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_nonsolid_wim_resource(blob, rdesc);
185 blob->is_metadata = (rdesc->flags & WIM_RESHDR_FLAG_METADATA) != 0;
189 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
194 read_blobs_from_pipe(struct apply_ctx *ctx,
195 const struct read_blob_list_callbacks *cbs)
197 struct blob_descriptor *found_blob = NULL;
198 struct wim_resource_descriptor *rdesc = NULL;
199 struct blob_table *blob_table;
202 ret = WIMLIB_ERR_NOMEM;
203 found_blob = new_blob_descriptor();
207 rdesc = MALLOC(sizeof(struct wim_resource_descriptor));
211 blob_table = ctx->wim->blob_table;
212 memcpy(ctx->progress.extract.guid, ctx->wim->hdr.guid, WIM_GUID_LEN);
213 ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
214 ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
215 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
219 while (ctx->num_blobs_remaining) {
220 struct wim_header_disk pwm_hdr;
221 struct blob_descriptor *needed_blob;
223 if (found_blob->blob_location != BLOB_NONEXISTENT)
224 blob_unset_is_located_in_wim_resource(found_blob);
225 ret = read_pwm_blob_header(ctx->wim, found_blob, rdesc,
226 PWM_ALLOW_WIM_HDR, &pwm_hdr);
230 if ((found_blob->blob_location != BLOB_NONEXISTENT)
231 && !found_blob->is_metadata
232 && (needed_blob = lookup_blob(blob_table, found_blob->hash))
233 && (needed_blob->out_refcnt))
235 blob_unset_is_located_in_wim_resource(found_blob);
236 blob_set_is_located_in_nonsolid_wim_resource(needed_blob, rdesc);
238 ret = (*cbs->begin_blob)(needed_blob,
239 cbs->begin_blob_ctx);
241 blob_unset_is_located_in_wim_resource(needed_blob);
245 ret = extract_blob(needed_blob, needed_blob->size,
247 cbs->consume_chunk_ctx);
249 ret = (*cbs->end_blob)(needed_blob, ret,
251 blob_unset_is_located_in_wim_resource(needed_blob);
254 ctx->num_blobs_remaining--;
255 } else if (found_blob->blob_location != BLOB_NONEXISTENT) {
256 ret = skip_wim_resource(found_blob->rdesc);
260 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
261 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
263 if (part_number != ctx->progress.extract.part_number ||
264 total_parts != ctx->progress.extract.total_parts ||
265 memcmp(pwm_hdr.guid, ctx->progress.extract.guid,
268 ctx->progress.extract.part_number = part_number;
269 ctx->progress.extract.total_parts = total_parts;
270 memcpy(ctx->progress.extract.guid,
271 pwm_hdr.guid, WIM_GUID_LEN);
272 ret = extract_progress(ctx,
273 WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
281 if (found_blob && found_blob->blob_location != BLOB_IN_WIM)
283 free_blob_descriptor(found_blob);
287 /* Creates a temporary file opened for writing. The open file descriptor is
288 * returned in @fd_ret and its name is returned in @name_ret (dynamically
291 create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
298 name = ttempnam(NULL, T("wimlib"));
300 ERROR_WITH_ERRNO("Failed to create temporary filename");
301 return WIMLIB_ERR_NOMEM;
304 open_flags = O_WRONLY | O_CREAT | O_EXCL | O_BINARY;
306 open_flags |= _O_SHORT_LIVED;
308 raw_fd = topen(name, open_flags, 0600);
311 if (errno == EEXIST) {
315 ERROR_WITH_ERRNO("Failed to create temporary file "
318 return WIMLIB_ERR_OPEN;
321 filedes_init(fd_ret, raw_fd);
327 begin_extract_blob_wrapper(struct blob_descriptor *blob, void *_ctx)
329 struct apply_ctx *ctx = _ctx;
331 ctx->cur_blob = blob;
332 ctx->cur_blob_offset = 0;
334 if (unlikely(blob->out_refcnt > MAX_OPEN_FILES))
335 return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
337 return (*ctx->saved_cbs->begin_blob)(blob, ctx->saved_cbs->begin_blob_ctx);
341 extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx)
343 struct apply_ctx *ctx = _ctx;
344 union wimlib_progress_info *progress = &ctx->progress;
347 ctx->cur_blob_offset += size;
349 if (likely(ctx->supported_features.hard_links)) {
350 progress->extract.completed_bytes +=
351 (u64)size * ctx->cur_blob->out_refcnt;
352 if (ctx->cur_blob_offset == ctx->cur_blob->size)
353 progress->extract.completed_streams += ctx->cur_blob->out_refcnt;
355 const struct blob_extraction_target *targets =
356 blob_extraction_targets(ctx->cur_blob);
357 for (u32 i = 0; i < ctx->cur_blob->out_refcnt; i++) {
358 const struct wim_inode *inode = targets[i].inode;
359 const struct wim_dentry *dentry;
361 list_for_each_entry(dentry,
362 &inode->i_extraction_aliases,
363 d_extraction_alias_node)
365 progress->extract.completed_bytes += size;
366 if (ctx->cur_blob_offset == ctx->cur_blob->size)
367 progress->extract.completed_streams++;
371 if (progress->extract.completed_bytes >= ctx->next_progress) {
373 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
377 if (progress->extract.completed_bytes >=
378 progress->extract.total_bytes)
380 ctx->next_progress = UINT64_MAX;
382 /* Send new message as soon as another 1/128 of the
383 * total has been extracted. (Arbitrary number.) */
385 progress->extract.completed_bytes +
386 progress->extract.total_bytes / 128;
388 /* ... Unless that would be more than 5000000 bytes, in
389 * which case send the next after the next 5000000
390 * bytes. (Another arbitrary number.) */
391 if (progress->extract.completed_bytes + 5000000 <
394 progress->extract.completed_bytes + 5000000;
396 /* ... But always send a message as soon as we're
397 * completely done. */
398 if (progress->extract.total_bytes < ctx->next_progress)
399 ctx->next_progress = progress->extract.total_bytes;
403 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
404 /* Just extracting to temporary file for now. */
405 ret = full_write(&ctx->tmpfile_fd, chunk, size);
407 ERROR_WITH_ERRNO("Error writing data to "
408 "temporary file \"%"TS"\"",
413 return (*ctx->saved_cbs->consume_chunk)(chunk, size,
414 ctx->saved_cbs->consume_chunk_ctx);
419 extract_from_tmpfile(const tchar *tmpfile_name, struct apply_ctx *ctx)
421 struct blob_descriptor tmpfile_blob;
422 struct blob_descriptor *orig_blob = ctx->cur_blob;
423 const struct read_blob_list_callbacks *cbs = ctx->saved_cbs;
425 const u32 orig_refcnt = orig_blob->out_refcnt;
427 BUILD_BUG_ON(MAX_OPEN_FILES <
428 ARRAY_LEN(orig_blob->inline_blob_extraction_targets));
430 struct blob_extraction_target *targets = orig_blob->blob_extraction_targets;
432 /* Copy the blob's data from the temporary file to each of its targets.
434 * This is executed only in the very uncommon case that a blob is being
435 * extracted to more than MAX_OPEN_FILES targets! */
437 memcpy(&tmpfile_blob, orig_blob, sizeof(struct blob_descriptor));
438 tmpfile_blob.blob_location = BLOB_IN_FILE_ON_DISK;
439 tmpfile_blob.file_on_disk = ctx->tmpfile_name;
441 for (u32 i = 0; i < orig_refcnt; i++) {
443 /* Note: it usually doesn't matter whether we pass the original
444 * blob descriptor to callbacks provided by the extraction
445 * backend as opposed to the tmpfile blob descriptor, since they
446 * shouldn't actually read data from the blob other than through
447 * the read_blob_prefix() call below. But for
448 * WIMLIB_EXTRACT_FLAG_WIMBOOT mode on Windows it does matter
449 * because it needs access to the original WIM resource
450 * descriptor in order to create the external backing reference.
453 orig_blob->out_refcnt = 1;
454 orig_blob->inline_blob_extraction_targets[0] = targets[i];
456 ret = (*cbs->begin_blob)(orig_blob, cbs->begin_blob_ctx);
460 /* Extra SHA-1 isn't necessary here, but it shouldn't hurt as
461 * this case is very rare anyway. */
462 ret = extract_blob(&tmpfile_blob, tmpfile_blob.size,
464 cbs->consume_chunk_ctx);
466 ret = (*cbs->end_blob)(orig_blob, ret, cbs->end_blob_ctx);
471 orig_blob->out_refcnt = 0;
476 end_extract_blob_wrapper(struct blob_descriptor *blob, int status, void *_ctx)
478 struct apply_ctx *ctx = _ctx;
480 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
481 filedes_close(&ctx->tmpfile_fd);
483 status = extract_from_tmpfile(ctx->tmpfile_name, ctx);
484 filedes_invalidate(&ctx->tmpfile_fd);
485 tunlink(ctx->tmpfile_name);
486 FREE(ctx->tmpfile_name);
489 return (*ctx->saved_cbs->end_blob)(blob, status,
490 ctx->saved_cbs->end_blob_ctx);
495 * Read the list of blobs to extract and feed their data into the specified
496 * callback functions.
498 * This handles checksumming each blob.
500 * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
502 * This also works if the WIM is being read from a pipe, whereas attempting to
503 * read blobs directly (e.g. with read_full_blob_into_buf()) will not.
505 * This also will split up blobs that will need to be extracted to more than
506 * MAX_OPEN_FILES locations, as measured by the 'out_refcnt' of each blob.
507 * Therefore, the apply_operations implementation need not worry about running
508 * out of file descriptors, unless it might open more than one file descriptor
509 * per 'blob_extraction_target' (e.g. Win32 currently might because the
510 * destination file system might not support hard links).
513 extract_blob_list(struct apply_ctx *ctx,
514 const struct read_blob_list_callbacks *cbs)
516 struct read_blob_list_callbacks wrapper_cbs = {
517 .begin_blob = begin_extract_blob_wrapper,
518 .begin_blob_ctx = ctx,
519 .consume_chunk = extract_chunk_wrapper,
520 .consume_chunk_ctx = ctx,
521 .end_blob = end_extract_blob_wrapper,
524 ctx->saved_cbs = cbs;
525 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
526 return read_blobs_from_pipe(ctx, &wrapper_cbs);
528 return read_blob_list(&ctx->blob_list,
529 offsetof(struct blob_descriptor,
531 &wrapper_cbs, VERIFY_BLOB_HASHES);
535 /* Extract a WIM dentry to standard output.
537 * This obviously doesn't make sense in all cases. We return an error if the
538 * dentry does not correspond to a regular file. Otherwise we extract the
539 * unnamed data stream only. */
541 extract_dentry_to_stdout(struct wim_dentry *dentry,
542 const struct blob_table *blob_table)
544 struct wim_inode *inode = dentry->d_inode;
545 struct blob_descriptor *blob;
546 struct filedes _stdout;
548 if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
549 FILE_ATTRIBUTE_DIRECTORY |
550 FILE_ATTRIBUTE_ENCRYPTED))
552 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
553 "extracted to standard output", dentry_full_path(dentry));
554 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
557 blob = inode_get_blob_for_unnamed_data_stream(inode, blob_table);
559 const u8 *hash = inode_get_hash_of_unnamed_data_stream(inode);
560 if (!is_zero_hash(hash))
561 return blob_not_found_error(inode, hash);
565 filedes_init(&_stdout, STDOUT_FILENO);
566 return extract_full_blob_to_fd(blob, &_stdout);
570 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
571 const struct blob_table *blob_table)
573 for (size_t i = 0; i < num_dentries; i++) {
574 int ret = extract_dentry_to_stdout(dentries[i], blob_table);
581 /**********************************************************************/
584 * Removes duplicate dentries from the array.
586 * Returns the new number of dentries, packed at the front of the array.
589 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
592 for (i = 0; i < num_trees; i++) {
593 if (!trees[i]->tmp_flag) {
594 /* Found distinct dentry. */
595 trees[i]->tmp_flag = 1;
596 trees[j++] = trees[i];
599 for (i = 0; i < j; i++)
600 trees[i]->tmp_flag = 0;
605 * Remove dentries that are descendants of other dentries in the array.
607 * Returns the new number of dentries, packed at the front of the array.
610 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
613 for (i = 0; i < num_trees; i++)
614 trees[i]->tmp_flag = 1;
615 for (i = 0; i < num_trees; i++) {
616 struct wim_dentry *d = trees[i];
617 while (!dentry_is_root(d)) {
622 trees[j++] = trees[i];
626 trees[i]->tmp_flag = 0;
629 for (i = 0; i < j; i++)
630 trees[i]->tmp_flag = 0;
635 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
637 struct list_head *dentry_list = _dentry_list;
638 list_add_tail(&dentry->d_extraction_list_node, dentry_list);
643 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
645 dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
649 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
651 list_del(&dentry->d_extraction_list_node);
652 dentry_reset_extraction_list_node(dentry);
657 * Build the preliminary list of dentries to be extracted.
659 * The list maintains the invariant that if d1 and d2 are in the list and d1 is
660 * an ancestor of d2, then d1 appears before d2 in the list.
663 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
664 size_t num_trees, bool add_ancestors)
666 INIT_LIST_HEAD(dentry_list);
668 /* Add the trees recursively. */
669 for (size_t i = 0; i < num_trees; i++)
670 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
672 /* If requested, add ancestors of the trees. */
674 for (size_t i = 0; i < num_trees; i++) {
675 struct wim_dentry *dentry = trees[i];
676 struct wim_dentry *ancestor;
677 struct list_head *place_after;
679 if (dentry_is_root(dentry))
682 place_after = dentry_list;
685 ancestor = ancestor->d_parent;
686 if (will_extract_dentry(ancestor)) {
687 place_after = &ancestor->d_extraction_list_node;
690 } while (!dentry_is_root(ancestor));
694 ancestor = ancestor->d_parent;
695 if (will_extract_dentry(ancestor))
697 list_add(&ancestor->d_extraction_list_node, place_after);
698 } while (!dentry_is_root(ancestor));
704 destroy_dentry_list(struct list_head *dentry_list)
706 struct wim_dentry *dentry, *tmp;
707 struct wim_inode *inode;
709 list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
710 inode = dentry->d_inode;
711 dentry_reset_extraction_list_node(dentry);
712 inode->i_visited = 0;
713 inode->i_can_externally_back = 0;
714 if ((void *)dentry->d_extraction_name != (void *)dentry->file_name)
715 FREE(dentry->d_extraction_name);
716 dentry->d_extraction_name = NULL;
717 dentry->d_extraction_name_nchars = 0;
722 destroy_blob_list(struct list_head *blob_list)
724 struct blob_descriptor *blob;
726 list_for_each_entry(blob, blob_list, extraction_list)
727 if (blob->out_refcnt > ARRAY_LEN(blob->inline_blob_extraction_targets))
728 FREE(blob->blob_extraction_targets);
732 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
734 static const utf16lechar replacement_char = cpu_to_le16('?');
738 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
744 for (i = 0; i < num_chars; i++) {
747 case cpu_to_le16('\\'):
748 case cpu_to_le16(':'):
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('|'):
756 case cpu_to_le16('/'):
757 case cpu_to_le16('\0'):
759 name[i] = replacement_char;
766 if (name[num_chars - 1] == cpu_to_le16(' ') ||
767 name[num_chars - 1] == cpu_to_le16('.'))
770 name[num_chars - 1] = replacement_char;
779 dentry_calculate_extraction_name(struct wim_dentry *dentry,
780 struct apply_ctx *ctx)
784 if (dentry_is_root(dentry))
788 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
789 dentry->d_extraction_name = dentry->file_name;
790 dentry->d_extraction_name_nchars = dentry->file_name_nbytes /
796 if (!ctx->supported_features.case_sensitive_filenames) {
797 struct wim_dentry *other;
798 list_for_each_entry(other, &dentry->d_ci_conflict_list,
801 if (will_extract_dentry(other)) {
802 if (ctx->extract_flags &
803 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
804 WARNING("\"%"TS"\" has the same "
805 "case-insensitive name as "
806 "\"%"TS"\"; extracting "
807 "dummy name instead",
808 dentry_full_path(dentry),
809 dentry_full_path(other));
812 WARNING("Not extracting \"%"TS"\": "
813 "has same case-insensitive "
815 dentry_full_path(dentry),
816 dentry_full_path(other));
823 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
824 ret = utf16le_get_tstr(dentry->file_name,
825 dentry->file_name_nbytes,
826 (const tchar **)&dentry->d_extraction_name,
827 &dentry->d_extraction_name_nchars);
828 dentry->d_extraction_name_nchars /= sizeof(tchar);
831 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
833 WARNING("\"%"TS"\" has an invalid filename "
834 "that is not supported on this platform; "
835 "extracting dummy name instead",
836 dentry_full_path(dentry));
839 WARNING("Not extracting \"%"TS"\": has an invalid filename "
840 "that is not supported on this platform",
841 dentry_full_path(dentry));
848 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
850 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
851 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
853 const tchar *tchar_name;
856 ret = utf16le_get_tstr(utf16_name_copy,
857 dentry->file_name_nbytes,
858 &tchar_name, &tchar_nchars);
862 tchar_nchars /= sizeof(tchar);
864 size_t fixed_name_num_chars = tchar_nchars;
865 tchar fixed_name[tchar_nchars + 50];
867 tmemcpy(fixed_name, tchar_name, tchar_nchars);
868 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
869 T(" (invalid filename #%lu)"),
870 ++ctx->invalid_sequence);
872 utf16le_put_tstr(tchar_name);
874 dentry->d_extraction_name = TSTRDUP(fixed_name);
875 if (!dentry->d_extraction_name)
876 return WIMLIB_ERR_NOMEM;
877 dentry->d_extraction_name_nchars = fixed_name_num_chars;
882 for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
887 * Calculate the actual filename component at which each WIM dentry will be
888 * extracted, with special handling for dentries that are unsupported by the
889 * extraction backend or have invalid names.
891 * ctx->supported_features must be filled in.
893 * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
896 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
897 struct apply_ctx *ctx)
899 struct list_head *prev, *cur;
901 /* Can't use list_for_each_entry() because a call to
902 * dentry_calculate_extraction_name() may delete the current dentry and
903 * its children from the list. */
907 struct wim_dentry *dentry;
911 if (cur == dentry_list)
914 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
916 ret = dentry_calculate_extraction_name(dentry, ctx);
920 if (prev->next == cur)
923 ; /* Current dentry and its children (which follow in
924 the list) were deleted. prev stays the same. */
930 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
931 struct blob_table *blob_table)
933 struct wim_inode *inode = dentry->d_inode;
934 struct blob_descriptor *blob;
938 /* Special case: when extracting from a pipe, the WIM blob table is
939 * initially empty, so "resolving" an inode's streams is initially not
940 * possible. However, we still need to keep track of which blobs,
941 * identified by SHA-1 message digests, need to be extracted, so we
942 * "resolve" the inode's streams anyway by allocating a 'struct
943 * blob_descriptor' for each one. */
944 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
946 ret = inode_resolve_streams(inode, blob_table, force);
949 for (unsigned i = 0; i < inode->i_num_streams; i++) {
950 blob = stream_blob_resolved(&inode->i_streams[i]);
952 blob->out_refcnt = 0;
958 * For each dentry to be extracted, resolve all streams in the corresponding
959 * inode and set 'out_refcnt' in all referenced blob_descriptors to 0.
961 * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
964 dentry_list_resolve_streams(struct list_head *dentry_list,
965 struct apply_ctx *ctx)
967 struct wim_dentry *dentry;
970 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
971 ret = dentry_resolve_streams(dentry,
973 ctx->wim->blob_table);
981 ref_stream(struct wim_inode_stream *strm, struct wim_dentry *dentry,
982 struct apply_ctx *ctx)
984 struct wim_inode *inode = dentry->d_inode;
985 struct blob_descriptor *blob = stream_blob_resolved(strm);
986 struct blob_extraction_target *targets;
991 /* Tally the size only for each actual extraction of the stream (not
992 * additional hard links to the inode). */
993 if (inode->i_visited && ctx->supported_features.hard_links)
996 ctx->progress.extract.total_bytes += blob->size;
997 ctx->progress.extract.total_streams++;
999 if (inode->i_visited)
1002 /* Add each blob to 'ctx->blob_list' only one time, regardless of how
1003 * many extraction targets it will have. */
1004 if (blob->out_refcnt == 0) {
1005 list_add_tail(&blob->extraction_list, &ctx->blob_list);
1006 ctx->num_blobs_remaining++;
1009 /* Set this stream as an extraction target of 'blob'. */
1011 if (blob->out_refcnt < ARRAY_LEN(blob->inline_blob_extraction_targets)) {
1012 targets = blob->inline_blob_extraction_targets;
1014 struct blob_extraction_target *prev_targets;
1015 size_t alloc_blob_extraction_targets;
1017 if (blob->out_refcnt == ARRAY_LEN(blob->inline_blob_extraction_targets)) {
1018 prev_targets = NULL;
1019 alloc_blob_extraction_targets = ARRAY_LEN(blob->inline_blob_extraction_targets);
1021 prev_targets = blob->blob_extraction_targets;
1022 alloc_blob_extraction_targets = blob->alloc_blob_extraction_targets;
1025 if (blob->out_refcnt == alloc_blob_extraction_targets) {
1026 alloc_blob_extraction_targets *= 2;
1027 targets = REALLOC(prev_targets,
1028 alloc_blob_extraction_targets *
1029 sizeof(targets[0]));
1031 return WIMLIB_ERR_NOMEM;
1032 if (!prev_targets) {
1034 blob->inline_blob_extraction_targets,
1035 sizeof(blob->inline_blob_extraction_targets));
1037 blob->blob_extraction_targets = targets;
1038 blob->alloc_blob_extraction_targets = alloc_blob_extraction_targets;
1040 targets = blob->blob_extraction_targets;
1042 targets[blob->out_refcnt].inode = inode;
1043 targets[blob->out_refcnt].stream = strm;
1049 ref_stream_if_needed(struct wim_dentry *dentry, struct wim_inode *inode,
1050 struct wim_inode_stream *strm, struct apply_ctx *ctx)
1052 bool need_stream = false;
1053 switch (strm->stream_type) {
1054 case STREAM_TYPE_DATA:
1055 if (stream_is_named(strm)) {
1056 /* Named data stream */
1057 if (ctx->supported_features.named_data_streams)
1059 } else if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
1060 FILE_ATTRIBUTE_ENCRYPTED))
1061 && !(inode_is_symlink(inode)
1062 && !ctx->supported_features.reparse_points
1063 && ctx->supported_features.symlink_reparse_points))
1066 * Unnamed data stream. Skip if any of the following is true:
1068 * - file is a directory
1069 * - file is encrypted
1070 * - backend needs to create the file as UNIX symlink
1071 * - backend will extract the stream as externally backed
1073 if (ctx->apply_ops->will_externally_back) {
1074 int ret = (*ctx->apply_ops->will_externally_back)(dentry, ctx);
1075 if (ret > 0) /* Error? */
1077 if (ret < 0) /* Won't externally back? */
1084 case STREAM_TYPE_REPARSE_POINT:
1085 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT);
1086 if (ctx->supported_features.reparse_points ||
1087 (inode_is_symlink(inode) &&
1088 ctx->supported_features.symlink_reparse_points))
1091 case STREAM_TYPE_EFSRPC_RAW_DATA:
1092 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED);
1093 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
1094 if (ctx->supported_features.encrypted_directories)
1097 if (ctx->supported_features.encrypted_files)
1103 return ref_stream(strm, dentry, ctx);
1108 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1110 struct wim_inode *inode = dentry->d_inode;
1111 for (unsigned i = 0; i < inode->i_num_streams; i++) {
1112 int ret = ref_stream_if_needed(dentry, inode,
1113 &inode->i_streams[i], ctx);
1117 inode->i_visited = 1;
1122 * Given a list of dentries to be extracted, build the list of blobs that need
1123 * to be extracted, and for each blob determine the streams to which that blob
1124 * will be extracted.
1126 * This also initializes the extract progress info with byte and blob
1129 * ctx->supported_features must be filled in.
1132 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1134 struct wim_dentry *dentry;
1137 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1138 ret = dentry_ref_streams(dentry, ctx);
1142 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1143 dentry->d_inode->i_visited = 0;
1148 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1150 struct wim_dentry *dentry;
1151 struct wim_inode *inode;
1153 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1154 inode = dentry->d_inode;
1155 if (!inode->i_visited)
1156 INIT_LIST_HEAD(&inode->i_extraction_aliases);
1157 list_add_tail(&dentry->d_extraction_alias_node,
1158 &inode->i_extraction_aliases);
1159 inode->i_visited = 1;
1161 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1162 dentry->d_inode->i_visited = 0;
1166 inode_tally_features(const struct wim_inode *inode,
1167 struct wim_features *features)
1169 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1170 features->archive_files++;
1171 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1172 features->hidden_files++;
1173 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1174 features->system_files++;
1175 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1176 features->compressed_files++;
1177 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1178 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1179 features->encrypted_directories++;
1181 features->encrypted_files++;
1183 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1184 features->not_context_indexed_files++;
1185 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1186 features->sparse_files++;
1187 if (inode_has_named_data_stream(inode))
1188 features->named_data_streams++;
1189 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1190 features->reparse_points++;
1191 if (inode_is_symlink(inode))
1192 features->symlink_reparse_points++;
1194 features->other_reparse_points++;
1196 if (inode->i_security_id != -1)
1197 features->security_descriptors++;
1198 if (inode_has_unix_data(inode))
1199 features->unix_data++;
1202 /* Tally features necessary to extract a dentry and the corresponding inode. */
1204 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1206 struct wim_inode *inode = dentry->d_inode;
1208 if (dentry_has_short_name(dentry))
1209 features->short_names++;
1211 if (inode->i_visited) {
1212 features->hard_links++;
1214 inode_tally_features(inode, features);
1215 inode->i_visited = 1;
1219 /* Tally the features necessary to extract the specified dentries. */
1221 dentry_list_get_features(struct list_head *dentry_list,
1222 struct wim_features *features)
1224 struct wim_dentry *dentry;
1226 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1227 dentry_tally_features(dentry, features);
1229 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1230 dentry->d_inode->i_visited = 0;
1234 do_feature_check(const struct wim_features *required_features,
1235 const struct wim_features *supported_features,
1238 /* Encrypted files. */
1239 if (required_features->encrypted_files &&
1240 !supported_features->encrypted_files)
1241 WARNING("Ignoring EFS-encrypted data of %lu files",
1242 required_features->encrypted_files);
1244 /* Named data streams. */
1245 if (required_features->named_data_streams &&
1246 !supported_features->named_data_streams)
1247 WARNING("Ignoring named data streams of %lu files",
1248 required_features->named_data_streams);
1250 /* File attributes. */
1251 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1252 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1253 * We're an archive program, so theoretically we can do what we
1256 if (required_features->hidden_files &&
1257 !supported_features->hidden_files)
1258 WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1259 required_features->hidden_files);
1261 if (required_features->system_files &&
1262 !supported_features->system_files)
1263 WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1264 required_features->system_files);
1266 if (required_features->compressed_files &&
1267 !supported_features->compressed_files)
1268 WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1269 required_features->compressed_files);
1271 if (required_features->not_context_indexed_files &&
1272 !supported_features->not_context_indexed_files)
1273 WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1274 required_features->not_context_indexed_files);
1276 if (required_features->sparse_files &&
1277 !supported_features->sparse_files)
1278 WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1279 required_features->sparse_files);
1281 if (required_features->encrypted_directories &&
1282 !supported_features->encrypted_directories)
1283 WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1284 required_features->encrypted_directories);
1288 if (required_features->hard_links && !supported_features->hard_links)
1289 WARNING("Extracting %lu hard links as independent files",
1290 required_features->hard_links);
1292 /* Symbolic links and reparse points. */
1293 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1294 required_features->symlink_reparse_points &&
1295 !supported_features->symlink_reparse_points &&
1296 !supported_features->reparse_points)
1298 ERROR("Extraction backend does not support symbolic links!");
1299 return WIMLIB_ERR_UNSUPPORTED;
1301 if (required_features->reparse_points &&
1302 !supported_features->reparse_points)
1304 if (supported_features->symlink_reparse_points) {
1305 if (required_features->other_reparse_points) {
1306 WARNING("Ignoring reparse data of %lu non-symlink/junction files",
1307 required_features->other_reparse_points);
1310 WARNING("Ignoring reparse data of %lu files",
1311 required_features->reparse_points);
1315 /* Security descriptors. */
1316 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1317 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1318 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1319 required_features->security_descriptors &&
1320 !supported_features->security_descriptors)
1322 ERROR("Extraction backend does not support security descriptors!");
1323 return WIMLIB_ERR_UNSUPPORTED;
1325 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1326 required_features->security_descriptors &&
1327 !supported_features->security_descriptors)
1328 WARNING("Ignoring Windows NT security descriptors of %lu files",
1329 required_features->security_descriptors);
1332 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1333 required_features->unix_data && !supported_features->unix_data)
1335 ERROR("Extraction backend does not support UNIX data!");
1336 return WIMLIB_ERR_UNSUPPORTED;
1339 if (required_features->unix_data &&
1340 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1342 WARNING("Ignoring UNIX metadata of %lu files",
1343 required_features->unix_data);
1347 if (required_features->short_names &&
1348 !supported_features->short_names)
1350 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1351 ERROR("Extraction backend does not support DOS names!");
1352 return WIMLIB_ERR_UNSUPPORTED;
1354 WARNING("Ignoring DOS names of %lu files",
1355 required_features->short_names);
1359 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1360 !supported_features->timestamps)
1362 ERROR("Extraction backend does not support timestamps!");
1363 return WIMLIB_ERR_UNSUPPORTED;
1369 static const struct apply_operations *
1370 select_apply_operations(int extract_flags)
1373 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1374 return &ntfs_3g_apply_ops;
1377 return &win32_apply_ops;
1379 return &unix_apply_ops;
1384 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1385 const tchar *target, int extract_flags)
1387 const struct apply_operations *ops;
1388 struct apply_ctx *ctx;
1390 LIST_HEAD(dentry_list);
1392 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1393 ret = extract_dentries_to_stdout(trees, num_trees,
1398 num_trees = remove_duplicate_trees(trees, num_trees);
1399 num_trees = remove_contained_trees(trees, num_trees);
1401 ops = select_apply_operations(extract_flags);
1403 if (num_trees > 1 && ops->single_tree_only) {
1404 ERROR("Extracting multiple directory trees "
1405 "at once is not supported in %s extraction mode!",
1407 ret = WIMLIB_ERR_UNSUPPORTED;
1411 ctx = CALLOC(1, ops->context_size);
1413 ret = WIMLIB_ERR_NOMEM;
1418 ctx->target = target;
1419 ctx->target_nchars = tstrlen(target);
1420 ctx->extract_flags = extract_flags;
1421 if (ctx->wim->progfunc) {
1422 ctx->progfunc = ctx->wim->progfunc;
1423 ctx->progctx = ctx->wim->progctx;
1424 ctx->progress.extract.image = wim->current_image;
1425 ctx->progress.extract.extract_flags = (extract_flags &
1426 WIMLIB_EXTRACT_MASK_PUBLIC);
1427 ctx->progress.extract.wimfile_name = wim->filename;
1428 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1429 wim->current_image);
1430 ctx->progress.extract.target = target;
1432 INIT_LIST_HEAD(&ctx->blob_list);
1433 filedes_invalidate(&ctx->tmpfile_fd);
1434 ctx->apply_ops = ops;
1436 ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1440 build_dentry_list(&dentry_list, trees, num_trees,
1442 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1444 dentry_list_get_features(&dentry_list, &ctx->required_features);
1446 ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1447 ctx->extract_flags);
1451 ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1455 if (unlikely(list_empty(&dentry_list))) {
1456 WARNING("There is nothing to extract!");
1460 ret = dentry_list_resolve_streams(&dentry_list, ctx);
1464 dentry_list_build_inode_alias_lists(&dentry_list);
1466 ret = dentry_list_ref_streams(&dentry_list, ctx);
1470 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1471 /* When extracting from a pipe, the number of bytes of data to
1472 * extract can't be determined in the normal way (examining the
1473 * blob table), since at this point all we have is a set of
1474 * SHA-1 message digests of blobs that need to be extracted.
1475 * However, we can get a reasonably accurate estimate by taking
1476 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1477 * data. This does assume that a full image is being extracted,
1478 * but currently there is no API for doing otherwise. (Also,
1479 * subtract <HARDLINKBYTES> from this if hard links are
1480 * supported by the extraction mode.) */
1481 ctx->progress.extract.total_bytes =
1482 wim_info_get_image_total_bytes(wim->wim_info,
1483 wim->current_image);
1484 if (ctx->supported_features.hard_links) {
1485 ctx->progress.extract.total_bytes -=
1486 wim_info_get_image_hard_link_bytes(wim->wim_info,
1487 wim->current_image);
1491 ret = extract_progress(ctx,
1492 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1493 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1494 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1498 ret = (*ops->extract)(&dentry_list, ctx);
1502 if (ctx->progress.extract.completed_bytes <
1503 ctx->progress.extract.total_bytes)
1505 ctx->progress.extract.completed_bytes =
1506 ctx->progress.extract.total_bytes;
1507 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1512 ret = extract_progress(ctx,
1513 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1514 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1515 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1517 destroy_blob_list(&ctx->blob_list);
1518 destroy_dentry_list(&dentry_list);
1525 mkdir_if_needed(const tchar *target)
1527 if (!tmkdir(target, 0755))
1530 if (errno == EEXIST)
1534 /* _wmkdir() fails with EACCES if called on a drive root directory. */
1535 if (errno == EACCES)
1539 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
1540 return WIMLIB_ERR_MKDIR;
1543 /* Make sure the extraction flags make sense, and update them if needed. */
1545 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1547 int extract_flags = *extract_flags_p;
1549 /* Check for invalid flag combinations */
1551 if ((extract_flags &
1552 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1553 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1554 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1555 return WIMLIB_ERR_INVALID_PARAM;
1557 if ((extract_flags &
1558 (WIMLIB_EXTRACT_FLAG_RPFIX |
1559 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1560 WIMLIB_EXTRACT_FLAG_NORPFIX))
1561 return WIMLIB_ERR_INVALID_PARAM;
1563 #ifndef WITH_NTFS_3G
1564 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1565 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
1566 " it cannot apply a WIM image directly to an NTFS volume.");
1567 return WIMLIB_ERR_UNSUPPORTED;
1571 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1574 return WIMLIB_ERR_NO_FILENAME;
1576 ERROR("WIMBoot extraction is only supported on Windows!");
1577 return WIMLIB_ERR_UNSUPPORTED;
1582 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1583 WIMLIB_EXTRACT_FLAG_NORPFIX |
1584 WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1585 WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1587 /* For full-image extraction, do reparse point fixups by default
1588 * if the WIM header says they are enabled. */
1589 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1590 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1593 *extract_flags_p = extract_flags;
1598 get_wildcard_flags(int extract_flags)
1600 u32 wildcard_flags = 0;
1602 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB)
1603 wildcard_flags |= WILDCARD_FLAG_ERROR_IF_NO_MATCH;
1605 wildcard_flags |= WILDCARD_FLAG_WARN_IF_NO_MATCH;
1607 if (default_ignore_case)
1608 wildcard_flags |= WILDCARD_FLAG_CASE_INSENSITIVE;
1610 return wildcard_flags;
1613 struct append_dentry_ctx {
1614 struct wim_dentry **dentries;
1615 size_t num_dentries;
1616 size_t num_alloc_dentries;
1620 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1622 struct append_dentry_ctx *ctx = _ctx;
1624 if (ctx->num_dentries == ctx->num_alloc_dentries) {
1625 struct wim_dentry **new_dentries;
1628 new_length = max(ctx->num_alloc_dentries + 8,
1629 ctx->num_alloc_dentries * 3 / 2);
1630 new_dentries = REALLOC(ctx->dentries,
1631 new_length * sizeof(ctx->dentries[0]));
1632 if (new_dentries == NULL)
1633 return WIMLIB_ERR_NOMEM;
1634 ctx->dentries = new_dentries;
1635 ctx->num_alloc_dentries = new_length;
1637 ctx->dentries[ctx->num_dentries++] = dentry;
1642 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1643 const tchar * const *paths, size_t num_paths,
1647 struct wim_dentry **trees;
1650 if (wim == NULL || target == NULL || target[0] == T('\0') ||
1651 (num_paths != 0 && paths == NULL))
1652 return WIMLIB_ERR_INVALID_PARAM;
1654 ret = check_extract_flags(wim, &extract_flags);
1658 ret = select_wim_image(wim, image);
1662 ret = wim_checksum_unhashed_blobs(wim);
1666 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1667 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1668 (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1670 ret = mkdir_if_needed(target);
1675 if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1677 struct append_dentry_ctx append_dentry_ctx = {
1680 .num_alloc_dentries = 0,
1683 u32 wildcard_flags = get_wildcard_flags(extract_flags);
1685 for (size_t i = 0; i < num_paths; i++) {
1686 tchar *path = canonicalize_wim_path(paths[i]);
1688 ret = WIMLIB_ERR_NOMEM;
1689 trees = append_dentry_ctx.dentries;
1690 goto out_free_trees;
1692 ret = expand_wildcard(wim, path,
1698 trees = append_dentry_ctx.dentries;
1699 goto out_free_trees;
1702 trees = append_dentry_ctx.dentries;
1703 num_trees = append_dentry_ctx.num_dentries;
1705 trees = MALLOC(num_paths * sizeof(trees[0]));
1707 return WIMLIB_ERR_NOMEM;
1709 for (size_t i = 0; i < num_paths; i++) {
1711 tchar *path = canonicalize_wim_path(paths[i]);
1713 ret = WIMLIB_ERR_NOMEM;
1714 goto out_free_trees;
1717 trees[i] = get_dentry(wim, path,
1718 WIMLIB_CASE_PLATFORM_DEFAULT);
1720 if (trees[i] == NULL) {
1721 ERROR("Path \"%"TS"\" does not exist "
1723 paths[i], wim->current_image);
1724 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1725 goto out_free_trees;
1728 num_trees = num_paths;
1731 if (num_trees == 0) {
1733 goto out_free_trees;
1736 ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1743 extract_single_image(WIMStruct *wim, int image,
1744 const tchar *target, int extract_flags)
1746 const tchar *path = WIMLIB_WIM_ROOT_PATH;
1747 extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1748 return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1751 static const tchar * const filename_forbidden_chars =
1760 /* This function checks if it is okay to use a WIM image's name as a directory
1763 image_name_ok_as_dir(const tchar *image_name)
1765 return image_name && *image_name &&
1766 !tstrpbrk(image_name, filename_forbidden_chars) &&
1767 tstrcmp(image_name, T(".")) &&
1768 tstrcmp(image_name, T(".."));
1771 /* Extracts all images from the WIM to the directory @target, with the images
1772 * placed in subdirectories named by their image names. */
1774 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1776 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
1777 size_t output_path_len = tstrlen(target);
1778 tchar buf[output_path_len + 1 + image_name_max_len + 1];
1781 const tchar *image_name;
1783 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1784 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1785 return WIMLIB_ERR_INVALID_PARAM;
1788 ret = mkdir_if_needed(target);
1791 tmemcpy(buf, target, output_path_len);
1792 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1793 for (image = 1; image <= wim->hdr.image_count; image++) {
1794 image_name = wimlib_get_image_name(wim, image);
1795 if (image_name_ok_as_dir(image_name)) {
1796 tstrcpy(buf + output_path_len + 1, image_name);
1798 /* Image name is empty or contains forbidden characters.
1799 * Use image number instead. */
1800 tsprintf(buf + output_path_len + 1, T("%d"), image);
1802 ret = extract_single_image(wim, image, buf, extract_flags);
1810 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1813 if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1814 WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1815 WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1816 return WIMLIB_ERR_INVALID_PARAM;
1818 if (image == WIMLIB_ALL_IMAGES)
1819 return extract_all_images(wim, target, extract_flags);
1821 return extract_single_image(wim, image, target, extract_flags);
1825 /****************************************************************************
1827 ****************************************************************************/
1830 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1831 const tchar * const *paths, size_t num_paths,
1834 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1835 return WIMLIB_ERR_INVALID_PARAM;
1837 return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1842 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1843 const tchar *path_list_file, int extract_flags)
1850 ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1852 ERROR("Failed to read path list file \"%"TS"\"",
1857 ret = wimlib_extract_paths(wim, image, target,
1858 (const tchar * const *)paths, num_paths,
1866 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1867 const tchar *image_num_or_name,
1868 const tchar *target,
1870 wimlib_progress_func_t progfunc,
1875 struct filedes *in_fd;
1879 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1880 return WIMLIB_ERR_INVALID_PARAM;
1882 /* Read the WIM header from the pipe and get a WIMStruct to represent
1883 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
1884 * wimlib_open_wim(), getting a WIMStruct in this way will result in an
1885 * empty blob table, no XML data read, and no filename set. */
1886 ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1891 /* Sanity check to make sure this is a pipable WIM. */
1892 if (pwm->hdr.magic != PWM_MAGIC) {
1893 ERROR("The WIM being read from file descriptor %d "
1894 "is not pipable!", pipe_fd);
1895 ret = WIMLIB_ERR_NOT_PIPABLE;
1896 goto out_wimlib_free;
1899 /* Sanity check to make sure the first part of a pipable split WIM is
1900 * sent over the pipe first. */
1901 if (pwm->hdr.part_number != 1) {
1902 ERROR("The first part of the split WIM must be "
1903 "sent over the pipe first.");
1904 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1905 goto out_wimlib_free;
1908 in_fd = &pwm->in_fd;
1909 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1911 /* As mentioned, the WIMStruct we created from the pipe does not have
1912 * XML data yet. Fix this by reading the extra copy of the XML data
1913 * that directly follows the header in pipable WIMs. (Note: see
1914 * write_pipable_wim() for more details about the format of pipable
1917 struct blob_descriptor xml_blob;
1918 struct wim_resource_descriptor xml_rdesc;
1919 ret = read_pwm_blob_header(pwm, &xml_blob, &xml_rdesc, 0, NULL);
1921 goto out_wimlib_free;
1923 if (!xml_blob.is_metadata) {
1924 ERROR("Expected XML data, but found non-metadata resource.");
1925 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1926 goto out_wimlib_free;
1929 wim_res_desc_to_hdr(&xml_rdesc, &pwm->hdr.xml_data_reshdr);
1931 ret = read_wim_xml_data(pwm);
1933 goto out_wimlib_free;
1935 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
1936 ERROR("Image count in XML data is not the same as in WIM header.");
1937 ret = WIMLIB_ERR_IMAGE_COUNT;
1938 goto out_wimlib_free;
1942 /* Get image index (this may use the XML data that was just read to
1943 * resolve an image name). */
1944 if (image_num_or_name) {
1945 image = wimlib_resolve_image(pwm, image_num_or_name);
1946 if (image == WIMLIB_NO_IMAGE) {
1947 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
1949 ret = WIMLIB_ERR_INVALID_IMAGE;
1950 goto out_wimlib_free;
1951 } else if (image == WIMLIB_ALL_IMAGES) {
1952 ERROR("Applying all images from a pipe is not supported!");
1953 ret = WIMLIB_ERR_INVALID_IMAGE;
1954 goto out_wimlib_free;
1957 if (pwm->hdr.image_count != 1) {
1958 ERROR("No image was specified, but the pipable WIM "
1959 "did not contain exactly 1 image");
1960 ret = WIMLIB_ERR_INVALID_IMAGE;
1961 goto out_wimlib_free;
1966 /* Load the needed metadata resource. */
1967 for (i = 1; i <= pwm->hdr.image_count; i++) {
1968 struct blob_descriptor *metadata_blob;
1969 struct wim_image_metadata *imd;
1970 struct wim_resource_descriptor *metadata_rdesc;
1972 metadata_blob = new_blob_descriptor();
1973 if (metadata_blob == NULL) {
1974 ret = WIMLIB_ERR_NOMEM;
1975 goto out_wimlib_free;
1977 metadata_rdesc = MALLOC(sizeof(struct wim_resource_descriptor));
1978 if (metadata_rdesc == NULL) {
1979 ret = WIMLIB_ERR_NOMEM;
1980 free_blob_descriptor(metadata_blob);
1981 goto out_wimlib_free;
1984 ret = read_pwm_blob_header(pwm, metadata_blob, metadata_rdesc, 0, NULL);
1985 imd = pwm->image_metadata[i - 1];
1986 imd->metadata_blob = metadata_blob;
1988 FREE(metadata_rdesc);
1989 goto out_wimlib_free;
1992 if (!metadata_blob->is_metadata) {
1993 ERROR("Expected metadata resource, but found "
1994 "non-metadata resource.");
1995 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1996 goto out_wimlib_free;
2000 /* Metadata resource is for the image being extracted.
2001 * Parse it and save the metadata in memory. */
2002 ret = read_metadata_resource(imd);
2004 goto out_wimlib_free;
2007 /* Metadata resource is not for the image being
2008 * extracted. Skip over it. */
2009 ret = skip_wim_resource(metadata_rdesc);
2011 goto out_wimlib_free;
2014 /* Extract the image. */
2015 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2016 ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
2017 /* Clean up and return. */
2025 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2026 const tchar *target, int extract_flags)
2028 return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
2037 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
2040 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
2041 return WIMLIB_ERR_INVALID_PARAM;
2042 return do_wimlib_extract_image(wim, image, target, extract_flags);