4 * Support for extracting WIM images, or files or directories contained in a WIM
9 * Copyright (C) 2012, 2013, 2014, 2015 Eric Biggers
11 * This file is free software; you can redistribute it and/or modify it under
12 * the terms of the GNU Lesser General Public License as published by the Free
13 * Software Foundation; either version 3 of the License, or (at your option) any
16 * This file is distributed in the hope that it will be useful, but WITHOUT
17 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
18 * FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
21 * You should have received a copy of the GNU Lesser General Public License
22 * along with this file; if not, see http://www.gnu.org/licenses/.
26 * This file provides the API functions wimlib_extract_image(),
27 * wimlib_extract_image_from_pipe(), wimlib_extract_paths(), and
28 * wimlib_extract_pathlist(). Internally, all end up calling
29 * do_wimlib_extract_paths() and extract_trees().
31 * Although wimlib supports multiple extraction modes/backends (NTFS-3g, UNIX,
32 * Win32), this file does not itself have code to extract files or directories
33 * to any specific target; instead, it handles generic functionality and relies
34 * on lower-level callback functions declared in `struct apply_operations' to do
35 * the actual extraction.
47 #include "wimlib/apply.h"
48 #include "wimlib/assert.h"
49 #include "wimlib/blob_table.h"
50 #include "wimlib/dentry.h"
51 #include "wimlib/encoding.h"
52 #include "wimlib/endianness.h"
53 #include "wimlib/error.h"
54 #include "wimlib/metadata.h"
55 #include "wimlib/pathlist.h"
56 #include "wimlib/paths.h"
57 #include "wimlib/reparse.h"
58 #include "wimlib/resource.h"
59 #include "wimlib/security.h"
60 #include "wimlib/unix_data.h"
61 #include "wimlib/wildcard.h"
62 #include "wimlib/wim.h"
63 #include "wimlib/win32.h" /* for realpath() equivalent */
64 #include "wimlib/xml.h"
66 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x80000000
67 #define WIMLIB_EXTRACT_FLAG_IMAGEMODE 0x40000000
69 /* Keep in sync with wimlib.h */
70 #define WIMLIB_EXTRACT_MASK_PUBLIC \
71 (WIMLIB_EXTRACT_FLAG_NTFS | \
72 WIMLIB_EXTRACT_FLAG_UNIX_DATA | \
73 WIMLIB_EXTRACT_FLAG_NO_ACLS | \
74 WIMLIB_EXTRACT_FLAG_STRICT_ACLS | \
75 WIMLIB_EXTRACT_FLAG_RPFIX | \
76 WIMLIB_EXTRACT_FLAG_NORPFIX | \
77 WIMLIB_EXTRACT_FLAG_TO_STDOUT | \
78 WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES | \
79 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS | \
80 WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS | \
81 WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES | \
82 WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS | \
83 WIMLIB_EXTRACT_FLAG_GLOB_PATHS | \
84 WIMLIB_EXTRACT_FLAG_STRICT_GLOB | \
85 WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES | \
86 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE | \
87 WIMLIB_EXTRACT_FLAG_WIMBOOT)
89 /* Send WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE or
90 * WIMLIB_PROGRESS_MSG_EXTRACT_METADATA. */
92 do_file_extract_progress(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
94 ctx->count_until_file_progress = 500; /* Arbitrary value to limit calls */
95 return extract_progress(ctx, msg);
99 start_file_phase(struct apply_ctx *ctx, u64 end_file_count, enum wimlib_progress_msg msg)
101 ctx->progress.extract.current_file_count = 0;
102 ctx->progress.extract.end_file_count = end_file_count;
103 return do_file_extract_progress(ctx, msg);
107 start_file_structure_phase(struct apply_ctx *ctx, u64 end_file_count)
109 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
113 start_file_metadata_phase(struct apply_ctx *ctx, u64 end_file_count)
115 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
119 end_file_phase(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
121 ctx->progress.extract.current_file_count = ctx->progress.extract.end_file_count;
122 return do_file_extract_progress(ctx, msg);
126 end_file_structure_phase(struct apply_ctx *ctx)
128 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
132 end_file_metadata_phase(struct apply_ctx *ctx)
134 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
137 #define PWM_ALLOW_WIM_HDR 0x00001
139 /* Read the header for a blob in a pipable WIM. */
141 read_pwm_blob_header(WIMStruct *pwm, struct blob_descriptor *blob,
142 struct wim_resource_descriptor *rdesc,
143 int flags, struct wim_header_disk *hdr_ret)
146 struct pwm_blob_hdr blob_hdr;
147 struct wim_header_disk pwm_hdr;
149 struct wim_reshdr reshdr;
152 ret = full_read(&pwm->in_fd, &buf.blob_hdr, sizeof(buf.blob_hdr));
156 if ((flags & PWM_ALLOW_WIM_HDR) &&
157 le64_to_cpu(buf.blob_hdr.magic) == PWM_MAGIC)
159 BUILD_BUG_ON(sizeof(buf.pwm_hdr) < sizeof(buf.blob_hdr));
160 ret = full_read(&pwm->in_fd, &buf.blob_hdr + 1,
161 sizeof(buf.pwm_hdr) - sizeof(buf.blob_hdr));
165 blob->blob_location = BLOB_NONEXISTENT;
166 memcpy(hdr_ret, &buf.pwm_hdr, sizeof(buf.pwm_hdr));
170 if (le64_to_cpu(buf.blob_hdr.magic) != PWM_BLOB_MAGIC) {
171 ERROR("Data read on pipe is invalid (expected blob header).");
172 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
175 copy_hash(blob->hash, buf.blob_hdr.hash);
177 reshdr.size_in_wim = 0;
178 reshdr.flags = le32_to_cpu(buf.blob_hdr.flags);
179 reshdr.offset_in_wim = pwm->in_fd.offset;
180 reshdr.uncompressed_size = le64_to_cpu(buf.blob_hdr.uncompressed_size);
181 wim_res_hdr_to_desc(&reshdr, pwm, rdesc);
182 blob_set_is_located_in_nonsolid_wim_resource(blob, rdesc);
183 blob->is_metadata = (rdesc->flags & WIM_RESHDR_FLAG_METADATA) != 0;
185 if (unlikely(blob->size == 0))
186 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
191 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
196 read_blobs_from_pipe(struct apply_ctx *ctx,
197 const struct read_blob_list_callbacks *cbs)
199 struct blob_descriptor *found_blob = NULL;
200 struct wim_resource_descriptor *rdesc = NULL;
201 struct blob_table *blob_table;
204 ret = WIMLIB_ERR_NOMEM;
205 found_blob = new_blob_descriptor();
209 rdesc = MALLOC(sizeof(struct wim_resource_descriptor));
213 blob_table = ctx->wim->blob_table;
214 memcpy(ctx->progress.extract.guid, ctx->wim->hdr.guid, WIM_GUID_LEN);
215 ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
216 ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
217 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
221 while (ctx->num_blobs_remaining) {
222 struct wim_header_disk pwm_hdr;
223 struct blob_descriptor *needed_blob;
225 if (found_blob->blob_location != BLOB_NONEXISTENT)
226 blob_unset_is_located_in_wim_resource(found_blob);
227 ret = read_pwm_blob_header(ctx->wim, found_blob, rdesc,
228 PWM_ALLOW_WIM_HDR, &pwm_hdr);
232 if ((found_blob->blob_location != BLOB_NONEXISTENT)
233 && !found_blob->is_metadata
234 && (needed_blob = lookup_blob(blob_table, found_blob->hash))
235 && (needed_blob->out_refcnt))
237 blob_unset_is_located_in_wim_resource(found_blob);
238 blob_set_is_located_in_nonsolid_wim_resource(needed_blob, rdesc);
240 ret = (*cbs->begin_blob)(needed_blob,
241 cbs->begin_blob_ctx);
243 blob_unset_is_located_in_wim_resource(needed_blob);
247 ret = extract_blob(needed_blob, needed_blob->size,
249 cbs->consume_chunk_ctx);
251 ret = (*cbs->end_blob)(needed_blob, ret,
253 blob_unset_is_located_in_wim_resource(needed_blob);
256 ctx->num_blobs_remaining--;
257 } else if (found_blob->blob_location != BLOB_NONEXISTENT) {
258 ret = skip_wim_resource(found_blob->rdesc);
262 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
263 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
265 if (part_number != ctx->progress.extract.part_number ||
266 total_parts != ctx->progress.extract.total_parts ||
267 memcmp(pwm_hdr.guid, ctx->progress.extract.guid,
270 ctx->progress.extract.part_number = part_number;
271 ctx->progress.extract.total_parts = total_parts;
272 memcpy(ctx->progress.extract.guid,
273 pwm_hdr.guid, WIM_GUID_LEN);
274 ret = extract_progress(ctx,
275 WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
283 if (found_blob && found_blob->blob_location != BLOB_IN_WIM)
285 free_blob_descriptor(found_blob);
289 /* Creates a temporary file opened for writing. The open file descriptor is
290 * returned in @fd_ret and its name is returned in @name_ret (dynamically
293 create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
300 name = ttempnam(NULL, T("wimlib"));
302 ERROR_WITH_ERRNO("Failed to create temporary filename");
303 return WIMLIB_ERR_NOMEM;
306 open_flags = O_WRONLY | O_CREAT | O_EXCL | O_BINARY;
308 open_flags |= _O_SHORT_LIVED;
310 raw_fd = topen(name, open_flags, 0600);
313 if (errno == EEXIST) {
317 ERROR_WITH_ERRNO("Failed to create temporary file "
320 return WIMLIB_ERR_OPEN;
323 filedes_init(fd_ret, raw_fd);
329 begin_extract_blob_wrapper(struct blob_descriptor *blob, void *_ctx)
331 struct apply_ctx *ctx = _ctx;
333 ctx->cur_blob = blob;
334 ctx->cur_blob_offset = 0;
336 if (unlikely(blob->out_refcnt > MAX_OPEN_FILES))
337 return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
339 return (*ctx->saved_cbs->begin_blob)(blob, ctx->saved_cbs->begin_blob_ctx);
343 extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx)
345 struct apply_ctx *ctx = _ctx;
346 union wimlib_progress_info *progress = &ctx->progress;
349 ctx->cur_blob_offset += size;
351 if (likely(ctx->supported_features.hard_links)) {
352 progress->extract.completed_bytes +=
353 (u64)size * ctx->cur_blob->out_refcnt;
354 if (ctx->cur_blob_offset == ctx->cur_blob->size)
355 progress->extract.completed_streams += ctx->cur_blob->out_refcnt;
357 const struct blob_extraction_target *targets =
358 blob_extraction_targets(ctx->cur_blob);
359 for (u32 i = 0; i < ctx->cur_blob->out_refcnt; i++) {
360 const struct wim_inode *inode = targets[i].inode;
361 const struct wim_dentry *dentry;
363 list_for_each_entry(dentry,
364 &inode->i_extraction_aliases,
365 d_extraction_alias_node)
367 progress->extract.completed_bytes += size;
368 if (ctx->cur_blob_offset == ctx->cur_blob->size)
369 progress->extract.completed_streams++;
373 if (progress->extract.completed_bytes >= ctx->next_progress) {
375 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
379 if (progress->extract.completed_bytes >=
380 progress->extract.total_bytes)
382 ctx->next_progress = UINT64_MAX;
384 /* Send new message as soon as another 1/128 of the
385 * total has been extracted. (Arbitrary number.) */
387 progress->extract.completed_bytes +
388 progress->extract.total_bytes / 128;
390 /* ... Unless that would be more than 5000000 bytes, in
391 * which case send the next after the next 5000000
392 * bytes. (Another arbitrary number.) */
393 if (progress->extract.completed_bytes + 5000000 <
396 progress->extract.completed_bytes + 5000000;
398 /* ... But always send a message as soon as we're
399 * completely done. */
400 if (progress->extract.total_bytes < ctx->next_progress)
401 ctx->next_progress = progress->extract.total_bytes;
405 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
406 /* Just extracting to temporary file for now. */
407 ret = full_write(&ctx->tmpfile_fd, chunk, size);
409 ERROR_WITH_ERRNO("Error writing data to "
410 "temporary file \"%"TS"\"",
415 return (*ctx->saved_cbs->consume_chunk)(chunk, size,
416 ctx->saved_cbs->consume_chunk_ctx);
421 extract_from_tmpfile(const tchar *tmpfile_name, struct apply_ctx *ctx)
423 struct blob_descriptor tmpfile_blob;
424 struct blob_descriptor *orig_blob = ctx->cur_blob;
425 const struct read_blob_list_callbacks *cbs = ctx->saved_cbs;
427 const u32 orig_refcnt = orig_blob->out_refcnt;
429 BUILD_BUG_ON(MAX_OPEN_FILES <
430 ARRAY_LEN(orig_blob->inline_blob_extraction_targets));
432 struct blob_extraction_target *targets = orig_blob->blob_extraction_targets;
434 /* Copy the blob's data from the temporary file to each of its targets.
436 * This is executed only in the very uncommon case that a blob is being
437 * extracted to more than MAX_OPEN_FILES targets! */
439 memcpy(&tmpfile_blob, orig_blob, sizeof(struct blob_descriptor));
440 tmpfile_blob.blob_location = BLOB_IN_FILE_ON_DISK;
441 tmpfile_blob.file_on_disk = ctx->tmpfile_name;
443 for (u32 i = 0; i < orig_refcnt; i++) {
445 /* Note: it usually doesn't matter whether we pass the original
446 * blob descriptor to callbacks provided by the extraction
447 * backend as opposed to the tmpfile blob descriptor, since they
448 * shouldn't actually read data from the blob other than through
449 * the read_blob_prefix() call below. But for
450 * WIMLIB_EXTRACT_FLAG_WIMBOOT mode on Windows it does matter
451 * because it needs access to the original WIM resource
452 * descriptor in order to create the external backing reference.
455 orig_blob->out_refcnt = 1;
456 orig_blob->inline_blob_extraction_targets[0] = targets[i];
458 ret = (*cbs->begin_blob)(orig_blob, cbs->begin_blob_ctx);
462 /* Extra SHA-1 isn't necessary here, but it shouldn't hurt as
463 * this case is very rare anyway. */
464 ret = extract_blob(&tmpfile_blob, tmpfile_blob.size,
466 cbs->consume_chunk_ctx);
468 ret = (*cbs->end_blob)(orig_blob, ret, cbs->end_blob_ctx);
473 orig_blob->out_refcnt = 0;
478 end_extract_blob_wrapper(struct blob_descriptor *blob, int status, void *_ctx)
480 struct apply_ctx *ctx = _ctx;
482 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
483 filedes_close(&ctx->tmpfile_fd);
485 status = extract_from_tmpfile(ctx->tmpfile_name, ctx);
486 filedes_invalidate(&ctx->tmpfile_fd);
487 tunlink(ctx->tmpfile_name);
488 FREE(ctx->tmpfile_name);
491 return (*ctx->saved_cbs->end_blob)(blob, status,
492 ctx->saved_cbs->end_blob_ctx);
497 * Read the list of blobs to extract and feed their data into the specified
498 * callback functions.
500 * This handles checksumming each blob.
502 * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
504 * This also works if the WIM is being read from a pipe, whereas attempting to
505 * read blobs directly (e.g. with read_full_blob_into_buf()) will not.
507 * This also will split up blobs that will need to be extracted to more than
508 * MAX_OPEN_FILES locations, as measured by the 'out_refcnt' of each blob.
509 * Therefore, the apply_operations implementation need not worry about running
510 * out of file descriptors, unless it might open more than one file descriptor
511 * per 'blob_extraction_target' (e.g. Win32 currently might because the
512 * destination file system might not support hard links).
515 extract_blob_list(struct apply_ctx *ctx,
516 const struct read_blob_list_callbacks *cbs)
518 struct read_blob_list_callbacks wrapper_cbs = {
519 .begin_blob = begin_extract_blob_wrapper,
520 .begin_blob_ctx = ctx,
521 .consume_chunk = extract_chunk_wrapper,
522 .consume_chunk_ctx = ctx,
523 .end_blob = end_extract_blob_wrapper,
526 ctx->saved_cbs = cbs;
527 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
528 return read_blobs_from_pipe(ctx, &wrapper_cbs);
530 return read_blob_list(&ctx->blob_list,
531 offsetof(struct blob_descriptor,
533 &wrapper_cbs, VERIFY_BLOB_HASHES);
537 /* Extract a WIM dentry to standard output.
539 * This obviously doesn't make sense in all cases. We return an error if the
540 * dentry does not correspond to a regular file. Otherwise we extract the
541 * unnamed data stream only. */
543 extract_dentry_to_stdout(struct wim_dentry *dentry,
544 const struct blob_table *blob_table)
546 struct wim_inode *inode = dentry->d_inode;
547 struct blob_descriptor *blob;
548 struct filedes _stdout;
550 if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
551 FILE_ATTRIBUTE_DIRECTORY |
552 FILE_ATTRIBUTE_ENCRYPTED))
554 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
555 "extracted to standard output", dentry_full_path(dentry));
556 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
559 blob = inode_get_blob_for_unnamed_data_stream(inode, blob_table);
561 const u8 *hash = inode_get_hash_of_unnamed_data_stream(inode);
562 if (!is_zero_hash(hash))
563 return blob_not_found_error(inode, hash);
567 filedes_init(&_stdout, STDOUT_FILENO);
568 return extract_full_blob_to_fd(blob, &_stdout);
572 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
573 const struct blob_table *blob_table)
575 for (size_t i = 0; i < num_dentries; i++) {
576 int ret = extract_dentry_to_stdout(dentries[i], blob_table);
583 /**********************************************************************/
586 * Removes duplicate dentries from the array.
588 * Returns the new number of dentries, packed at the front of the array.
591 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
594 for (i = 0; i < num_trees; i++) {
595 if (!trees[i]->tmp_flag) {
596 /* Found distinct dentry. */
597 trees[i]->tmp_flag = 1;
598 trees[j++] = trees[i];
601 for (i = 0; i < j; i++)
602 trees[i]->tmp_flag = 0;
607 * Remove dentries that are descendants of other dentries in the array.
609 * Returns the new number of dentries, packed at the front of the array.
612 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
615 for (i = 0; i < num_trees; i++)
616 trees[i]->tmp_flag = 1;
617 for (i = 0; i < num_trees; i++) {
618 struct wim_dentry *d = trees[i];
619 while (!dentry_is_root(d)) {
624 trees[j++] = trees[i];
628 trees[i]->tmp_flag = 0;
631 for (i = 0; i < j; i++)
632 trees[i]->tmp_flag = 0;
637 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
639 struct list_head *dentry_list = _dentry_list;
640 list_add_tail(&dentry->d_extraction_list_node, dentry_list);
645 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
647 dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
651 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
653 list_del(&dentry->d_extraction_list_node);
654 dentry_reset_extraction_list_node(dentry);
659 * Build the preliminary list of dentries to be extracted.
661 * The list maintains the invariant that if d1 and d2 are in the list and d1 is
662 * an ancestor of d2, then d1 appears before d2 in the list.
665 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
666 size_t num_trees, bool add_ancestors)
668 INIT_LIST_HEAD(dentry_list);
670 /* Add the trees recursively. */
671 for (size_t i = 0; i < num_trees; i++)
672 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
674 /* If requested, add ancestors of the trees. */
676 for (size_t i = 0; i < num_trees; i++) {
677 struct wim_dentry *dentry = trees[i];
678 struct wim_dentry *ancestor;
679 struct list_head *place_after;
681 if (dentry_is_root(dentry))
684 place_after = dentry_list;
687 ancestor = ancestor->d_parent;
688 if (will_extract_dentry(ancestor)) {
689 place_after = &ancestor->d_extraction_list_node;
692 } while (!dentry_is_root(ancestor));
696 ancestor = ancestor->d_parent;
697 if (will_extract_dentry(ancestor))
699 list_add(&ancestor->d_extraction_list_node, place_after);
700 } while (!dentry_is_root(ancestor));
706 destroy_dentry_list(struct list_head *dentry_list)
708 struct wim_dentry *dentry, *tmp;
709 struct wim_inode *inode;
711 list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
712 inode = dentry->d_inode;
713 dentry_reset_extraction_list_node(dentry);
714 inode->i_visited = 0;
715 inode->i_can_externally_back = 0;
716 if ((void *)dentry->d_extraction_name != (void *)dentry->file_name)
717 FREE(dentry->d_extraction_name);
718 dentry->d_extraction_name = NULL;
719 dentry->d_extraction_name_nchars = 0;
724 destroy_blob_list(struct list_head *blob_list)
726 struct blob_descriptor *blob;
728 list_for_each_entry(blob, blob_list, extraction_list)
729 if (blob->out_refcnt > ARRAY_LEN(blob->inline_blob_extraction_targets))
730 FREE(blob->blob_extraction_targets);
734 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
736 static const utf16lechar replacement_char = cpu_to_le16('?');
740 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
746 for (i = 0; i < num_chars; i++) {
749 case cpu_to_le16('\\'):
750 case cpu_to_le16(':'):
751 case cpu_to_le16('*'):
752 case cpu_to_le16('?'):
753 case cpu_to_le16('"'):
754 case cpu_to_le16('<'):
755 case cpu_to_le16('>'):
756 case cpu_to_le16('|'):
758 case cpu_to_le16('/'):
759 case cpu_to_le16('\0'):
761 name[i] = replacement_char;
768 if (name[num_chars - 1] == cpu_to_le16(' ') ||
769 name[num_chars - 1] == cpu_to_le16('.'))
772 name[num_chars - 1] = replacement_char;
781 dentry_calculate_extraction_name(struct wim_dentry *dentry,
782 struct apply_ctx *ctx)
786 if (dentry_is_root(dentry))
790 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
791 dentry->d_extraction_name = dentry->file_name;
792 dentry->d_extraction_name_nchars = dentry->file_name_nbytes /
798 if (!ctx->supported_features.case_sensitive_filenames) {
799 struct wim_dentry *other;
800 list_for_each_entry(other, &dentry->d_ci_conflict_list,
803 if (will_extract_dentry(other)) {
804 if (ctx->extract_flags &
805 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
806 WARNING("\"%"TS"\" has the same "
807 "case-insensitive name as "
808 "\"%"TS"\"; extracting "
809 "dummy name instead",
810 dentry_full_path(dentry),
811 dentry_full_path(other));
814 WARNING("Not extracting \"%"TS"\": "
815 "has same case-insensitive "
817 dentry_full_path(dentry),
818 dentry_full_path(other));
825 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
826 ret = utf16le_get_tstr(dentry->file_name,
827 dentry->file_name_nbytes,
828 (const tchar **)&dentry->d_extraction_name,
829 &dentry->d_extraction_name_nchars);
830 dentry->d_extraction_name_nchars /= sizeof(tchar);
833 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
835 WARNING("\"%"TS"\" has an invalid filename "
836 "that is not supported on this platform; "
837 "extracting dummy name instead",
838 dentry_full_path(dentry));
841 WARNING("Not extracting \"%"TS"\": has an invalid filename "
842 "that is not supported on this platform",
843 dentry_full_path(dentry));
850 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
852 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
853 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
855 const tchar *tchar_name;
858 ret = utf16le_get_tstr(utf16_name_copy,
859 dentry->file_name_nbytes,
860 &tchar_name, &tchar_nchars);
864 tchar_nchars /= sizeof(tchar);
866 size_t fixed_name_num_chars = tchar_nchars;
867 tchar fixed_name[tchar_nchars + 50];
869 tmemcpy(fixed_name, tchar_name, tchar_nchars);
870 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
871 T(" (invalid filename #%lu)"),
872 ++ctx->invalid_sequence);
874 utf16le_put_tstr(tchar_name);
876 dentry->d_extraction_name = TSTRDUP(fixed_name);
877 if (!dentry->d_extraction_name)
878 return WIMLIB_ERR_NOMEM;
879 dentry->d_extraction_name_nchars = fixed_name_num_chars;
884 for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
889 * Calculate the actual filename component at which each WIM dentry will be
890 * extracted, with special handling for dentries that are unsupported by the
891 * extraction backend or have invalid names.
893 * ctx->supported_features must be filled in.
895 * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
898 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
899 struct apply_ctx *ctx)
901 struct list_head *prev, *cur;
903 /* Can't use list_for_each_entry() because a call to
904 * dentry_calculate_extraction_name() may delete the current dentry and
905 * its children from the list. */
909 struct wim_dentry *dentry;
913 if (cur == dentry_list)
916 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
918 ret = dentry_calculate_extraction_name(dentry, ctx);
922 if (prev->next == cur)
925 ; /* Current dentry and its children (which follow in
926 the list) were deleted. prev stays the same. */
932 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
933 struct blob_table *blob_table)
935 struct wim_inode *inode = dentry->d_inode;
936 struct blob_descriptor *blob;
940 /* Special case: when extracting from a pipe, the WIM blob table is
941 * initially empty, so "resolving" an inode's streams is initially not
942 * possible. However, we still need to keep track of which blobs,
943 * identified by SHA-1 message digests, need to be extracted, so we
944 * "resolve" the inode's streams anyway by allocating a 'struct
945 * blob_descriptor' for each one. */
946 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
948 ret = inode_resolve_streams(inode, blob_table, force);
951 for (unsigned i = 0; i < inode->i_num_streams; i++) {
952 blob = stream_blob_resolved(&inode->i_streams[i]);
954 blob->out_refcnt = 0;
960 * For each dentry to be extracted, resolve all streams in the corresponding
961 * inode and set 'out_refcnt' in all referenced blob_descriptors to 0.
963 * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
966 dentry_list_resolve_streams(struct list_head *dentry_list,
967 struct apply_ctx *ctx)
969 struct wim_dentry *dentry;
972 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
973 ret = dentry_resolve_streams(dentry,
975 ctx->wim->blob_table);
983 ref_stream(struct wim_inode_stream *strm, struct wim_dentry *dentry,
984 struct apply_ctx *ctx)
986 struct wim_inode *inode = dentry->d_inode;
987 struct blob_descriptor *blob = stream_blob_resolved(strm);
988 struct blob_extraction_target *targets;
993 /* Tally the size only for each actual extraction of the stream (not
994 * additional hard links to the inode). */
995 if (inode->i_visited && ctx->supported_features.hard_links)
998 ctx->progress.extract.total_bytes += blob->size;
999 ctx->progress.extract.total_streams++;
1001 if (inode->i_visited)
1004 /* Add each blob to 'ctx->blob_list' only one time, regardless of how
1005 * many extraction targets it will have. */
1006 if (blob->out_refcnt == 0) {
1007 list_add_tail(&blob->extraction_list, &ctx->blob_list);
1008 ctx->num_blobs_remaining++;
1011 /* Set this stream as an extraction target of 'blob'. */
1013 if (blob->out_refcnt < ARRAY_LEN(blob->inline_blob_extraction_targets)) {
1014 targets = blob->inline_blob_extraction_targets;
1016 struct blob_extraction_target *prev_targets;
1017 size_t alloc_blob_extraction_targets;
1019 if (blob->out_refcnt == ARRAY_LEN(blob->inline_blob_extraction_targets)) {
1020 prev_targets = NULL;
1021 alloc_blob_extraction_targets = ARRAY_LEN(blob->inline_blob_extraction_targets);
1023 prev_targets = blob->blob_extraction_targets;
1024 alloc_blob_extraction_targets = blob->alloc_blob_extraction_targets;
1027 if (blob->out_refcnt == alloc_blob_extraction_targets) {
1028 alloc_blob_extraction_targets *= 2;
1029 targets = REALLOC(prev_targets,
1030 alloc_blob_extraction_targets *
1031 sizeof(targets[0]));
1033 return WIMLIB_ERR_NOMEM;
1034 if (!prev_targets) {
1036 blob->inline_blob_extraction_targets,
1037 sizeof(blob->inline_blob_extraction_targets));
1039 blob->blob_extraction_targets = targets;
1040 blob->alloc_blob_extraction_targets = alloc_blob_extraction_targets;
1042 targets = blob->blob_extraction_targets;
1044 targets[blob->out_refcnt].inode = inode;
1045 targets[blob->out_refcnt].stream = strm;
1051 ref_stream_if_needed(struct wim_dentry *dentry, struct wim_inode *inode,
1052 struct wim_inode_stream *strm, struct apply_ctx *ctx)
1054 bool need_stream = false;
1055 switch (strm->stream_type) {
1056 case STREAM_TYPE_DATA:
1057 if (stream_is_named(strm)) {
1058 /* Named data stream */
1059 if (ctx->supported_features.named_data_streams)
1061 } else if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
1062 FILE_ATTRIBUTE_ENCRYPTED))
1063 && !(inode_is_symlink(inode)
1064 && !ctx->supported_features.reparse_points
1065 && ctx->supported_features.symlink_reparse_points))
1068 * Unnamed data stream. Skip if any of the following is true:
1070 * - file is a directory
1071 * - file is encrypted
1072 * - backend needs to create the file as UNIX symlink
1073 * - backend will extract the stream as externally backed
1075 if (ctx->apply_ops->will_externally_back) {
1076 int ret = (*ctx->apply_ops->will_externally_back)(dentry, ctx);
1077 if (ret > 0) /* Error? */
1079 if (ret < 0) /* Won't externally back? */
1086 case STREAM_TYPE_REPARSE_POINT:
1087 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT);
1088 if (ctx->supported_features.reparse_points ||
1089 (inode_is_symlink(inode) &&
1090 ctx->supported_features.symlink_reparse_points))
1093 case STREAM_TYPE_EFSRPC_RAW_DATA:
1094 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED);
1095 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
1096 if (ctx->supported_features.encrypted_directories)
1099 if (ctx->supported_features.encrypted_files)
1105 return ref_stream(strm, dentry, ctx);
1110 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1112 struct wim_inode *inode = dentry->d_inode;
1113 for (unsigned i = 0; i < inode->i_num_streams; i++) {
1114 int ret = ref_stream_if_needed(dentry, inode,
1115 &inode->i_streams[i], ctx);
1119 inode->i_visited = 1;
1124 * Given a list of dentries to be extracted, build the list of blobs that need
1125 * to be extracted, and for each blob determine the streams to which that blob
1126 * will be extracted.
1128 * This also initializes the extract progress info with byte and blob
1131 * ctx->supported_features must be filled in.
1134 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1136 struct wim_dentry *dentry;
1139 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1140 ret = dentry_ref_streams(dentry, ctx);
1144 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1145 dentry->d_inode->i_visited = 0;
1150 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1152 struct wim_dentry *dentry;
1153 struct wim_inode *inode;
1155 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1156 inode = dentry->d_inode;
1157 if (!inode->i_visited)
1158 INIT_LIST_HEAD(&inode->i_extraction_aliases);
1159 list_add_tail(&dentry->d_extraction_alias_node,
1160 &inode->i_extraction_aliases);
1161 inode->i_visited = 1;
1163 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1164 dentry->d_inode->i_visited = 0;
1168 inode_tally_features(const struct wim_inode *inode,
1169 struct wim_features *features)
1171 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1172 features->archive_files++;
1173 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1174 features->hidden_files++;
1175 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1176 features->system_files++;
1177 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1178 features->compressed_files++;
1179 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1180 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1181 features->encrypted_directories++;
1183 features->encrypted_files++;
1185 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1186 features->not_context_indexed_files++;
1187 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1188 features->sparse_files++;
1189 if (inode_has_named_data_stream(inode))
1190 features->named_data_streams++;
1191 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1192 features->reparse_points++;
1193 if (inode_is_symlink(inode))
1194 features->symlink_reparse_points++;
1196 features->other_reparse_points++;
1198 if (inode->i_security_id != -1)
1199 features->security_descriptors++;
1200 if (inode_has_unix_data(inode))
1201 features->unix_data++;
1204 /* Tally features necessary to extract a dentry and the corresponding inode. */
1206 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1208 struct wim_inode *inode = dentry->d_inode;
1210 if (dentry_has_short_name(dentry))
1211 features->short_names++;
1213 if (inode->i_visited) {
1214 features->hard_links++;
1216 inode_tally_features(inode, features);
1217 inode->i_visited = 1;
1221 /* Tally the features necessary to extract the specified dentries. */
1223 dentry_list_get_features(struct list_head *dentry_list,
1224 struct wim_features *features)
1226 struct wim_dentry *dentry;
1228 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1229 dentry_tally_features(dentry, features);
1231 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1232 dentry->d_inode->i_visited = 0;
1236 do_feature_check(const struct wim_features *required_features,
1237 const struct wim_features *supported_features,
1240 /* Encrypted files. */
1241 if (required_features->encrypted_files &&
1242 !supported_features->encrypted_files)
1243 WARNING("Ignoring EFS-encrypted data of %lu files",
1244 required_features->encrypted_files);
1246 /* Named data streams. */
1247 if (required_features->named_data_streams &&
1248 !supported_features->named_data_streams)
1249 WARNING("Ignoring named data streams of %lu files",
1250 required_features->named_data_streams);
1252 /* File attributes. */
1253 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1254 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1255 * We're an archive program, so theoretically we can do what we
1258 if (required_features->hidden_files &&
1259 !supported_features->hidden_files)
1260 WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1261 required_features->hidden_files);
1263 if (required_features->system_files &&
1264 !supported_features->system_files)
1265 WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1266 required_features->system_files);
1268 if (required_features->compressed_files &&
1269 !supported_features->compressed_files)
1270 WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1271 required_features->compressed_files);
1273 if (required_features->not_context_indexed_files &&
1274 !supported_features->not_context_indexed_files)
1275 WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1276 required_features->not_context_indexed_files);
1278 if (required_features->sparse_files &&
1279 !supported_features->sparse_files)
1280 WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1281 required_features->sparse_files);
1283 if (required_features->encrypted_directories &&
1284 !supported_features->encrypted_directories)
1285 WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1286 required_features->encrypted_directories);
1290 if (required_features->hard_links && !supported_features->hard_links)
1291 WARNING("Extracting %lu hard links as independent files",
1292 required_features->hard_links);
1294 /* Symbolic links and reparse points. */
1295 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1296 required_features->symlink_reparse_points &&
1297 !supported_features->symlink_reparse_points &&
1298 !supported_features->reparse_points)
1300 ERROR("Extraction backend does not support symbolic links!");
1301 return WIMLIB_ERR_UNSUPPORTED;
1303 if (required_features->reparse_points &&
1304 !supported_features->reparse_points)
1306 if (supported_features->symlink_reparse_points) {
1307 if (required_features->other_reparse_points) {
1308 WARNING("Ignoring reparse data of %lu non-symlink/junction files",
1309 required_features->other_reparse_points);
1312 WARNING("Ignoring reparse data of %lu files",
1313 required_features->reparse_points);
1317 /* Security descriptors. */
1318 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1319 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1320 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1321 required_features->security_descriptors &&
1322 !supported_features->security_descriptors)
1324 ERROR("Extraction backend does not support security descriptors!");
1325 return WIMLIB_ERR_UNSUPPORTED;
1327 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1328 required_features->security_descriptors &&
1329 !supported_features->security_descriptors)
1330 WARNING("Ignoring Windows NT security descriptors of %lu files",
1331 required_features->security_descriptors);
1334 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1335 required_features->unix_data && !supported_features->unix_data)
1337 ERROR("Extraction backend does not support UNIX data!");
1338 return WIMLIB_ERR_UNSUPPORTED;
1341 if (required_features->unix_data &&
1342 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1344 WARNING("Ignoring UNIX metadata of %lu files",
1345 required_features->unix_data);
1349 if (required_features->short_names &&
1350 !supported_features->short_names)
1352 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1353 ERROR("Extraction backend does not support DOS names!");
1354 return WIMLIB_ERR_UNSUPPORTED;
1356 WARNING("Ignoring DOS names of %lu files",
1357 required_features->short_names);
1361 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1362 !supported_features->timestamps)
1364 ERROR("Extraction backend does not support timestamps!");
1365 return WIMLIB_ERR_UNSUPPORTED;
1371 static const struct apply_operations *
1372 select_apply_operations(int extract_flags)
1375 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1376 return &ntfs_3g_apply_ops;
1379 return &win32_apply_ops;
1381 return &unix_apply_ops;
1386 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1387 const tchar *target, int extract_flags)
1389 const struct apply_operations *ops;
1390 struct apply_ctx *ctx;
1392 LIST_HEAD(dentry_list);
1394 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1395 ret = extract_dentries_to_stdout(trees, num_trees,
1400 num_trees = remove_duplicate_trees(trees, num_trees);
1401 num_trees = remove_contained_trees(trees, num_trees);
1403 ops = select_apply_operations(extract_flags);
1405 if (num_trees > 1 && ops->single_tree_only) {
1406 ERROR("Extracting multiple directory trees "
1407 "at once is not supported in %s extraction mode!",
1409 ret = WIMLIB_ERR_UNSUPPORTED;
1413 ctx = CALLOC(1, ops->context_size);
1415 ret = WIMLIB_ERR_NOMEM;
1420 ctx->target = target;
1421 ctx->target_nchars = tstrlen(target);
1422 ctx->extract_flags = extract_flags;
1423 if (ctx->wim->progfunc) {
1424 ctx->progfunc = ctx->wim->progfunc;
1425 ctx->progctx = ctx->wim->progctx;
1426 ctx->progress.extract.image = wim->current_image;
1427 ctx->progress.extract.extract_flags = (extract_flags &
1428 WIMLIB_EXTRACT_MASK_PUBLIC);
1429 ctx->progress.extract.wimfile_name = wim->filename;
1430 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1431 wim->current_image);
1432 ctx->progress.extract.target = target;
1434 INIT_LIST_HEAD(&ctx->blob_list);
1435 filedes_invalidate(&ctx->tmpfile_fd);
1436 ctx->apply_ops = ops;
1438 ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1442 build_dentry_list(&dentry_list, trees, num_trees,
1444 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1446 dentry_list_get_features(&dentry_list, &ctx->required_features);
1448 ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1449 ctx->extract_flags);
1453 ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1457 if (unlikely(list_empty(&dentry_list))) {
1458 WARNING("There is nothing to extract!");
1462 ret = dentry_list_resolve_streams(&dentry_list, ctx);
1466 dentry_list_build_inode_alias_lists(&dentry_list);
1468 ret = dentry_list_ref_streams(&dentry_list, ctx);
1472 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1473 /* When extracting from a pipe, the number of bytes of data to
1474 * extract can't be determined in the normal way (examining the
1475 * blob table), since at this point all we have is a set of
1476 * SHA-1 message digests of blobs that need to be extracted.
1477 * However, we can get a reasonably accurate estimate by taking
1478 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1479 * data. This does assume that a full image is being extracted,
1480 * but currently there is no API for doing otherwise. (Also,
1481 * subtract <HARDLINKBYTES> from this if hard links are
1482 * supported by the extraction mode.) */
1483 ctx->progress.extract.total_bytes =
1484 wim_info_get_image_total_bytes(wim->wim_info,
1485 wim->current_image);
1486 if (ctx->supported_features.hard_links) {
1487 ctx->progress.extract.total_bytes -=
1488 wim_info_get_image_hard_link_bytes(wim->wim_info,
1489 wim->current_image);
1493 ret = extract_progress(ctx,
1494 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1495 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1496 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1500 ret = (*ops->extract)(&dentry_list, ctx);
1504 if (ctx->progress.extract.completed_bytes <
1505 ctx->progress.extract.total_bytes)
1507 ctx->progress.extract.completed_bytes =
1508 ctx->progress.extract.total_bytes;
1509 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1514 ret = extract_progress(ctx,
1515 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1516 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1517 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1519 destroy_blob_list(&ctx->blob_list);
1520 destroy_dentry_list(&dentry_list);
1527 mkdir_if_needed(const tchar *target)
1529 if (!tmkdir(target, 0755))
1532 if (errno == EEXIST)
1536 /* _wmkdir() fails with EACCES if called on a drive root directory. */
1537 if (errno == EACCES)
1541 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
1542 return WIMLIB_ERR_MKDIR;
1545 /* Make sure the extraction flags make sense, and update them if needed. */
1547 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1549 int extract_flags = *extract_flags_p;
1551 /* Check for invalid flag combinations */
1553 if ((extract_flags &
1554 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1555 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1556 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1557 return WIMLIB_ERR_INVALID_PARAM;
1559 if ((extract_flags &
1560 (WIMLIB_EXTRACT_FLAG_RPFIX |
1561 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1562 WIMLIB_EXTRACT_FLAG_NORPFIX))
1563 return WIMLIB_ERR_INVALID_PARAM;
1565 #ifndef WITH_NTFS_3G
1566 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1567 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
1568 " it cannot apply a WIM image directly to an NTFS volume.");
1569 return WIMLIB_ERR_UNSUPPORTED;
1573 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1576 return WIMLIB_ERR_NO_FILENAME;
1578 ERROR("WIMBoot extraction is only supported on Windows!");
1579 return WIMLIB_ERR_UNSUPPORTED;
1584 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1585 WIMLIB_EXTRACT_FLAG_NORPFIX |
1586 WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1587 WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1589 /* For full-image extraction, do reparse point fixups by default
1590 * if the WIM header says they are enabled. */
1591 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1592 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1595 *extract_flags_p = extract_flags;
1600 get_wildcard_flags(int extract_flags)
1602 u32 wildcard_flags = 0;
1604 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB)
1605 wildcard_flags |= WILDCARD_FLAG_ERROR_IF_NO_MATCH;
1607 wildcard_flags |= WILDCARD_FLAG_WARN_IF_NO_MATCH;
1609 if (default_ignore_case)
1610 wildcard_flags |= WILDCARD_FLAG_CASE_INSENSITIVE;
1612 return wildcard_flags;
1615 struct append_dentry_ctx {
1616 struct wim_dentry **dentries;
1617 size_t num_dentries;
1618 size_t num_alloc_dentries;
1622 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1624 struct append_dentry_ctx *ctx = _ctx;
1626 if (ctx->num_dentries == ctx->num_alloc_dentries) {
1627 struct wim_dentry **new_dentries;
1630 new_length = max(ctx->num_alloc_dentries + 8,
1631 ctx->num_alloc_dentries * 3 / 2);
1632 new_dentries = REALLOC(ctx->dentries,
1633 new_length * sizeof(ctx->dentries[0]));
1634 if (new_dentries == NULL)
1635 return WIMLIB_ERR_NOMEM;
1636 ctx->dentries = new_dentries;
1637 ctx->num_alloc_dentries = new_length;
1639 ctx->dentries[ctx->num_dentries++] = dentry;
1644 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1645 const tchar * const *paths, size_t num_paths,
1649 struct wim_dentry **trees;
1652 if (wim == NULL || target == NULL || target[0] == T('\0') ||
1653 (num_paths != 0 && paths == NULL))
1654 return WIMLIB_ERR_INVALID_PARAM;
1656 ret = check_extract_flags(wim, &extract_flags);
1660 ret = select_wim_image(wim, image);
1664 ret = wim_checksum_unhashed_blobs(wim);
1668 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1669 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1670 (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1672 ret = mkdir_if_needed(target);
1677 if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1679 struct append_dentry_ctx append_dentry_ctx = {
1682 .num_alloc_dentries = 0,
1685 u32 wildcard_flags = get_wildcard_flags(extract_flags);
1687 for (size_t i = 0; i < num_paths; i++) {
1688 tchar *path = canonicalize_wim_path(paths[i]);
1690 ret = WIMLIB_ERR_NOMEM;
1691 trees = append_dentry_ctx.dentries;
1692 goto out_free_trees;
1694 ret = expand_wildcard(wim, path,
1700 trees = append_dentry_ctx.dentries;
1701 goto out_free_trees;
1704 trees = append_dentry_ctx.dentries;
1705 num_trees = append_dentry_ctx.num_dentries;
1707 trees = MALLOC(num_paths * sizeof(trees[0]));
1709 return WIMLIB_ERR_NOMEM;
1711 for (size_t i = 0; i < num_paths; i++) {
1713 tchar *path = canonicalize_wim_path(paths[i]);
1715 ret = WIMLIB_ERR_NOMEM;
1716 goto out_free_trees;
1719 trees[i] = get_dentry(wim, path,
1720 WIMLIB_CASE_PLATFORM_DEFAULT);
1722 if (trees[i] == NULL) {
1723 ERROR("Path \"%"TS"\" does not exist "
1725 paths[i], wim->current_image);
1726 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1727 goto out_free_trees;
1730 num_trees = num_paths;
1733 if (num_trees == 0) {
1735 goto out_free_trees;
1738 ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1745 extract_single_image(WIMStruct *wim, int image,
1746 const tchar *target, int extract_flags)
1748 const tchar *path = WIMLIB_WIM_ROOT_PATH;
1749 extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1750 return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1753 static const tchar * const filename_forbidden_chars =
1762 /* This function checks if it is okay to use a WIM image's name as a directory
1765 image_name_ok_as_dir(const tchar *image_name)
1767 return image_name && *image_name &&
1768 !tstrpbrk(image_name, filename_forbidden_chars) &&
1769 tstrcmp(image_name, T(".")) &&
1770 tstrcmp(image_name, T(".."));
1773 /* Extracts all images from the WIM to the directory @target, with the images
1774 * placed in subdirectories named by their image names. */
1776 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1778 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
1779 size_t output_path_len = tstrlen(target);
1780 tchar buf[output_path_len + 1 + image_name_max_len + 1];
1783 const tchar *image_name;
1785 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1786 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1787 return WIMLIB_ERR_INVALID_PARAM;
1790 ret = mkdir_if_needed(target);
1793 tmemcpy(buf, target, output_path_len);
1794 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1795 for (image = 1; image <= wim->hdr.image_count; image++) {
1796 image_name = wimlib_get_image_name(wim, image);
1797 if (image_name_ok_as_dir(image_name)) {
1798 tstrcpy(buf + output_path_len + 1, image_name);
1800 /* Image name is empty or contains forbidden characters.
1801 * Use image number instead. */
1802 tsprintf(buf + output_path_len + 1, T("%d"), image);
1804 ret = extract_single_image(wim, image, buf, extract_flags);
1812 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1815 if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1816 WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1817 WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1818 return WIMLIB_ERR_INVALID_PARAM;
1820 if (image == WIMLIB_ALL_IMAGES)
1821 return extract_all_images(wim, target, extract_flags);
1823 return extract_single_image(wim, image, target, extract_flags);
1827 /****************************************************************************
1829 ****************************************************************************/
1832 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1833 const tchar * const *paths, size_t num_paths,
1836 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1837 return WIMLIB_ERR_INVALID_PARAM;
1839 return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1844 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1845 const tchar *path_list_file, int extract_flags)
1852 ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1854 ERROR("Failed to read path list file \"%"TS"\"",
1859 ret = wimlib_extract_paths(wim, image, target,
1860 (const tchar * const *)paths, num_paths,
1868 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1869 const tchar *image_num_or_name,
1870 const tchar *target,
1872 wimlib_progress_func_t progfunc,
1877 struct filedes *in_fd;
1881 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1882 return WIMLIB_ERR_INVALID_PARAM;
1884 /* Read the WIM header from the pipe and get a WIMStruct to represent
1885 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
1886 * wimlib_open_wim(), getting a WIMStruct in this way will result in an
1887 * empty blob table, no XML data read, and no filename set. */
1888 ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1893 /* Sanity check to make sure this is a pipable WIM. */
1894 if (pwm->hdr.magic != PWM_MAGIC) {
1895 ERROR("The WIM being read from file descriptor %d "
1896 "is not pipable!", pipe_fd);
1897 ret = WIMLIB_ERR_NOT_PIPABLE;
1898 goto out_wimlib_free;
1901 /* Sanity check to make sure the first part of a pipable split WIM is
1902 * sent over the pipe first. */
1903 if (pwm->hdr.part_number != 1) {
1904 ERROR("The first part of the split WIM must be "
1905 "sent over the pipe first.");
1906 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1907 goto out_wimlib_free;
1910 in_fd = &pwm->in_fd;
1911 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1913 /* As mentioned, the WIMStruct we created from the pipe does not have
1914 * XML data yet. Fix this by reading the extra copy of the XML data
1915 * that directly follows the header in pipable WIMs. (Note: see
1916 * write_pipable_wim() for more details about the format of pipable
1919 struct blob_descriptor xml_blob;
1920 struct wim_resource_descriptor xml_rdesc;
1921 ret = read_pwm_blob_header(pwm, &xml_blob, &xml_rdesc, 0, NULL);
1923 goto out_wimlib_free;
1925 if (!xml_blob.is_metadata) {
1926 ERROR("Expected XML data, but found non-metadata resource.");
1927 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1928 goto out_wimlib_free;
1931 wim_res_desc_to_hdr(&xml_rdesc, &pwm->hdr.xml_data_reshdr);
1933 ret = read_wim_xml_data(pwm);
1935 goto out_wimlib_free;
1937 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
1938 ERROR("Image count in XML data is not the same as in WIM header.");
1939 ret = WIMLIB_ERR_IMAGE_COUNT;
1940 goto out_wimlib_free;
1944 /* Get image index (this may use the XML data that was just read to
1945 * resolve an image name). */
1946 if (image_num_or_name) {
1947 image = wimlib_resolve_image(pwm, image_num_or_name);
1948 if (image == WIMLIB_NO_IMAGE) {
1949 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
1951 ret = WIMLIB_ERR_INVALID_IMAGE;
1952 goto out_wimlib_free;
1953 } else if (image == WIMLIB_ALL_IMAGES) {
1954 ERROR("Applying all images from a pipe is not supported!");
1955 ret = WIMLIB_ERR_INVALID_IMAGE;
1956 goto out_wimlib_free;
1959 if (pwm->hdr.image_count != 1) {
1960 ERROR("No image was specified, but the pipable WIM "
1961 "did not contain exactly 1 image");
1962 ret = WIMLIB_ERR_INVALID_IMAGE;
1963 goto out_wimlib_free;
1968 /* Load the needed metadata resource. */
1969 for (i = 1; i <= pwm->hdr.image_count; i++) {
1970 struct blob_descriptor *metadata_blob;
1971 struct wim_image_metadata *imd;
1972 struct wim_resource_descriptor *metadata_rdesc;
1974 metadata_blob = new_blob_descriptor();
1975 if (metadata_blob == NULL) {
1976 ret = WIMLIB_ERR_NOMEM;
1977 goto out_wimlib_free;
1979 metadata_rdesc = MALLOC(sizeof(struct wim_resource_descriptor));
1980 if (metadata_rdesc == NULL) {
1981 ret = WIMLIB_ERR_NOMEM;
1982 free_blob_descriptor(metadata_blob);
1983 goto out_wimlib_free;
1986 ret = read_pwm_blob_header(pwm, metadata_blob, metadata_rdesc, 0, NULL);
1987 imd = pwm->image_metadata[i - 1];
1988 imd->metadata_blob = metadata_blob;
1990 FREE(metadata_rdesc);
1991 goto out_wimlib_free;
1994 if (!metadata_blob->is_metadata) {
1995 ERROR("Expected metadata resource, but found "
1996 "non-metadata resource.");
1997 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1998 goto out_wimlib_free;
2002 /* Metadata resource is for the image being extracted.
2003 * Parse it and save the metadata in memory. */
2004 ret = read_metadata_resource(imd);
2006 goto out_wimlib_free;
2009 /* Metadata resource is not for the image being
2010 * extracted. Skip over it. */
2011 ret = skip_wim_resource(metadata_rdesc);
2013 goto out_wimlib_free;
2016 /* Extract the image. */
2017 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2018 ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
2019 /* Clean up and return. */
2027 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2028 const tchar *target, int extract_flags)
2030 return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
2039 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
2042 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
2043 return WIMLIB_ERR_INVALID_PARAM;
2044 return do_wimlib_extract_image(wim, image, target, extract_flags);