = dentry->d_inode;
1016 struct stream_owner *stream_owners;
1021 /* Tally the size only for each extraction of the stream (not hard
1023 if (inode->i_visited && ctx->supported_features.hard_links)
1026 ctx->progress.extract.total_bytes += lte->size;
1027 ctx->progress.extract.total_streams++;
1029 if (inode->i_visited)
1032 /* Add stream to the dentry_list only one time, even if it's going
1033 * to be extracted to multiple inodes. */
1034 if (lte->out_refcnt == 0) {
1035 list_add_tail(<e->extraction_list, &ctx->stream_list);
1036 ctx->num_streams_remaining++;
1039 /* If inode not yet been visited, append it to the stream_owners array. */
1040 if (lte->out_refcnt < ARRAY_LEN(lte->inline_stream_owners)) {
1041 stream_owners = lte->inline_stream_owners;
1043 struct stream_owner *prev_stream_owners;
1044 size_t alloc_stream_owners;
1046 if (lte->out_refcnt == ARRAY_LEN(lte->inline_stream_owners)) {
1047 prev_stream_owners = NULL;
1048 alloc_stream_owners = ARRAY_LEN(lte->inline_stream_owners);
1050 prev_stream_owners = lte->stream_owners;
1051 alloc_stream_owners = lte->alloc_stream_owners;
1054 if (lte->out_refcnt == alloc_stream_owners) {
1055 alloc_stream_owners *= 2;
1056 stream_owners = REALLOC(prev_stream_owners,
1057 alloc_stream_owners *
1058 sizeof(stream_owners[0]));
1060 return WIMLIB_ERR_NOMEM;
1061 if (!prev_stream_owners) {
1062 memcpy(stream_owners,
1063 lte->inline_stream_owners,
1064 sizeof(lte->inline_stream_owners));
1066 lte->stream_owners = stream_owners;
1067 lte->alloc_stream_owners = alloc_stream_owners;
1069 stream_owners = lte->stream_owners;
1071 stream_owners[lte->out_refcnt].inode = inode;
1072 if (stream_idx == 0) {
1073 stream_owners[lte->out_refcnt].stream_name = NULL;
1075 stream_owners[lte->out_refcnt].stream_name =
1076 inode->i_ads_entries[stream_idx - 1].stream_name;
1083 ref_unnamed_stream(struct wim_dentry *dentry, struct apply_ctx *ctx)
1085 struct wim_inode *inode = dentry->d_inode;
1087 unsigned stream_idx;
1088 struct wim_lookup_table_entry *stream;
1090 if (unlikely(inode_is_encrypted_directory(inode)))
1093 if (unlikely(ctx->apply_ops->will_externally_back)) {
1094 ret = (*ctx->apply_ops->will_externally_back)(dentry, ctx);
1096 if (ret) /* Error */
1098 /* Will externally back */
1101 /* Won't externally back */
1104 stream = inode_unnamed_stream_resolved(inode, &stream_idx);
1105 return ref_stream(stream, stream_idx, dentry, ctx);
1109 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1111 struct wim_inode *inode = dentry->d_inode;
1114 /* The unnamed data stream will almost always be extracted, but there
1115 * exist cases in which it won't be. */
1116 ret = ref_unnamed_stream(dentry, ctx);
1120 /* Named data streams will be extracted only if supported in the current
1121 * extraction mode and volume, and to avoid complications, if not doing
1122 * a linked extraction. */
1123 if (ctx->supported_features.named_data_streams) {
1124 for (unsigned i = 0; i < inode->i_num_ads; i++) {
1125 if (!inode->i_ads_entries[i].stream_name_nbytes)
1127 ret = ref_stream(inode->i_ads_entries[i].lte, i + 1,
1133 inode->i_visited = 1;
1138 * For each dentry to be extracted, iterate through the data streams of the
1139 * corresponding inode. For each such stream that is not to be ignored due to
1140 * the supported features or extraction flags, add it to the list of streams to
1141 * be extracted (ctx->stream_list) if not already done so.
1143 * Also builds a mapping from each stream to the inodes referencing it.
1145 * This also initializes the extract progress info with byte and stream
1148 * ctx->supported_features must be filled in.
1150 * Possible error codes: WIMLIB_ERR_NOMEM.
1153 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1155 struct wim_dentry *dentry;
1158 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1159 ret = dentry_ref_streams(dentry, ctx);
1163 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1164 dentry->d_inode->i_visited = 0;
1169 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1171 struct wim_dentry *dentry;
1172 struct wim_inode *inode;
1174 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1175 inode = dentry->d_inode;
1176 if (!inode->i_visited)
1177 INIT_LIST_HEAD(&inode->i_extraction_aliases);
1178 list_add_tail(&dentry->d_extraction_alias_node,
1179 &inode->i_extraction_aliases);
1180 inode->i_visited = 1;
1182 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1183 dentry->d_inode->i_visited = 0;
1187 inode_tally_features(const struct wim_inode *inode,
1188 struct wim_features *features)
1190 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1191 features->archive_files++;
1192 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1193 features->hidden_files++;
1194 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1195 features->system_files++;
1196 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1197 features->compressed_files++;
1198 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1199 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1200 features->encrypted_directories++;
1202 features->encrypted_files++;
1204 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1205 features->not_context_indexed_files++;
1206 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1207 features->sparse_files++;
1208 if (inode_has_named_stream(inode))
1209 features->named_data_streams++;
1210 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1211 features->reparse_points++;
1212 if (inode_is_symlink(inode))
1213 features->symlink_reparse_points++;
1215 features->other_reparse_points++;
1217 if (inode->i_security_id != -1)
1218 features->security_descriptors++;
1219 if (inode_has_unix_data(inode))
1220 features->unix_data++;
1223 /* Tally features necessary to extract a dentry and the corresponding inode. */
1225 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1227 struct wim_inode *inode = dentry->d_inode;
1229 if (dentry_has_short_name(dentry))
1230 features->short_names++;
1232 if (inode->i_visited) {
1233 features->hard_links++;
1235 inode_tally_features(inode, features);
1236 inode->i_visited = 1;
1240 /* Tally the features necessary to extract the specified dentries. */
1242 dentry_list_get_features(struct list_head *dentry_list,
1243 struct wim_features *features)
1245 struct wim_dentry *dentry;
1247 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1248 dentry_tally_features(dentry, features);
1250 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1251 dentry->d_inode->i_visited = 0;
1255 do_feature_check(const struct wim_features *required_features,
1256 const struct wim_features *supported_features,
1259 /* File attributes. */
1260 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1261 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1262 * We're an archive program, so theoretically we can do what we
1265 if (required_features->hidden_files &&
1266 !supported_features->hidden_files)
1267 WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1268 required_features->hidden_files);
1270 if (required_features->system_files &&
1271 !supported_features->system_files)
1272 WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1273 required_features->system_files);
1275 if (required_features->compressed_files &&
1276 !supported_features->compressed_files)
1277 WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1278 required_features->compressed_files);
1280 if (required_features->not_context_indexed_files &&
1281 !supported_features->not_context_indexed_files)
1282 WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1283 required_features->not_context_indexed_files);
1285 if (required_features->sparse_files &&
1286 !supported_features->sparse_files)
1287 WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1288 required_features->sparse_files);
1290 if (required_features->encrypted_directories &&
1291 !supported_features->encrypted_directories)
1292 WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1293 required_features->encrypted_directories);
1296 /* Encrypted files. */
1297 if (required_features->encrypted_files &&
1298 !supported_features->encrypted_files)
1299 WARNING("Ignoring %lu encrypted files",
1300 required_features->encrypted_files);
1302 /* Named data streams. */
1303 if (required_features->named_data_streams &&
1304 (!supported_features->named_data_streams))
1305 WARNING("Ignoring named data streams of %lu files",
1306 required_features->named_data_streams);
1309 if (required_features->hard_links && !supported_features->hard_links)
1310 WARNING("Extracting %lu hard links as independent files",
1311 required_features->hard_links);
1313 /* Symbolic links and reparse points. */
1314 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1315 required_features->symlink_reparse_points &&
1316 !supported_features->symlink_reparse_points &&
1317 !supported_features->reparse_points)
1319 ERROR("Extraction backend does not support symbolic links!");
1320 return WIMLIB_ERR_UNSUPPORTED;
1322 if (required_features->reparse_points &&
1323 !supported_features->reparse_points)
1325 if (supported_features->symlink_reparse_points) {
1326 if (required_features->other_reparse_points) {
1327 WARNING("Ignoring %lu non-symlink/junction "
1328 "reparse point files",
1329 required_features->other_reparse_points);
1332 WARNING("Ignoring %lu reparse point files",
1333 required_features->reparse_points);
1337 /* Security descriptors. */
1338 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1339 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1340 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1341 required_features->security_descriptors &&
1342 !supported_features->security_descriptors)
1344 ERROR("Extraction backend does not support security descriptors!");
1345 return WIMLIB_ERR_UNSUPPORTED;
1347 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1348 required_features->security_descriptors &&
1349 !supported_features->security_descriptors)
1350 WARNING("Ignoring Windows NT security descriptors of %lu files",
1351 required_features->security_descriptors);
1354 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1355 required_features->unix_data && !supported_features->unix_data)
1357 ERROR("Extraction backend does not support UNIX data!");
1358 return WIMLIB_ERR_UNSUPPORTED;
1361 if (required_features->unix_data &&
1362 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1364 WARNING("Ignoring UNIX metadata of %lu files",
1365 required_features->unix_data);
1369 if (required_features->short_names &&
1370 !supported_features->short_names)
1372 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1373 ERROR("Extraction backend does not support DOS names!");
1374 return WIMLIB_ERR_UNSUPPORTED;
1376 WARNING("Ignoring DOS names of %lu files",
1377 required_features->short_names);
1381 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1382 !supported_features->timestamps)
1384 ERROR("Extraction backend does not support timestamps!");
1385 return WIMLIB_ERR_UNSUPPORTED;
1391 static const struct apply_operations *
1392 select_apply_operations(int extract_flags)
1395 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1396 return &ntfs_3g_apply_ops;
1399 return &win32_apply_ops;
1401 return &unix_apply_ops;
1406 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1407 const tchar *target, int extract_flags)
1409 const struct apply_operations *ops;
1410 struct apply_ctx *ctx;
1412 LIST_HEAD(dentry_list);
1414 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1415 ret = extract_dentries_to_stdout(trees, num_trees,
1420 num_trees = remove_duplicate_trees(trees, num_trees);
1421 num_trees = remove_contained_trees(trees, num_trees);
1423 ops = select_apply_operations(extract_flags);
1425 if (num_trees > 1 && ops->single_tree_only) {
1426 ERROR("Extracting multiple directory trees "
1427 "at once is not supported in %s extraction mode!",
1429 ret = WIMLIB_ERR_UNSUPPORTED;
1433 ctx = CALLOC(1, ops->context_size);
1435 ret = WIMLIB_ERR_NOMEM;
1440 ctx->target = target;
1441 ctx->target_nchars = tstrlen(target);
1442 ctx->extract_flags = extract_flags;
1443 if (ctx->wim->progfunc) {
1444 ctx->progfunc = ctx->wim->progfunc;
1445 ctx->progctx = ctx->wim->progctx;
1446 ctx->progress.extract.image = wim->current_image;
1447 ctx->progress.extract.extract_flags = (extract_flags &
1448 WIMLIB_EXTRACT_MASK_PUBLIC);
1449 ctx->progress.extract.wimfile_name = wim->filename;
1450 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1451 wim->current_image);
1452 ctx->progress.extract.target = target;
1454 INIT_LIST_HEAD(&ctx->stream_list);
1455 filedes_invalidate(&ctx->tmpfile_fd);
1456 ctx->apply_ops = ops;
1458 ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1462 build_dentry_list(&dentry_list, trees, num_trees,
1464 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1466 dentry_list_get_features(&dentry_list, &ctx->required_features);
1468 ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1469 ctx->extract_flags);
1473 ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1477 ret = dentry_list_resolve_streams(&dentry_list, ctx);
1481 dentry_list_build_inode_alias_lists(&dentry_list);
1483 ret = dentry_list_ref_streams(&dentry_list, ctx);
1487 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1488 /* When extracting from a pipe, the number of bytes of data to
1489 * extract can't be determined in the normal way (examining the
1490 * lookup table), since at this point all we have is a set of
1491 * SHA1 message digests of streams that need to be extracted.
1492 * However, we can get a reasonably accurate estimate by taking
1493 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1494 * data. This does assume that a full image is being extracted,
1495 * but currently there is no API for doing otherwise. (Also,
1496 * subtract <HARDLINKBYTES> from this if hard links are
1497 * supported by the extraction mode.) */
1498 ctx->progress.extract.total_bytes =
1499 wim_info_get_image_total_bytes(wim->wim_info,
1500 wim->current_image);
1501 if (ctx->supported_features.hard_links) {
1502 ctx->progress.extract.total_bytes -=
1503 wim_info_get_image_hard_link_bytes(wim->wim_info,
1504 wim->current_image);
1508 ret = extract_progress(ctx,
1509 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1510 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1511 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1515 ret = (*ops->extract)(&dentry_list, ctx);
1519 if (ctx->progress.extract.completed_bytes <
1520 ctx->progress.extract.total_bytes)
1522 ctx->progress.extract.completed_bytes =
1523 ctx->progress.extract.total_bytes;
1524 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1529 ret = extract_progress(ctx,
1530 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1531 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1532 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1534 destroy_stream_list(&ctx->stream_list);
1535 destroy_dentry_list(&dentry_list);
1542 mkdir_if_needed(const tchar *target)
1544 if (!tmkdir(target, 0755))
1547 if (errno == EEXIST)
1551 /* _wmkdir() fails with EACCES if called on a drive root directory. */
1552 if (errno == EACCES)
1556 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
1557 return WIMLIB_ERR_MKDIR;
1560 /* Make sure the extraction flags make sense, and update them if needed. */
1562 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1564 int extract_flags = *extract_flags_p;
1566 /* Check for invalid flag combinations */
1568 if ((extract_flags &
1569 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1570 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1571 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1572 return WIMLIB_ERR_INVALID_PARAM;
1574 if ((extract_flags &
1575 (WIMLIB_EXTRACT_FLAG_RPFIX |
1576 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1577 WIMLIB_EXTRACT_FLAG_NORPFIX))
1578 return WIMLIB_ERR_INVALID_PARAM;
1580 #ifndef WITH_NTFS_3G
1581 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1582 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
1583 " it cannot apply a WIM image directly to an NTFS volume.");
1584 return WIMLIB_ERR_UNSUPPORTED;
1588 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1591 return WIMLIB_ERR_NO_FILENAME;
1593 ERROR("WIMBoot extraction is only supported on Windows!");
1594 return WIMLIB_ERR_UNSUPPORTED;
1599 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1600 WIMLIB_EXTRACT_FLAG_NORPFIX |
1601 WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1602 WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1604 /* For full-image extraction, do reparse point fixups by default
1605 * if the WIM header says they are enabled. */
1606 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1607 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1610 *extract_flags_p = extract_flags;
1615 get_wildcard_flags(int extract_flags)
1617 u32 wildcard_flags = 0;
1619 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB)
1620 wildcard_flags |= WILDCARD_FLAG_ERROR_IF_NO_MATCH;
1622 wildcard_flags |= WILDCARD_FLAG_WARN_IF_NO_MATCH;
1624 if (default_ignore_case)
1625 wildcard_flags |= WILDCARD_FLAG_CASE_INSENSITIVE;
1627 return wildcard_flags;
1630 struct append_dentry_ctx {
1631 struct wim_dentry **dentries;
1632 size_t num_dentries;
1633 size_t num_alloc_dentries;
1637 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1639 struct append_dentry_ctx *ctx = _ctx;
1641 if (ctx->num_dentries == ctx->num_alloc_dentries) {
1642 struct wim_dentry **new_dentries;
1645 new_length = max(ctx->num_alloc_dentries + 8,
1646 ctx->num_alloc_dentries * 3 / 2);
1647 new_dentries = REALLOC(ctx->dentries,
1648 new_length * sizeof(ctx->dentries[0]));
1649 if (new_dentries == NULL)
1650 return WIMLIB_ERR_NOMEM;
1651 ctx->dentries = new_dentries;
1652 ctx->num_alloc_dentries = new_length;
1654 ctx->dentries[ctx->num_dentries++] = dentry;
1659 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1660 const tchar * const *paths, size_t num_paths,
1664 struct wim_dentry **trees;
1667 if (wim == NULL || target == NULL || target[0] == T('\0') ||
1668 (num_paths != 0 && paths == NULL))
1669 return WIMLIB_ERR_INVALID_PARAM;
1671 ret = check_extract_flags(wim, &extract_flags);
1675 ret = select_wim_image(wim, image);
1679 ret = wim_checksum_unhashed_streams(wim);
1683 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1684 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1685 (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1687 ret = mkdir_if_needed(target);
1692 if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1694 struct append_dentry_ctx append_dentry_ctx = {
1697 .num_alloc_dentries = 0,
1700 u32 wildcard_flags = get_wildcard_flags(extract_flags);
1702 for (size_t i = 0; i < num_paths; i++) {
1703 tchar *path = canonicalize_wim_path(paths[i]);
1705 ret = WIMLIB_ERR_NOMEM;
1706 trees = append_dentry_ctx.dentries;
1707 goto out_free_trees;
1709 ret = expand_wildcard(wim, path,
1715 trees = append_dentry_ctx.dentries;
1716 goto out_free_trees;
1719 trees = append_dentry_ctx.dentries;
1720 num_trees = append_dentry_ctx.num_dentries;
1722 trees = MALLOC(num_paths * sizeof(trees[0]));
1724 return WIMLIB_ERR_NOMEM;
1726 for (size_t i = 0; i < num_paths; i++) {
1728 tchar *path = canonicalize_wim_path(paths[i]);
1730 ret = WIMLIB_ERR_NOMEM;
1731 goto out_free_trees;
1734 trees[i] = get_dentry(wim, path,
1735 WIMLIB_CASE_PLATFORM_DEFAULT);
1737 if (trees[i] == NULL) {
1738 ERROR("Path \"%"TS"\" does not exist "
1740 paths[i], wim->current_image);
1741 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1742 goto out_free_trees;
1745 num_trees = num_paths;
1748 if (num_trees == 0) {
1750 goto out_free_trees;
1753 ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1760 extract_single_image(WIMStruct *wim, int image,
1761 const tchar *target, int extract_flags)
1763 const tchar *path = WIMLIB_WIM_ROOT_PATH;
1764 extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1765 return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1768 static const tchar * const filename_forbidden_chars =
1777 /* This function checks if it is okay to use a WIM image's name as a directory
1780 image_name_ok_as_dir(const tchar *image_name)
1782 return image_name && *image_name &&
1783 !tstrpbrk(image_name, filename_forbidden_chars) &&
1784 tstrcmp(image_name, T(".")) &&
1785 tstrcmp(image_name, T(".."));
1788 /* Extracts all images from the WIM to the directory @target, with the images
1789 * placed in subdirectories named by their image names. */
1791 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1793 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
1794 size_t output_path_len = tstrlen(target);
1795 tchar buf[output_path_len + 1 + image_name_max_len + 1];
1798 const tchar *image_name;
1800 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1801 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1802 return WIMLIB_ERR_INVALID_PARAM;
1805 ret = mkdir_if_needed(target);
1808 tmemcpy(buf, target, output_path_len);
1809 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1810 for (image = 1; image <= wim->hdr.image_count; image++) {
1811 image_name = wimlib_get_image_name(wim, image);
1812 if (image_name_ok_as_dir(image_name)) {
1813 tstrcpy(buf + output_path_len + 1, image_name);
1815 /* Image name is empty or contains forbidden characters.
1816 * Use image number instead. */
1817 tsprintf(buf + output_path_len + 1, T("%d"), image);
1819 ret = extract_single_image(wim, image, buf, extract_flags);
1827 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1830 if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1831 WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1832 WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1833 return WIMLIB_ERR_INVALID_PARAM;
1835 if (image == WIMLIB_ALL_IMAGES)
1836 return extract_all_images(wim, target, extract_flags);
1838 return extract_single_image(wim, image, target, extract_flags);
1842 /****************************************************************************
1844 ****************************************************************************/
1847 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1848 const tchar * const *paths, size_t num_paths,
1851 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1852 return WIMLIB_ERR_INVALID_PARAM;
1854 return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1859 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1860 const tchar *path_list_file, int extract_flags)
1867 ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1869 ERROR("Failed to read path list file \"%"TS"\"",
1874 ret = wimlib_extract_paths(wim, image, target,
1875 (const tchar * const *)paths, num_paths,
1883 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1884 const tchar *image_num_or_name,
1885 const tchar *target,
1887 wimlib_progress_func_t progfunc,
1892 struct filedes *in_fd;
1896 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1897 return WIMLIB_ERR_INVALID_PARAM;
1899 /* Read the WIM header from the pipe and get a WIMStruct to represent
1900 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
1901 * wimlib_open_wim(), getting a WIMStruct in this way will result in
1902 * an empty lookup table, no XML data read, and no filename set. */
1903 ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1908 /* Sanity check to make sure this is a pipable WIM. */
1909 if (pwm->hdr.magic != PWM_MAGIC) {
1910 ERROR("The WIM being read from file descriptor %d "
1911 "is not pipable!", pipe_fd);
1912 ret = WIMLIB_ERR_NOT_PIPABLE;
1913 goto out_wimlib_free;
1916 /* Sanity check to make sure the first part of a pipable split WIM is
1917 * sent over the pipe first. */
1918 if (pwm->hdr.part_number != 1) {
1919 ERROR("The first part of the split WIM must be "
1920 "sent over the pipe first.");
1921 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1922 goto out_wimlib_free;
1925 in_fd = &pwm->in_fd;
1926 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1928 /* As mentioned, the WIMStruct we created from the pipe does not have
1929 * XML data yet. Fix this by reading the extra copy of the XML data
1930 * that directly follows the header in pipable WIMs. (Note: see
1931 * write_pipable_wim() for more details about the format of pipable
1934 struct wim_lookup_table_entry xml_lte;
1935 struct wim_resource_spec xml_rspec;
1936 ret = read_pwm_stream_header(pwm, &xml_lte, &xml_rspec, 0, NULL);
1938 goto out_wimlib_free;
1940 if (!(xml_lte.flags & WIM_RESHDR_FLAG_METADATA))
1942 ERROR("Expected XML data, but found non-metadata "
1944 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1945 goto out_wimlib_free;
1948 wim_res_spec_to_hdr(&xml_rspec, &pwm->hdr.xml_data_reshdr);
1950 ret = read_wim_xml_data(pwm);
1952 goto out_wimlib_free;
1954 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
1955 ERROR("Image count in XML data is not the same as in WIM header.");
1956 ret = WIMLIB_ERR_IMAGE_COUNT;
1957 goto out_wimlib_free;
1961 /* Get image index (this may use the XML data that was just read to
1962 * resolve an image name). */
1963 if (image_num_or_name) {
1964 image = wimlib_resolve_image(pwm, image_num_or_name);
1965 if (image == WIMLIB_NO_IMAGE) {
1966 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
1968 ret = WIMLIB_ERR_INVALID_IMAGE;
1969 goto out_wimlib_free;
1970 } else if (image == WIMLIB_ALL_IMAGES) {
1971 ERROR("Applying all images from a pipe is not supported!");
1972 ret = WIMLIB_ERR_INVALID_IMAGE;
1973 goto out_wimlib_free;
1976 if (pwm->hdr.image_count != 1) {
1977 ERROR("No image was specified, but the pipable WIM "
1978 "did not contain exactly 1 image");
1979 ret = WIMLIB_ERR_INVALID_IMAGE;
1980 goto out_wimlib_free;
1985 /* Load the needed metadata resource. */
1986 for (i = 1; i <= pwm->hdr.image_count; i++) {
1987 struct wim_lookup_table_entry *metadata_lte;
1988 struct wim_image_metadata *imd;
1989 struct wim_resource_spec *metadata_rspec;
1991 metadata_lte = new_lookup_table_entry();
1992 if (metadata_lte == NULL) {
1993 ret = WIMLIB_ERR_NOMEM;
1994 goto out_wimlib_free;
1996 metadata_rspec = MALLOC(sizeof(struct wim_resource_spec));
1997 if (metadata_rspec == NULL) {
1998 ret = WIMLIB_ERR_NOMEM;
1999 free_lookup_table_entry(metadata_lte);
2000 goto out_wimlib_free;
2003 ret = read_pwm_stream_header(pwm, metadata_lte, metadata_rspec, 0, NULL);
2004 imd = pwm->image_metadata[i - 1];
2005 imd->metadata_lte = metadata_lte;
2007 FREE(metadata_rspec);
2008 goto out_wimlib_free;
2011 if (!(metadata_lte->flags & WIM_RESHDR_FLAG_METADATA)) {
2012 ERROR("Expected metadata resource, but found "
2013 "non-metadata stream.");
2014 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2015 goto out_wimlib_free;
2019 /* Metadata resource is for the image being extracted.
2020 * Parse it and save the metadata in memory. */
2021 ret = read_metadata_resource(imd);
2023 goto out_wimlib_free;
2026 /* Metadata resource is not for the image being
2027 * extracted. Skip over it. */
2028 ret = skip_wim_stream(metadata_lte);
2030 goto out_wimlib_free;
2033 /* Extract the image. */
2034 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2035 ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
2036 /* Clean up and return. */
2044 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2045 const tchar *target, int extract_flags)
2047 return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
2056 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
2059 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
2060 return WIMLIB_ERR_INVALID_PARAM;
2061 return do_wimlib_extract_image(wim, image, target, extract_flags);