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"
62 # include "wimlib/win32.h" /* for realpath() equivalent */
64 #include "wimlib/xml.h"
65 #include "wimlib/wildcard.h"
66 #include "wimlib/wim.h"
68 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x80000000
69 #define WIMLIB_EXTRACT_FLAG_IMAGEMODE 0x40000000
71 /* Keep in sync with wimlib.h */
72 #define WIMLIB_EXTRACT_MASK_PUBLIC \
73 (WIMLIB_EXTRACT_FLAG_NTFS | \
74 WIMLIB_EXTRACT_FLAG_UNIX_DATA | \
75 WIMLIB_EXTRACT_FLAG_NO_ACLS | \
76 WIMLIB_EXTRACT_FLAG_STRICT_ACLS | \
77 WIMLIB_EXTRACT_FLAG_RPFIX | \
78 WIMLIB_EXTRACT_FLAG_NORPFIX | \
79 WIMLIB_EXTRACT_FLAG_TO_STDOUT | \
80 WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES | \
81 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS | \
82 WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS | \
83 WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES | \
84 WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS | \
85 WIMLIB_EXTRACT_FLAG_GLOB_PATHS | \
86 WIMLIB_EXTRACT_FLAG_STRICT_GLOB | \
87 WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES | \
88 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE | \
89 WIMLIB_EXTRACT_FLAG_WIMBOOT)
91 /* Send WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE or
92 * WIMLIB_PROGRESS_MSG_EXTRACT_METADATA. */
94 do_file_extract_progress(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
96 ctx->count_until_file_progress = 500; /* Arbitrary value to limit calls */
97 return extract_progress(ctx, msg);
101 start_file_phase(struct apply_ctx *ctx, uint64_t end_file_count, enum wimlib_progress_msg msg)
103 ctx->progress.extract.current_file_count = 0;
104 ctx->progress.extract.end_file_count = end_file_count;
105 return do_file_extract_progress(ctx, msg);
109 start_file_structure_phase(struct apply_ctx *ctx, uint64_t end_file_count)
111 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
115 start_file_metadata_phase(struct apply_ctx *ctx, uint64_t end_file_count)
117 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
121 end_file_phase(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
123 ctx->progress.extract.current_file_count = ctx->progress.extract.end_file_count;
124 return do_file_extract_progress(ctx, msg);
128 end_file_structure_phase(struct apply_ctx *ctx)
130 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
134 end_file_metadata_phase(struct apply_ctx *ctx)
136 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
139 #define PWM_ALLOW_WIM_HDR 0x00001
141 /* Read the header for a blob in a pipable WIM. */
143 read_pwm_blob_header(WIMStruct *pwm, struct blob_descriptor *blob,
144 struct wim_resource_descriptor *rdesc,
145 int flags, struct wim_header_disk *hdr_ret)
148 struct pwm_blob_hdr blob_hdr;
149 struct wim_header_disk pwm_hdr;
151 struct wim_reshdr reshdr;
154 ret = full_read(&pwm->in_fd, &buf.blob_hdr, sizeof(buf.blob_hdr));
158 if ((flags & PWM_ALLOW_WIM_HDR) &&
159 le64_to_cpu(buf.blob_hdr.magic) == PWM_MAGIC)
161 BUILD_BUG_ON(sizeof(buf.pwm_hdr) < sizeof(buf.blob_hdr));
162 ret = full_read(&pwm->in_fd, &buf.blob_hdr + 1,
163 sizeof(buf.pwm_hdr) - sizeof(buf.blob_hdr));
167 blob->blob_location = BLOB_NONEXISTENT;
168 memcpy(hdr_ret, &buf.pwm_hdr, sizeof(buf.pwm_hdr));
172 if (le64_to_cpu(buf.blob_hdr.magic) != PWM_BLOB_MAGIC) {
173 ERROR("Data read on pipe is invalid (expected blob header).");
174 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
177 copy_hash(blob->hash, buf.blob_hdr.hash);
179 reshdr.size_in_wim = 0;
180 reshdr.flags = le32_to_cpu(buf.blob_hdr.flags);
181 reshdr.offset_in_wim = pwm->in_fd.offset;
182 reshdr.uncompressed_size = le64_to_cpu(buf.blob_hdr.uncompressed_size);
183 wim_res_hdr_to_desc(&reshdr, pwm, rdesc);
184 blob_set_is_located_in_wim_resource(blob, rdesc);
185 blob->size = rdesc->uncompressed_size;
186 blob->offset_in_res = 0;
187 blob->is_metadata = (rdesc->flags & WIM_RESHDR_FLAG_METADATA) != 0;
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 needed_blob->offset_in_res = found_blob->offset_in_res;
238 needed_blob->size = found_blob->size;
240 blob_unset_is_located_in_wim_resource(found_blob);
241 blob_set_is_located_in_wim_resource(needed_blob, rdesc);
243 ret = (*cbs->begin_blob)(needed_blob,
244 cbs->begin_blob_ctx);
246 blob_unset_is_located_in_wim_resource(needed_blob);
250 ret = extract_blob(needed_blob, needed_blob->size,
252 cbs->consume_chunk_ctx);
254 ret = (*cbs->end_blob)(needed_blob, ret,
256 blob_unset_is_located_in_wim_resource(needed_blob);
259 ctx->num_blobs_remaining--;
260 } else if (found_blob->blob_location != BLOB_NONEXISTENT) {
261 ret = skip_wim_resource(found_blob->rdesc);
265 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
266 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
268 if (part_number != ctx->progress.extract.part_number ||
269 total_parts != ctx->progress.extract.total_parts ||
270 memcmp(pwm_hdr.guid, ctx->progress.extract.guid,
273 ctx->progress.extract.part_number = part_number;
274 ctx->progress.extract.total_parts = total_parts;
275 memcpy(ctx->progress.extract.guid,
276 pwm_hdr.guid, WIM_GUID_LEN);
277 ret = extract_progress(ctx,
278 WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
286 if (found_blob && found_blob->blob_location != BLOB_IN_WIM)
288 free_blob_descriptor(found_blob);
292 /* Creates a temporary file opened for writing. The open file descriptor is
293 * returned in @fd_ret and its name is returned in @name_ret (dynamically
296 create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
303 name = ttempnam(NULL, T("wimlib"));
305 ERROR_WITH_ERRNO("Failed to create temporary filename");
306 return WIMLIB_ERR_NOMEM;
309 open_flags = O_WRONLY | O_CREAT | O_EXCL | O_BINARY;
311 open_flags |= _O_SHORT_LIVED;
313 raw_fd = topen(name, open_flags, 0600);
316 if (errno == EEXIST) {
320 ERROR_WITH_ERRNO("Failed to create temporary file "
323 return WIMLIB_ERR_OPEN;
326 filedes_init(fd_ret, raw_fd);
332 begin_extract_blob_wrapper(struct blob_descriptor *blob, void *_ctx)
334 struct apply_ctx *ctx = _ctx;
336 ctx->cur_blob = blob;
337 ctx->cur_blob_offset = 0;
339 if (unlikely(blob->out_refcnt > MAX_OPEN_FILES))
340 return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
342 return (*ctx->saved_cbs->begin_blob)(blob, ctx->saved_cbs->begin_blob_ctx);
346 extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx)
348 struct apply_ctx *ctx = _ctx;
349 union wimlib_progress_info *progress = &ctx->progress;
352 ctx->cur_blob_offset += size;
354 if (likely(ctx->supported_features.hard_links)) {
355 progress->extract.completed_bytes +=
356 (u64)size * ctx->cur_blob->out_refcnt;
357 if (ctx->cur_blob_offset == ctx->cur_blob->size)
358 progress->extract.completed_streams += ctx->cur_blob->out_refcnt;
360 const struct blob_extraction_target *targets =
361 blob_extraction_targets(ctx->cur_blob);
362 for (u32 i = 0; i < ctx->cur_blob->out_refcnt; i++) {
363 const struct wim_inode *inode = targets[i].inode;
364 const struct wim_dentry *dentry;
366 list_for_each_entry(dentry,
367 &inode->i_extraction_aliases,
368 d_extraction_alias_node)
370 progress->extract.completed_bytes += size;
371 if (ctx->cur_blob_offset == ctx->cur_blob->size)
372 progress->extract.completed_streams++;
376 if (progress->extract.completed_bytes >= ctx->next_progress) {
378 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
382 if (progress->extract.completed_bytes >=
383 progress->extract.total_bytes)
385 ctx->next_progress = UINT64_MAX;
387 /* Send new message as soon as another 1/128 of the
388 * total has been extracted. (Arbitrary number.) */
390 progress->extract.completed_bytes +
391 progress->extract.total_bytes / 128;
393 /* ... Unless that would be more than 5000000 bytes, in
394 * which case send the next after the next 5000000
395 * bytes. (Another arbitrary number.) */
396 if (progress->extract.completed_bytes + 5000000 <
399 progress->extract.completed_bytes + 5000000;
401 /* ... But always send a message as soon as we're
402 * completely done. */
403 if (progress->extract.total_bytes < ctx->next_progress)
404 ctx->next_progress = progress->extract.total_bytes;
408 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
409 /* Just extracting to temporary file for now. */
410 ret = full_write(&ctx->tmpfile_fd, chunk, size);
412 ERROR_WITH_ERRNO("Error writing data to "
413 "temporary file \"%"TS"\"",
418 return (*ctx->saved_cbs->consume_chunk)(chunk, size,
419 ctx->saved_cbs->consume_chunk_ctx);
424 extract_from_tmpfile(const tchar *tmpfile_name, struct apply_ctx *ctx)
426 struct blob_descriptor tmpfile_blob;
427 struct blob_descriptor *orig_blob = ctx->cur_blob;
428 const struct read_blob_list_callbacks *cbs = ctx->saved_cbs;
430 const u32 orig_refcnt = orig_blob->out_refcnt;
432 BUILD_BUG_ON(MAX_OPEN_FILES <
433 ARRAY_LEN(orig_blob->inline_blob_extraction_targets));
435 struct blob_extraction_target *targets = orig_blob->blob_extraction_targets;
437 /* Copy the blob's data from the temporary file to each of its targets.
439 * This is executed only in the very uncommon case that a blob is being
440 * extracted to more than MAX_OPEN_FILES targets! */
442 memcpy(&tmpfile_blob, orig_blob, sizeof(struct blob_descriptor));
443 tmpfile_blob.blob_location = BLOB_IN_FILE_ON_DISK;
444 tmpfile_blob.file_on_disk = ctx->tmpfile_name;
446 for (u32 i = 0; i < orig_refcnt; i++) {
448 /* Note: it usually doesn't matter whether we pass the original
449 * blob descriptor to callbacks provided by the extraction
450 * backend as opposed to the tmpfile blob descriptor, since they
451 * shouldn't actually read data from the blob other than through
452 * the read_blob_prefix() call below. But for
453 * WIMLIB_EXTRACT_FLAG_WIMBOOT mode on Windows it does matter
454 * because it needs access to the original WIM resource
455 * descriptor in order to create the external backing reference.
458 orig_blob->out_refcnt = 1;
459 orig_blob->inline_blob_extraction_targets[0] = targets[i];
461 ret = (*cbs->begin_blob)(orig_blob, cbs->begin_blob_ctx);
465 /* Extra SHA-1 isn't necessary here, but it shouldn't hurt as
466 * this case is very rare anyway. */
467 ret = extract_blob(&tmpfile_blob, tmpfile_blob.size,
469 cbs->consume_chunk_ctx);
471 ret = (*cbs->end_blob)(orig_blob, ret, cbs->end_blob_ctx);
476 orig_blob->out_refcnt = 0;
481 end_extract_blob_wrapper(struct blob_descriptor *blob, int status, void *_ctx)
483 struct apply_ctx *ctx = _ctx;
485 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
486 filedes_close(&ctx->tmpfile_fd);
488 status = extract_from_tmpfile(ctx->tmpfile_name, ctx);
489 filedes_invalidate(&ctx->tmpfile_fd);
490 tunlink(ctx->tmpfile_name);
491 FREE(ctx->tmpfile_name);
494 return (*ctx->saved_cbs->end_blob)(blob, status,
495 ctx->saved_cbs->end_blob_ctx);
500 * Read the list of blobs to extract and feed their data into the specified
501 * callback functions.
503 * This handles checksumming each blob.
505 * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
507 * This also works if the WIM is being read from a pipe, whereas attempting to
508 * read blobs directly (e.g. with read_full_blob_into_buf()) will not.
510 * This also will split up blobs that will need to be extracted to more than
511 * MAX_OPEN_FILES locations, as measured by the 'out_refcnt' of each blob.
512 * Therefore, the apply_operations implementation need not worry about running
513 * out of file descriptors, unless it might open more than one file descriptor
514 * per nominal destination (e.g. Win32 currently might because the destination
515 * file system might not support hard links).
518 extract_blob_list(struct apply_ctx *ctx,
519 const struct read_blob_list_callbacks *cbs)
521 struct read_blob_list_callbacks wrapper_cbs = {
522 .begin_blob = begin_extract_blob_wrapper,
523 .begin_blob_ctx = ctx,
524 .consume_chunk = extract_chunk_wrapper,
525 .consume_chunk_ctx = ctx,
526 .end_blob = end_extract_blob_wrapper,
529 ctx->saved_cbs = cbs;
530 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
531 return read_blobs_from_pipe(ctx, &wrapper_cbs);
533 return read_blob_list(&ctx->blob_list,
534 offsetof(struct blob_descriptor,
536 &wrapper_cbs, VERIFY_BLOB_HASHES);
540 /* Extract a WIM dentry to standard output.
542 * This obviously doesn't make sense in all cases. We return an error if the
543 * dentry does not correspond to a regular file. Otherwise we extract the
544 * unnamed data stream only. */
546 extract_dentry_to_stdout(struct wim_dentry *dentry,
547 const struct blob_table *blob_table)
549 struct wim_inode *inode = dentry->d_inode;
550 struct blob_descriptor *blob;
551 struct filedes _stdout;
553 if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
554 FILE_ATTRIBUTE_DIRECTORY |
555 FILE_ATTRIBUTE_ENCRYPTED))
557 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
558 "extracted to standard output", dentry_full_path(dentry));
559 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
562 blob = inode_get_blob_for_unnamed_data_stream(inode, blob_table);
564 const u8 *hash = inode_get_hash_of_unnamed_data_stream(inode);
565 if (!is_zero_hash(hash))
566 return blob_not_found_error(inode, hash);
570 filedes_init(&_stdout, STDOUT_FILENO);
571 return extract_full_blob_to_fd(blob, &_stdout);
575 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
576 const struct blob_table *blob_table)
578 for (size_t i = 0; i < num_dentries; i++) {
579 int ret = extract_dentry_to_stdout(dentries[i], blob_table);
586 /**********************************************************************/
589 * Removes duplicate dentries from the array.
591 * Returns the new number of dentries, packed at the front of the array.
594 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
597 for (i = 0; i < num_trees; i++) {
598 if (!trees[i]->tmp_flag) {
599 /* Found distinct dentry. */
600 trees[i]->tmp_flag = 1;
601 trees[j++] = trees[i];
604 for (i = 0; i < j; i++)
605 trees[i]->tmp_flag = 0;
610 * Remove dentries that are descendants of other dentries in the array.
612 * Returns the new number of dentries, packed at the front of the array.
615 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
618 for (i = 0; i < num_trees; i++)
619 trees[i]->tmp_flag = 1;
620 for (i = 0; i < num_trees; i++) {
621 struct wim_dentry *d = trees[i];
622 while (!dentry_is_root(d)) {
627 trees[j++] = trees[i];
631 trees[i]->tmp_flag = 0;
634 for (i = 0; i < j; i++)
635 trees[i]->tmp_flag = 0;
640 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
642 struct list_head *dentry_list = _dentry_list;
643 list_add_tail(&dentry->d_extraction_list_node, dentry_list);
648 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
650 dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
654 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
656 list_del(&dentry->d_extraction_list_node);
657 dentry_reset_extraction_list_node(dentry);
662 * Build the preliminary list of dentries to be extracted.
664 * The list maintains the invariant that if d1 and d2 are in the list and d1 is
665 * an ancestor of d2, then d1 appears before d2 in the list.
668 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
669 size_t num_trees, bool add_ancestors)
671 INIT_LIST_HEAD(dentry_list);
673 /* Add the trees recursively. */
674 for (size_t i = 0; i < num_trees; i++)
675 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
677 /* If requested, add ancestors of the trees. */
679 for (size_t i = 0; i < num_trees; i++) {
680 struct wim_dentry *dentry = trees[i];
681 struct wim_dentry *ancestor;
682 struct list_head *place_after;
684 if (dentry_is_root(dentry))
687 place_after = dentry_list;
690 ancestor = ancestor->d_parent;
691 if (will_extract_dentry(ancestor)) {
692 place_after = &ancestor->d_extraction_list_node;
695 } while (!dentry_is_root(ancestor));
699 ancestor = ancestor->d_parent;
700 if (will_extract_dentry(ancestor))
702 list_add(&ancestor->d_extraction_list_node, place_after);
703 } while (!dentry_is_root(ancestor));
709 destroy_dentry_list(struct list_head *dentry_list)
711 struct wim_dentry *dentry, *tmp;
712 struct wim_inode *inode;
714 list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
715 inode = dentry->d_inode;
716 dentry_reset_extraction_list_node(dentry);
717 inode->i_visited = 0;
718 inode->i_can_externally_back = 0;
719 if ((void *)dentry->d_extraction_name != (void *)dentry->file_name)
720 FREE(dentry->d_extraction_name);
721 dentry->d_extraction_name = NULL;
722 dentry->d_extraction_name_nchars = 0;
727 destroy_blob_list(struct list_head *blob_list)
729 struct blob_descriptor *blob;
731 list_for_each_entry(blob, blob_list, extraction_list)
732 if (blob->out_refcnt > ARRAY_LEN(blob->inline_blob_extraction_targets))
733 FREE(blob->blob_extraction_targets);
737 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
739 static const utf16lechar replacement_char = cpu_to_le16('?');
743 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
749 for (i = 0; i < num_chars; i++) {
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('"'):
757 case cpu_to_le16('<'):
758 case cpu_to_le16('>'):
759 case cpu_to_le16('|'):
761 case cpu_to_le16('/'):
762 case cpu_to_le16('\0'):
764 name[i] = replacement_char;
771 if (name[num_chars - 1] == cpu_to_le16(' ') ||
772 name[num_chars - 1] == cpu_to_le16('.'))
775 name[num_chars - 1] = replacement_char;
784 dentry_calculate_extraction_name(struct wim_dentry *dentry,
785 struct apply_ctx *ctx)
789 if (dentry_is_root(dentry))
793 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
794 dentry->d_extraction_name = dentry->file_name;
795 dentry->d_extraction_name_nchars = dentry->file_name_nbytes /
801 if (!ctx->supported_features.case_sensitive_filenames) {
802 struct wim_dentry *other;
803 list_for_each_entry(other, &dentry->d_ci_conflict_list,
806 if (will_extract_dentry(other)) {
807 if (ctx->extract_flags &
808 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
809 WARNING("\"%"TS"\" has the same "
810 "case-insensitive name as "
811 "\"%"TS"\"; extracting "
812 "dummy name instead",
813 dentry_full_path(dentry),
814 dentry_full_path(other));
817 WARNING("Not extracting \"%"TS"\": "
818 "has same case-insensitive "
820 dentry_full_path(dentry),
821 dentry_full_path(other));
828 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
829 ret = utf16le_get_tstr(dentry->file_name,
830 dentry->file_name_nbytes,
831 (const tchar **)&dentry->d_extraction_name,
832 &dentry->d_extraction_name_nchars);
833 dentry->d_extraction_name_nchars /= sizeof(tchar);
836 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
838 WARNING("\"%"TS"\" has an invalid filename "
839 "that is not supported on this platform; "
840 "extracting dummy name instead",
841 dentry_full_path(dentry));
844 WARNING("Not extracting \"%"TS"\": has an invalid filename "
845 "that is not supported on this platform",
846 dentry_full_path(dentry));
853 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
855 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
856 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
858 const tchar *tchar_name;
861 ret = utf16le_get_tstr(utf16_name_copy,
862 dentry->file_name_nbytes,
863 &tchar_name, &tchar_nchars);
867 tchar_nchars /= sizeof(tchar);
869 size_t fixed_name_num_chars = tchar_nchars;
870 tchar fixed_name[tchar_nchars + 50];
872 tmemcpy(fixed_name, tchar_name, tchar_nchars);
873 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
874 T(" (invalid filename #%lu)"),
875 ++ctx->invalid_sequence);
877 utf16le_put_tstr(tchar_name);
879 dentry->d_extraction_name = TSTRDUP(fixed_name);
880 if (!dentry->d_extraction_name)
881 return WIMLIB_ERR_NOMEM;
882 dentry->d_extraction_name_nchars = fixed_name_num_chars;
887 for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
892 * Calculate the actual filename component at which each WIM dentry will be
893 * extracted, with special handling for dentries that are unsupported by the
894 * extraction backend or have invalid names.
896 * ctx->supported_features must be filled in.
898 * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
901 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
902 struct apply_ctx *ctx)
904 struct list_head *prev, *cur;
906 /* Can't use list_for_each_entry() because a call to
907 * dentry_calculate_extraction_name() may delete the current dentry and
908 * its children from the list. */
912 struct wim_dentry *dentry;
916 if (cur == dentry_list)
919 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
921 ret = dentry_calculate_extraction_name(dentry, ctx);
925 if (prev->next == cur)
928 ; /* Current dentry and its children (which follow in
929 the list) were deleted. prev stays the same. */
935 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
936 struct blob_table *blob_table)
938 struct wim_inode *inode = dentry->d_inode;
939 struct blob_descriptor *blob;
943 /* Special case: when extracting from a pipe, the WIM blob table is
944 * initially empty, so "resolving" an inode's streams is initially not
945 * possible. However, we still need to keep track of which blobs,
946 * identified by SHA-1 message digests, need to be extracted, so we
947 * "resolve" the inode's streams anyway by allocating a 'struct
948 * blob_descriptor' for each one. */
949 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
951 ret = inode_resolve_streams(inode, blob_table, force);
954 for (unsigned i = 0; i < inode->i_num_streams; i++) {
955 blob = stream_blob_resolved(&inode->i_streams[i]);
957 blob->out_refcnt = 0;
963 * For each dentry to be extracted, resolve all streams in the corresponding
964 * inode and set 'out_refcnt' in all referenced blob_descriptors to 0.
966 * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
969 dentry_list_resolve_streams(struct list_head *dentry_list,
970 struct apply_ctx *ctx)
972 struct wim_dentry *dentry;
975 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
976 ret = dentry_resolve_streams(dentry,
978 ctx->wim->blob_table);
986 ref_stream(struct wim_inode_stream *strm, struct wim_dentry *dentry,
987 struct apply_ctx *ctx)
989 struct wim_inode *inode = dentry->d_inode;
990 struct blob_descriptor *blob = stream_blob_resolved(strm);
991 struct blob_extraction_target *targets;
996 /* Tally the size only for each actual extraction of the stream (not
997 * additional hard links to the inode). */
998 if (inode->i_visited && ctx->supported_features.hard_links)
1001 ctx->progress.extract.total_bytes += blob->size;
1002 ctx->progress.extract.total_streams++;
1004 if (inode->i_visited)
1007 /* Add each blob to 'ctx->blob_list' only one time, regardless of how
1008 * many extraction targets it will have. */
1009 if (blob->out_refcnt == 0) {
1010 list_add_tail(&blob->extraction_list, &ctx->blob_list);
1011 ctx->num_blobs_remaining++;
1014 /* Set this stream as an extraction target of 'blob'. */
1016 if (blob->out_refcnt < ARRAY_LEN(blob->inline_blob_extraction_targets)) {
1017 targets = blob->inline_blob_extraction_targets;
1019 struct blob_extraction_target *prev_targets;
1020 size_t alloc_blob_extraction_targets;
1022 if (blob->out_refcnt == ARRAY_LEN(blob->inline_blob_extraction_targets)) {
1023 prev_targets = NULL;
1024 alloc_blob_extraction_targets = ARRAY_LEN(blob->inline_blob_extraction_targets);
1026 prev_targets = blob->blob_extraction_targets;
1027 alloc_blob_extraction_targets = blob->alloc_blob_extraction_targets;
1030 if (blob->out_refcnt == alloc_blob_extraction_targets) {
1031 alloc_blob_extraction_targets *= 2;
1032 targets = REALLOC(prev_targets,
1033 alloc_blob_extraction_targets *
1034 sizeof(targets[0]));
1036 return WIMLIB_ERR_NOMEM;
1037 if (!prev_targets) {
1039 blob->inline_blob_extraction_targets,
1040 sizeof(blob->inline_blob_extraction_targets));
1042 blob->blob_extraction_targets = targets;
1043 blob->alloc_blob_extraction_targets = alloc_blob_extraction_targets;
1045 targets = blob->blob_extraction_targets;
1047 targets[blob->out_refcnt].inode = inode;
1048 targets[blob->out_refcnt].stream = strm;
1054 ref_stream_if_needed(struct wim_dentry *dentry, struct wim_inode *inode,
1055 struct wim_inode_stream *strm, struct apply_ctx *ctx)
1057 bool need_stream = false;
1058 switch (strm->stream_type) {
1059 case STREAM_TYPE_DATA:
1060 if (stream_is_named(strm)) {
1061 /* Named data stream */
1062 if (ctx->supported_features.named_data_streams)
1064 } else if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
1065 FILE_ATTRIBUTE_ENCRYPTED))
1066 && !(inode_is_symlink(inode)
1067 && !ctx->supported_features.reparse_points
1068 && ctx->supported_features.symlink_reparse_points))
1071 * Unnamed data stream. Skip if any of the following is true:
1073 * - file is a directory
1074 * - file is encrypted
1075 * - backend needs to create the file as UNIX symlink
1076 * - backend will extract the stream as externally backed
1078 if (ctx->apply_ops->will_externally_back) {
1079 int ret = (*ctx->apply_ops->will_externally_back)(dentry, ctx);
1080 if (ret > 0) /* Error? */
1082 if (ret < 0) /* Won't externally back? */
1089 case STREAM_TYPE_REPARSE_POINT:
1090 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT);
1091 if (ctx->supported_features.reparse_points ||
1092 (inode_is_symlink(inode) &&
1093 ctx->supported_features.symlink_reparse_points))
1096 case STREAM_TYPE_EFSRPC_RAW_DATA:
1097 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED);
1098 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
1099 if (ctx->supported_features.encrypted_directories)
1102 if (ctx->supported_features.encrypted_files)
1108 return ref_stream(strm, dentry, ctx);
1113 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1115 struct wim_inode *inode = dentry->d_inode;
1116 for (unsigned i = 0; i < inode->i_num_streams; i++) {
1117 int ret = ref_stream_if_needed(dentry, inode,
1118 &inode->i_streams[i], ctx);
1122 inode->i_visited = 1;
1127 * Given a list of dentries to be extracted, build the list of blobs that need
1128 * to be extracted, and for each blob determine the streams to which that blob
1129 * will be extracted.
1131 * This also initializes the extract progress info with byte and blob
1134 * ctx->supported_features must be filled in.
1137 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1139 struct wim_dentry *dentry;
1142 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1143 ret = dentry_ref_streams(dentry, ctx);
1147 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1148 dentry->d_inode->i_visited = 0;
1153 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1155 struct wim_dentry *dentry;
1156 struct wim_inode *inode;
1158 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1159 inode = dentry->d_inode;
1160 if (!inode->i_visited)
1161 INIT_LIST_HEAD(&inode->i_extraction_aliases);
1162 list_add_tail(&dentry->d_extraction_alias_node,
1163 &inode->i_extraction_aliases);
1164 inode->i_visited = 1;
1166 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1167 dentry->d_inode->i_visited = 0;
1171 inode_tally_features(const struct wim_inode *inode,
1172 struct wim_features *features)
1174 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1175 features->archive_files++;
1176 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1177 features->hidden_files++;
1178 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1179 features->system_files++;
1180 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1181 features->compressed_files++;
1182 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1183 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1184 features->encrypted_directories++;
1186 features->encrypted_files++;
1188 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1189 features->not_context_indexed_files++;
1190 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1191 features->sparse_files++;
1192 if (inode_has_named_data_stream(inode))
1193 features->named_data_streams++;
1194 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1195 features->reparse_points++;
1196 if (inode_is_symlink(inode))
1197 features->symlink_reparse_points++;
1199 features->other_reparse_points++;
1201 if (inode->i_security_id != -1)
1202 features->security_descriptors++;
1203 if (inode_has_unix_data(inode))
1204 features->unix_data++;
1207 /* Tally features necessary to extract a dentry and the corresponding inode. */
1209 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1211 struct wim_inode *inode = dentry->d_inode;
1213 if (dentry_has_short_name(dentry))
1214 features->short_names++;
1216 if (inode->i_visited) {
1217 features->hard_links++;
1219 inode_tally_features(inode, features);
1220 inode->i_visited = 1;
1224 /* Tally the features necessary to extract the specified dentries. */
1226 dentry_list_get_features(struct list_head *dentry_list,
1227 struct wim_features *features)
1229 struct wim_dentry *dentry;
1231 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1232 dentry_tally_features(dentry, features);
1234 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1235 dentry->d_inode->i_visited = 0;
1239 do_feature_check(const struct wim_features *required_features,
1240 const struct wim_features *supported_features,
1243 /* Encrypted files. */
1244 if (required_features->encrypted_files &&
1245 !supported_features->encrypted_files)
1246 WARNING("Ignoring EFS-encrypted data of %lu files",
1247 required_features->encrypted_files);
1249 /* Named data streams. */
1250 if (required_features->named_data_streams &&
1251 !supported_features->named_data_streams)
1252 WARNING("Ignoring named data streams of %lu files",
1253 required_features->named_data_streams);
1255 /* File attributes. */
1256 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1257 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1258 * We're an archive program, so theoretically we can do what we
1261 if (required_features->hidden_files &&
1262 !supported_features->hidden_files)
1263 WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1264 required_features->hidden_files);
1266 if (required_features->system_files &&
1267 !supported_features->system_files)
1268 WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1269 required_features->system_files);
1271 if (required_features->compressed_files &&
1272 !supported_features->compressed_files)
1273 WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1274 required_features->compressed_files);
1276 if (required_features->not_context_indexed_files &&
1277 !supported_features->not_context_indexed_files)
1278 WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1279 required_features->not_context_indexed_files);
1281 if (required_features->sparse_files &&
1282 !supported_features->sparse_files)
1283 WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1284 required_features->sparse_files);
1286 if (required_features->encrypted_directories &&
1287 !supported_features->encrypted_directories)
1288 WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1289 required_features->encrypted_directories);
1293 if (required_features->hard_links && !supported_features->hard_links)
1294 WARNING("Extracting %lu hard links as independent files",
1295 required_features->hard_links);
1297 /* Symbolic links and reparse points. */
1298 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1299 required_features->symlink_reparse_points &&
1300 !supported_features->symlink_reparse_points &&
1301 !supported_features->reparse_points)
1303 ERROR("Extraction backend does not support symbolic links!");
1304 return WIMLIB_ERR_UNSUPPORTED;
1306 if (required_features->reparse_points &&
1307 !supported_features->reparse_points)
1309 if (supported_features->symlink_reparse_points) {
1310 if (required_features->other_reparse_points) {
1311 WARNING("Ignoring reparse data of %lu non-symlink/junction files",
1312 required_features->other_reparse_points);
1315 WARNING("Ignoring reparse data of %lu files",
1316 required_features->reparse_points);
1320 /* Security descriptors. */
1321 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1322 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1323 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1324 required_features->security_descriptors &&
1325 !supported_features->security_descriptors)
1327 ERROR("Extraction backend does not support security descriptors!");
1328 return WIMLIB_ERR_UNSUPPORTED;
1330 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1331 required_features->security_descriptors &&
1332 !supported_features->security_descriptors)
1333 WARNING("Ignoring Windows NT security descriptors of %lu files",
1334 required_features->security_descriptors);
1337 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1338 required_features->unix_data && !supported_features->unix_data)
1340 ERROR("Extraction backend does not support UNIX data!");
1341 return WIMLIB_ERR_UNSUPPORTED;
1344 if (required_features->unix_data &&
1345 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1347 WARNING("Ignoring UNIX metadata of %lu files",
1348 required_features->unix_data);
1352 if (required_features->short_names &&
1353 !supported_features->short_names)
1355 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1356 ERROR("Extraction backend does not support DOS names!");
1357 return WIMLIB_ERR_UNSUPPORTED;
1359 WARNING("Ignoring DOS names of %lu files",
1360 required_features->short_names);
1364 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1365 !supported_features->timestamps)
1367 ERROR("Extraction backend does not support timestamps!");
1368 return WIMLIB_ERR_UNSUPPORTED;
1374 static const struct apply_operations *
1375 select_apply_operations(int extract_flags)
1378 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1379 return &ntfs_3g_apply_ops;
1382 return &win32_apply_ops;
1384 return &unix_apply_ops;
1389 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1390 const tchar *target, int extract_flags)
1392 const struct apply_operations *ops;
1393 struct apply_ctx *ctx;
1395 LIST_HEAD(dentry_list);
1397 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1398 ret = extract_dentries_to_stdout(trees, num_trees,
1403 num_trees = remove_duplicate_trees(trees, num_trees);
1404 num_trees = remove_contained_trees(trees, num_trees);
1406 ops = select_apply_operations(extract_flags);
1408 if (num_trees > 1 && ops->single_tree_only) {
1409 ERROR("Extracting multiple directory trees "
1410 "at once is not supported in %s extraction mode!",
1412 ret = WIMLIB_ERR_UNSUPPORTED;
1416 ctx = CALLOC(1, ops->context_size);
1418 ret = WIMLIB_ERR_NOMEM;
1423 ctx->target = target;
1424 ctx->target_nchars = tstrlen(target);
1425 ctx->extract_flags = extract_flags;
1426 if (ctx->wim->progfunc) {
1427 ctx->progfunc = ctx->wim->progfunc;
1428 ctx->progctx = ctx->wim->progctx;
1429 ctx->progress.extract.image = wim->current_image;
1430 ctx->progress.extract.extract_flags = (extract_flags &
1431 WIMLIB_EXTRACT_MASK_PUBLIC);
1432 ctx->progress.extract.wimfile_name = wim->filename;
1433 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1434 wim->current_image);
1435 ctx->progress.extract.target = target;
1437 INIT_LIST_HEAD(&ctx->blob_list);
1438 filedes_invalidate(&ctx->tmpfile_fd);
1439 ctx->apply_ops = ops;
1441 ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1445 build_dentry_list(&dentry_list, trees, num_trees,
1447 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1449 dentry_list_get_features(&dentry_list, &ctx->required_features);
1451 ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1452 ctx->extract_flags);
1456 ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1460 if (unlikely(list_empty(&dentry_list))) {
1461 WARNING("There is nothing to extract!");
1465 ret = dentry_list_resolve_streams(&dentry_list, ctx);
1469 dentry_list_build_inode_alias_lists(&dentry_list);
1471 ret = dentry_list_ref_streams(&dentry_list, ctx);
1475 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1476 /* When extracting from a pipe, the number of bytes of data to
1477 * extract can't be determined in the normal way (examining the
1478 * blob table), since at this point all we have is a set of
1479 * SHA-1 message digests of blobs that need to be extracted.
1480 * However, we can get a reasonably accurate estimate by taking
1481 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1482 * data. This does assume that a full image is being extracted,
1483 * but currently there is no API for doing otherwise. (Also,
1484 * subtract <HARDLINKBYTES> from this if hard links are
1485 * supported by the extraction mode.) */
1486 ctx->progress.extract.total_bytes =
1487 wim_info_get_image_total_bytes(wim->wim_info,
1488 wim->current_image);
1489 if (ctx->supported_features.hard_links) {
1490 ctx->progress.extract.total_bytes -=
1491 wim_info_get_image_hard_link_bytes(wim->wim_info,
1492 wim->current_image);
1496 ret = extract_progress(ctx,
1497 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1498 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1499 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1503 ret = (*ops->extract)(&dentry_list, ctx);
1507 if (ctx->progress.extract.completed_bytes <
1508 ctx->progress.extract.total_bytes)
1510 ctx->progress.extract.completed_bytes =
1511 ctx->progress.extract.total_bytes;
1512 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1517 ret = extract_progress(ctx,
1518 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1519 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1520 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1522 destroy_blob_list(&ctx->blob_list);
1523 destroy_dentry_list(&dentry_list);
1530 mkdir_if_needed(const tchar *target)
1532 if (!tmkdir(target, 0755))
1535 if (errno == EEXIST)
1539 /* _wmkdir() fails with EACCES if called on a drive root directory. */
1540 if (errno == EACCES)
1544 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
1545 return WIMLIB_ERR_MKDIR;
1548 /* Make sure the extraction flags make sense, and update them if needed. */
1550 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1552 int extract_flags = *extract_flags_p;
1554 /* Check for invalid flag combinations */
1556 if ((extract_flags &
1557 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1558 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1559 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1560 return WIMLIB_ERR_INVALID_PARAM;
1562 if ((extract_flags &
1563 (WIMLIB_EXTRACT_FLAG_RPFIX |
1564 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1565 WIMLIB_EXTRACT_FLAG_NORPFIX))
1566 return WIMLIB_ERR_INVALID_PARAM;
1568 #ifndef WITH_NTFS_3G
1569 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1570 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
1571 " it cannot apply a WIM image directly to an NTFS volume.");
1572 return WIMLIB_ERR_UNSUPPORTED;
1576 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1579 return WIMLIB_ERR_NO_FILENAME;
1581 ERROR("WIMBoot extraction is only supported on Windows!");
1582 return WIMLIB_ERR_UNSUPPORTED;
1587 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1588 WIMLIB_EXTRACT_FLAG_NORPFIX |
1589 WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1590 WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1592 /* For full-image extraction, do reparse point fixups by default
1593 * if the WIM header says they are enabled. */
1594 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1595 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1598 *extract_flags_p = extract_flags;
1603 get_wildcard_flags(int extract_flags)
1605 u32 wildcard_flags = 0;
1607 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB)
1608 wildcard_flags |= WILDCARD_FLAG_ERROR_IF_NO_MATCH;
1610 wildcard_flags |= WILDCARD_FLAG_WARN_IF_NO_MATCH;
1612 if (default_ignore_case)
1613 wildcard_flags |= WILDCARD_FLAG_CASE_INSENSITIVE;
1615 return wildcard_flags;
1618 struct append_dentry_ctx {
1619 struct wim_dentry **dentries;
1620 size_t num_dentries;
1621 size_t num_alloc_dentries;
1625 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1627 struct append_dentry_ctx *ctx = _ctx;
1629 if (ctx->num_dentries == ctx->num_alloc_dentries) {
1630 struct wim_dentry **new_dentries;
1633 new_length = max(ctx->num_alloc_dentries + 8,
1634 ctx->num_alloc_dentries * 3 / 2);
1635 new_dentries = REALLOC(ctx->dentries,
1636 new_length * sizeof(ctx->dentries[0]));
1637 if (new_dentries == NULL)
1638 return WIMLIB_ERR_NOMEM;
1639 ctx->dentries = new_dentries;
1640 ctx->num_alloc_dentries = new_length;
1642 ctx->dentries[ctx->num_dentries++] = dentry;
1647 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1648 const tchar * const *paths, size_t num_paths,
1652 struct wim_dentry **trees;
1655 if (wim == NULL || target == NULL || target[0] == T('\0') ||
1656 (num_paths != 0 && paths == NULL))
1657 return WIMLIB_ERR_INVALID_PARAM;
1659 ret = check_extract_flags(wim, &extract_flags);
1663 ret = select_wim_image(wim, image);
1667 ret = wim_checksum_unhashed_blobs(wim);
1671 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1672 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1673 (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1675 ret = mkdir_if_needed(target);
1680 if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1682 struct append_dentry_ctx append_dentry_ctx = {
1685 .num_alloc_dentries = 0,
1688 u32 wildcard_flags = get_wildcard_flags(extract_flags);
1690 for (size_t i = 0; i < num_paths; i++) {
1691 tchar *path = canonicalize_wim_path(paths[i]);
1693 ret = WIMLIB_ERR_NOMEM;
1694 trees = append_dentry_ctx.dentries;
1695 goto out_free_trees;
1697 ret = expand_wildcard(wim, path,
1703 trees = append_dentry_ctx.dentries;
1704 goto out_free_trees;
1707 trees = append_dentry_ctx.dentries;
1708 num_trees = append_dentry_ctx.num_dentries;
1710 trees = MALLOC(num_paths * sizeof(trees[0]));
1712 return WIMLIB_ERR_NOMEM;
1714 for (size_t i = 0; i < num_paths; i++) {
1716 tchar *path = canonicalize_wim_path(paths[i]);
1718 ret = WIMLIB_ERR_NOMEM;
1719 goto out_free_trees;
1722 trees[i] = get_dentry(wim, path,
1723 WIMLIB_CASE_PLATFORM_DEFAULT);
1725 if (trees[i] == NULL) {
1726 ERROR("Path \"%"TS"\" does not exist "
1728 paths[i], wim->current_image);
1729 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1730 goto out_free_trees;
1733 num_trees = num_paths;
1736 if (num_trees == 0) {
1738 goto out_free_trees;
1741 ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1748 extract_single_image(WIMStruct *wim, int image,
1749 const tchar *target, int extract_flags)
1751 const tchar *path = WIMLIB_WIM_ROOT_PATH;
1752 extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1753 return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1756 static const tchar * const filename_forbidden_chars =
1765 /* This function checks if it is okay to use a WIM image's name as a directory
1768 image_name_ok_as_dir(const tchar *image_name)
1770 return image_name && *image_name &&
1771 !tstrpbrk(image_name, filename_forbidden_chars) &&
1772 tstrcmp(image_name, T(".")) &&
1773 tstrcmp(image_name, T(".."));
1776 /* Extracts all images from the WIM to the directory @target, with the images
1777 * placed in subdirectories named by their image names. */
1779 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1781 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
1782 size_t output_path_len = tstrlen(target);
1783 tchar buf[output_path_len + 1 + image_name_max_len + 1];
1786 const tchar *image_name;
1788 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1789 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1790 return WIMLIB_ERR_INVALID_PARAM;
1793 ret = mkdir_if_needed(target);
1796 tmemcpy(buf, target, output_path_len);
1797 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1798 for (image = 1; image <= wim->hdr.image_count; image++) {
1799 image_name = wimlib_get_image_name(wim, image);
1800 if (image_name_ok_as_dir(image_name)) {
1801 tstrcpy(buf + output_path_len + 1, image_name);
1803 /* Image name is empty or contains forbidden characters.
1804 * Use image number instead. */
1805 tsprintf(buf + output_path_len + 1, T("%d"), image);
1807 ret = extract_single_image(wim, image, buf, extract_flags);
1815 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1818 if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1819 WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1820 WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1821 return WIMLIB_ERR_INVALID_PARAM;
1823 if (image == WIMLIB_ALL_IMAGES)
1824 return extract_all_images(wim, target, extract_flags);
1826 return extract_single_image(wim, image, target, extract_flags);
1830 /****************************************************************************
1832 ****************************************************************************/
1835 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1836 const tchar * const *paths, size_t num_paths,
1839 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1840 return WIMLIB_ERR_INVALID_PARAM;
1842 return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1847 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1848 const tchar *path_list_file, int extract_flags)
1855 ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1857 ERROR("Failed to read path list file \"%"TS"\"",
1862 ret = wimlib_extract_paths(wim, image, target,
1863 (const tchar * const *)paths, num_paths,
1871 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1872 const tchar *image_num_or_name,
1873 const tchar *target,
1875 wimlib_progress_func_t progfunc,
1880 struct filedes *in_fd;
1884 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1885 return WIMLIB_ERR_INVALID_PARAM;
1887 /* Read the WIM header from the pipe and get a WIMStruct to represent
1888 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
1889 * wimlib_open_wim(), getting a WIMStruct in this way will result in an
1890 * empty blob table, no XML data read, and no filename set. */
1891 ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1896 /* Sanity check to make sure this is a pipable WIM. */
1897 if (pwm->hdr.magic != PWM_MAGIC) {
1898 ERROR("The WIM being read from file descriptor %d "
1899 "is not pipable!", pipe_fd);
1900 ret = WIMLIB_ERR_NOT_PIPABLE;
1901 goto out_wimlib_free;
1904 /* Sanity check to make sure the first part of a pipable split WIM is
1905 * sent over the pipe first. */
1906 if (pwm->hdr.part_number != 1) {
1907 ERROR("The first part of the split WIM must be "
1908 "sent over the pipe first.");
1909 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1910 goto out_wimlib_free;
1913 in_fd = &pwm->in_fd;
1914 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1916 /* As mentioned, the WIMStruct we created from the pipe does not have
1917 * XML data yet. Fix this by reading the extra copy of the XML data
1918 * that directly follows the header in pipable WIMs. (Note: see
1919 * write_pipable_wim() for more details about the format of pipable
1922 struct blob_descriptor xml_blob;
1923 struct wim_resource_descriptor xml_rdesc;
1924 ret = read_pwm_blob_header(pwm, &xml_blob, &xml_rdesc, 0, NULL);
1926 goto out_wimlib_free;
1928 if (!xml_blob.is_metadata) {
1929 ERROR("Expected XML data, but found non-metadata resource.");
1930 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1931 goto out_wimlib_free;
1934 wim_res_desc_to_hdr(&xml_rdesc, &pwm->hdr.xml_data_reshdr);
1936 ret = read_wim_xml_data(pwm);
1938 goto out_wimlib_free;
1940 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
1941 ERROR("Image count in XML data is not the same as in WIM header.");
1942 ret = WIMLIB_ERR_IMAGE_COUNT;
1943 goto out_wimlib_free;
1947 /* Get image index (this may use the XML data that was just read to
1948 * resolve an image name). */
1949 if (image_num_or_name) {
1950 image = wimlib_resolve_image(pwm, image_num_or_name);
1951 if (image == WIMLIB_NO_IMAGE) {
1952 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
1954 ret = WIMLIB_ERR_INVALID_IMAGE;
1955 goto out_wimlib_free;
1956 } else if (image == WIMLIB_ALL_IMAGES) {
1957 ERROR("Applying all images from a pipe is not supported!");
1958 ret = WIMLIB_ERR_INVALID_IMAGE;
1959 goto out_wimlib_free;
1962 if (pwm->hdr.image_count != 1) {
1963 ERROR("No image was specified, but the pipable WIM "
1964 "did not contain exactly 1 image");
1965 ret = WIMLIB_ERR_INVALID_IMAGE;
1966 goto out_wimlib_free;
1971 /* Load the needed metadata resource. */
1972 for (i = 1; i <= pwm->hdr.image_count; i++) {
1973 struct blob_descriptor *metadata_blob;
1974 struct wim_image_metadata *imd;
1975 struct wim_resource_descriptor *metadata_rdesc;
1977 metadata_blob = new_blob_descriptor();
1978 if (metadata_blob == NULL) {
1979 ret = WIMLIB_ERR_NOMEM;
1980 goto out_wimlib_free;
1982 metadata_rdesc = MALLOC(sizeof(struct wim_resource_descriptor));
1983 if (metadata_rdesc == NULL) {
1984 ret = WIMLIB_ERR_NOMEM;
1985 free_blob_descriptor(metadata_blob);
1986 goto out_wimlib_free;
1989 ret = read_pwm_blob_header(pwm, metadata_blob, metadata_rdesc, 0, NULL);
1990 imd = pwm->image_metadata[i - 1];
1991 imd->metadata_blob = metadata_blob;
1993 FREE(metadata_rdesc);
1994 goto out_wimlib_free;
1997 if (!metadata_blob->is_metadata) {
1998 ERROR("Expected metadata resource, but found "
1999 "non-metadata resource.");
2000 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2001 goto out_wimlib_free;
2005 /* Metadata resource is for the image being extracted.
2006 * Parse it and save the metadata in memory. */
2007 ret = read_metadata_resource(imd);
2009 goto out_wimlib_free;
2012 /* Metadata resource is not for the image being
2013 * extracted. Skip over it. */
2014 ret = skip_wim_resource(metadata_rdesc);
2016 goto out_wimlib_free;
2019 /* Extract the image. */
2020 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2021 ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
2022 /* Clean up and return. */
2030 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2031 const tchar *target, int extract_flags)
2033 return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
2042 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
2045 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
2046 return WIMLIB_ERR_INVALID_PARAM;
2047 return do_wimlib_extract_image(wim, image, target, extract_flags);