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->flags = rdesc->flags;
186 blob->size = rdesc->uncompressed_size;
187 blob->offset_in_res = 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->flags & WIM_RESHDR_FLAG_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->flags = found_blob->flags;
239 needed_blob->size = found_blob->size;
241 blob_unset_is_located_in_wim_resource(found_blob);
242 blob_set_is_located_in_wim_resource(needed_blob, rdesc);
244 ret = (*cbs->begin_blob)(needed_blob,
245 cbs->begin_blob_ctx);
247 blob_unset_is_located_in_wim_resource(needed_blob);
251 ret = extract_blob(needed_blob, needed_blob->size,
253 cbs->consume_chunk_ctx);
255 ret = (*cbs->end_blob)(needed_blob, ret,
257 blob_unset_is_located_in_wim_resource(needed_blob);
260 ctx->num_blobs_remaining--;
261 } else if (found_blob->blob_location != BLOB_NONEXISTENT) {
262 ret = skip_wim_resource(found_blob->rdesc);
266 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
267 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
269 if (part_number != ctx->progress.extract.part_number ||
270 total_parts != ctx->progress.extract.total_parts ||
271 memcmp(pwm_hdr.guid, ctx->progress.extract.guid,
274 ctx->progress.extract.part_number = part_number;
275 ctx->progress.extract.total_parts = total_parts;
276 memcpy(ctx->progress.extract.guid,
277 pwm_hdr.guid, WIM_GUID_LEN);
278 ret = extract_progress(ctx,
279 WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
287 if (found_blob && found_blob->blob_location != BLOB_IN_WIM)
289 free_blob_descriptor(found_blob);
293 /* Creates a temporary file opened for writing. The open file descriptor is
294 * returned in @fd_ret and its name is returned in @name_ret (dynamically
297 create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
304 name = ttempnam(NULL, T("wimlib"));
306 ERROR_WITH_ERRNO("Failed to create temporary filename");
307 return WIMLIB_ERR_NOMEM;
310 open_flags = O_WRONLY | O_CREAT | O_EXCL | O_BINARY;
312 open_flags |= _O_SHORT_LIVED;
314 raw_fd = topen(name, open_flags, 0600);
317 if (errno == EEXIST) {
321 ERROR_WITH_ERRNO("Failed to create temporary file "
324 return WIMLIB_ERR_OPEN;
327 filedes_init(fd_ret, raw_fd);
333 begin_extract_blob_wrapper(struct blob_descriptor *blob, void *_ctx)
335 struct apply_ctx *ctx = _ctx;
337 ctx->cur_blob = blob;
338 ctx->cur_blob_offset = 0;
340 if (unlikely(blob->out_refcnt > MAX_OPEN_FILES))
341 return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
343 return (*ctx->saved_cbs->begin_blob)(blob, ctx->saved_cbs->begin_blob_ctx);
347 extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx)
349 struct apply_ctx *ctx = _ctx;
350 union wimlib_progress_info *progress = &ctx->progress;
353 ctx->cur_blob_offset += size;
355 if (likely(ctx->supported_features.hard_links)) {
356 progress->extract.completed_bytes +=
357 (u64)size * ctx->cur_blob->out_refcnt;
358 if (ctx->cur_blob_offset == ctx->cur_blob->size)
359 progress->extract.completed_streams += ctx->cur_blob->out_refcnt;
361 const struct blob_extraction_target *targets =
362 blob_extraction_targets(ctx->cur_blob);
363 for (u32 i = 0; i < ctx->cur_blob->out_refcnt; i++) {
364 const struct wim_inode *inode = targets[i].inode;
365 const struct wim_dentry *dentry;
367 list_for_each_entry(dentry,
368 &inode->i_extraction_aliases,
369 d_extraction_alias_node)
371 progress->extract.completed_bytes += size;
372 if (ctx->cur_blob_offset == ctx->cur_blob->size)
373 progress->extract.completed_streams++;
377 if (progress->extract.completed_bytes >= ctx->next_progress) {
379 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
383 if (progress->extract.completed_bytes >=
384 progress->extract.total_bytes)
386 ctx->next_progress = UINT64_MAX;
388 /* Send new message as soon as another 1/128 of the
389 * total has been extracted. (Arbitrary number.) */
391 progress->extract.completed_bytes +
392 progress->extract.total_bytes / 128;
394 /* ... Unless that would be more than 5000000 bytes, in
395 * which case send the next after the next 5000000
396 * bytes. (Another arbitrary number.) */
397 if (progress->extract.completed_bytes + 5000000 <
400 progress->extract.completed_bytes + 5000000;
402 /* ... But always send a message as soon as we're
403 * completely done. */
404 if (progress->extract.total_bytes < ctx->next_progress)
405 ctx->next_progress = progress->extract.total_bytes;
409 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
410 /* Just extracting to temporary file for now. */
411 ret = full_write(&ctx->tmpfile_fd, chunk, size);
413 ERROR_WITH_ERRNO("Error writing data to "
414 "temporary file \"%"TS"\"",
419 return (*ctx->saved_cbs->consume_chunk)(chunk, size,
420 ctx->saved_cbs->consume_chunk_ctx);
425 extract_from_tmpfile(const tchar *tmpfile_name, struct apply_ctx *ctx)
427 struct blob_descriptor tmpfile_blob;
428 struct blob_descriptor *orig_blob = ctx->cur_blob;
429 const struct read_blob_list_callbacks *cbs = ctx->saved_cbs;
431 const u32 orig_refcnt = orig_blob->out_refcnt;
433 BUILD_BUG_ON(MAX_OPEN_FILES <
434 ARRAY_LEN(orig_blob->inline_blob_extraction_targets));
436 struct blob_extraction_target *targets = orig_blob->blob_extraction_targets;
438 /* Copy the blob's data from the temporary file to each of its targets.
440 * This is executed only in the very uncommon case that a blob is being
441 * extracted to more than MAX_OPEN_FILES targets! */
443 memcpy(&tmpfile_blob, orig_blob, sizeof(struct blob_descriptor));
444 tmpfile_blob.blob_location = BLOB_IN_FILE_ON_DISK;
445 tmpfile_blob.file_on_disk = ctx->tmpfile_name;
447 for (u32 i = 0; i < orig_refcnt; i++) {
449 /* Note: it usually doesn't matter whether we pass the original
450 * blob descriptor to callbacks provided by the extraction
451 * backend as opposed to the tmpfile blob descriptor, since they
452 * shouldn't actually read data from the blob other than through
453 * the read_blob_prefix() call below. But for
454 * WIMLIB_EXTRACT_FLAG_WIMBOOT mode on Windows it does matter
455 * because it needs access to the original WIM resource
456 * descriptor in order to create the external backing reference.
459 orig_blob->out_refcnt = 1;
460 orig_blob->inline_blob_extraction_targets[0] = targets[i];
462 ret = (*cbs->begin_blob)(orig_blob, cbs->begin_blob_ctx);
466 /* Extra SHA-1 isn't necessary here, but it shouldn't hurt as
467 * this case is very rare anyway. */
468 ret = extract_blob(&tmpfile_blob, tmpfile_blob.size,
470 cbs->consume_chunk_ctx);
472 ret = (*cbs->end_blob)(orig_blob, ret, cbs->end_blob_ctx);
477 orig_blob->out_refcnt = 0;
482 end_extract_blob_wrapper(struct blob_descriptor *blob, int status, void *_ctx)
484 struct apply_ctx *ctx = _ctx;
486 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
487 filedes_close(&ctx->tmpfile_fd);
489 status = extract_from_tmpfile(ctx->tmpfile_name, ctx);
490 filedes_invalidate(&ctx->tmpfile_fd);
491 tunlink(ctx->tmpfile_name);
492 FREE(ctx->tmpfile_name);
495 return (*ctx->saved_cbs->end_blob)(blob, status,
496 ctx->saved_cbs->end_blob_ctx);
501 * Read the list of blobs to extract and feed their data into the specified
502 * callback functions.
504 * This handles checksumming each blob.
506 * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
508 * This also works if the WIM is being read from a pipe, whereas attempting to
509 * read blobs directly (e.g. with read_full_blob_into_buf()) will not.
511 * This also will split up blobs that will need to be extracted to more than
512 * MAX_OPEN_FILES locations, as measured by the 'out_refcnt' of each blob.
513 * Therefore, the apply_operations implementation need not worry about running
514 * out of file descriptors, unless it might open more than one file descriptor
515 * per nominal destination (e.g. Win32 currently might because the destination
516 * file system might not support hard links).
519 extract_blob_list(struct apply_ctx *ctx,
520 const struct read_blob_list_callbacks *cbs)
522 struct read_blob_list_callbacks wrapper_cbs = {
523 .begin_blob = begin_extract_blob_wrapper,
524 .begin_blob_ctx = ctx,
525 .consume_chunk = extract_chunk_wrapper,
526 .consume_chunk_ctx = ctx,
527 .end_blob = end_extract_blob_wrapper,
530 ctx->saved_cbs = cbs;
531 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
532 return read_blobs_from_pipe(ctx, &wrapper_cbs);
534 return read_blob_list(&ctx->blob_list,
535 offsetof(struct blob_descriptor,
537 &wrapper_cbs, VERIFY_BLOB_HASHES);
541 /* Extract a WIM dentry to standard output.
543 * This obviously doesn't make sense in all cases. We return an error if the
544 * dentry does not correspond to a regular file. Otherwise we extract the
545 * unnamed data stream only. */
547 extract_dentry_to_stdout(struct wim_dentry *dentry,
548 const struct blob_table *blob_table)
550 struct wim_inode *inode = dentry->d_inode;
551 struct blob_descriptor *blob;
552 struct filedes _stdout;
554 if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
555 FILE_ATTRIBUTE_DIRECTORY |
556 FILE_ATTRIBUTE_ENCRYPTED))
558 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
559 "extracted to standard output", dentry_full_path(dentry));
560 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
563 blob = inode_get_blob_for_unnamed_data_stream(inode, blob_table);
565 const u8 *hash = inode_get_hash_of_unnamed_data_stream(inode);
566 if (!is_zero_hash(hash))
567 return blob_not_found_error(inode, hash);
571 filedes_init(&_stdout, STDOUT_FILENO);
572 return extract_full_blob_to_fd(blob, &_stdout);
576 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
577 const struct blob_table *blob_table)
579 for (size_t i = 0; i < num_dentries; i++) {
580 int ret = extract_dentry_to_stdout(dentries[i], blob_table);
587 /**********************************************************************/
590 * Removes duplicate dentries from the array.
592 * Returns the new number of dentries, packed at the front of the array.
595 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
598 for (i = 0; i < num_trees; i++) {
599 if (!trees[i]->tmp_flag) {
600 /* Found distinct dentry. */
601 trees[i]->tmp_flag = 1;
602 trees[j++] = trees[i];
605 for (i = 0; i < j; i++)
606 trees[i]->tmp_flag = 0;
611 * Remove dentries that are descendants of other dentries in the array.
613 * Returns the new number of dentries, packed at the front of the array.
616 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
619 for (i = 0; i < num_trees; i++)
620 trees[i]->tmp_flag = 1;
621 for (i = 0; i < num_trees; i++) {
622 struct wim_dentry *d = trees[i];
623 while (!dentry_is_root(d)) {
628 trees[j++] = trees[i];
632 trees[i]->tmp_flag = 0;
635 for (i = 0; i < j; i++)
636 trees[i]->tmp_flag = 0;
641 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
643 struct list_head *dentry_list = _dentry_list;
644 list_add_tail(&dentry->d_extraction_list_node, dentry_list);
649 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
651 dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
655 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
657 list_del(&dentry->d_extraction_list_node);
658 dentry_reset_extraction_list_node(dentry);
663 * Build the preliminary list of dentries to be extracted.
665 * The list maintains the invariant that if d1 and d2 are in the list and d1 is
666 * an ancestor of d2, then d1 appears before d2 in the list.
669 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
670 size_t num_trees, bool add_ancestors)
672 INIT_LIST_HEAD(dentry_list);
674 /* Add the trees recursively. */
675 for (size_t i = 0; i < num_trees; i++)
676 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
678 /* If requested, add ancestors of the trees. */
680 for (size_t i = 0; i < num_trees; i++) {
681 struct wim_dentry *dentry = trees[i];
682 struct wim_dentry *ancestor;
683 struct list_head *place_after;
685 if (dentry_is_root(dentry))
688 place_after = dentry_list;
691 ancestor = ancestor->d_parent;
692 if (will_extract_dentry(ancestor)) {
693 place_after = &ancestor->d_extraction_list_node;
696 } while (!dentry_is_root(ancestor));
700 ancestor = ancestor->d_parent;
701 if (will_extract_dentry(ancestor))
703 list_add(&ancestor->d_extraction_list_node, place_after);
704 } while (!dentry_is_root(ancestor));
710 destroy_dentry_list(struct list_head *dentry_list)
712 struct wim_dentry *dentry, *tmp;
713 struct wim_inode *inode;
715 list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
716 inode = dentry->d_inode;
717 dentry_reset_extraction_list_node(dentry);
718 inode->i_visited = 0;
719 inode->i_can_externally_back = 0;
720 if ((void *)dentry->d_extraction_name != (void *)dentry->file_name)
721 FREE(dentry->d_extraction_name);
722 dentry->d_extraction_name = NULL;
723 dentry->d_extraction_name_nchars = 0;
728 destroy_blob_list(struct list_head *blob_list)
730 struct blob_descriptor *blob;
732 list_for_each_entry(blob, blob_list, extraction_list)
733 if (blob->out_refcnt > ARRAY_LEN(blob->inline_blob_extraction_targets))
734 FREE(blob->blob_extraction_targets);
738 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
740 static const utf16lechar replacement_char = cpu_to_le16('?');
744 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
750 for (i = 0; i < num_chars; i++) {
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('>'):
760 case cpu_to_le16('|'):
762 case cpu_to_le16('/'):
763 case cpu_to_le16('\0'):
765 name[i] = replacement_char;
772 if (name[num_chars - 1] == cpu_to_le16(' ') ||
773 name[num_chars - 1] == cpu_to_le16('.'))
776 name[num_chars - 1] = replacement_char;
785 dentry_calculate_extraction_name(struct wim_dentry *dentry,
786 struct apply_ctx *ctx)
790 if (dentry_is_root(dentry))
794 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
795 dentry->d_extraction_name = dentry->file_name;
796 dentry->d_extraction_name_nchars = dentry->file_name_nbytes /
802 if (!ctx->supported_features.case_sensitive_filenames) {
803 struct wim_dentry *other;
804 list_for_each_entry(other, &dentry->d_ci_conflict_list,
807 if (will_extract_dentry(other)) {
808 if (ctx->extract_flags &
809 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
810 WARNING("\"%"TS"\" has the same "
811 "case-insensitive name as "
812 "\"%"TS"\"; extracting "
813 "dummy name instead",
814 dentry_full_path(dentry),
815 dentry_full_path(other));
818 WARNING("Not extracting \"%"TS"\": "
819 "has same case-insensitive "
821 dentry_full_path(dentry),
822 dentry_full_path(other));
829 if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
830 ret = utf16le_get_tstr(dentry->file_name,
831 dentry->file_name_nbytes,
832 (const tchar **)&dentry->d_extraction_name,
833 &dentry->d_extraction_name_nchars);
834 dentry->d_extraction_name_nchars /= sizeof(tchar);
837 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
839 WARNING("\"%"TS"\" has an invalid filename "
840 "that is not supported on this platform; "
841 "extracting dummy name instead",
842 dentry_full_path(dentry));
845 WARNING("Not extracting \"%"TS"\": has an invalid filename "
846 "that is not supported on this platform",
847 dentry_full_path(dentry));
854 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
856 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
857 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
859 const tchar *tchar_name;
862 ret = utf16le_get_tstr(utf16_name_copy,
863 dentry->file_name_nbytes,
864 &tchar_name, &tchar_nchars);
868 tchar_nchars /= sizeof(tchar);
870 size_t fixed_name_num_chars = tchar_nchars;
871 tchar fixed_name[tchar_nchars + 50];
873 tmemcpy(fixed_name, tchar_name, tchar_nchars);
874 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
875 T(" (invalid filename #%lu)"),
876 ++ctx->invalid_sequence);
878 utf16le_put_tstr(tchar_name);
880 dentry->d_extraction_name = memdup(fixed_name,
881 2 * fixed_name_num_chars + 2);
882 if (!dentry->d_extraction_name)
883 return WIMLIB_ERR_NOMEM;
884 dentry->d_extraction_name_nchars = fixed_name_num_chars;
889 for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
894 * Calculate the actual filename component at which each WIM dentry will be
895 * extracted, with special handling for dentries that are unsupported by the
896 * extraction backend or have invalid names.
898 * ctx->supported_features must be filled in.
900 * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
903 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
904 struct apply_ctx *ctx)
906 struct list_head *prev, *cur;
908 /* Can't use list_for_each_entry() because a call to
909 * dentry_calculate_extraction_name() may delete the current dentry and
910 * its children from the list. */
914 struct wim_dentry *dentry;
918 if (cur == dentry_list)
921 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
923 ret = dentry_calculate_extraction_name(dentry, ctx);
927 if (prev->next == cur)
930 ; /* Current dentry and its children (which follow in
931 the list) were deleted. prev stays the same. */
937 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
938 struct blob_table *blob_table)
940 struct wim_inode *inode = dentry->d_inode;
941 struct blob_descriptor *blob;
945 /* Special case: when extracting from a pipe, the WIM blob table is
946 * initially empty, so "resolving" an inode's streams is initially not
947 * possible. However, we still need to keep track of which blobs,
948 * identified by SHA-1 message digests, need to be extracted, so we
949 * "resolve" the inode's streams anyway by allocating a 'struct
950 * blob_descriptor' for each one. */
951 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
953 ret = inode_resolve_streams(inode, blob_table, force);
956 for (unsigned i = 0; i < inode->i_num_streams; i++) {
957 blob = stream_blob_resolved(&inode->i_streams[i]);
959 blob->out_refcnt = 0;
965 * For each dentry to be extracted, resolve all streams in the corresponding
966 * inode and set 'out_refcnt' in all referenced blob_descriptors to 0.
968 * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
971 dentry_list_resolve_streams(struct list_head *dentry_list,
972 struct apply_ctx *ctx)
974 struct wim_dentry *dentry;
977 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
978 ret = dentry_resolve_streams(dentry,
980 ctx->wim->blob_table);
988 ref_stream(struct wim_inode_stream *strm, struct wim_dentry *dentry,
989 struct apply_ctx *ctx)
991 struct wim_inode *inode = dentry->d_inode;
992 struct blob_descriptor *blob = stream_blob_resolved(strm);
993 struct blob_extraction_target *targets;
998 /* Tally the size only for each actual extraction of the stream (not
999 * additional hard links to the inode). */
1000 if (inode->i_visited && ctx->supported_features.hard_links)
1003 ctx->progress.extract.total_bytes += blob->size;
1004 ctx->progress.extract.total_streams++;
1006 if (inode->i_visited)
1009 /* Add each blob to 'ctx->blob_list' only one time, regardless of how
1010 * many extraction targets it will have. */
1011 if (blob->out_refcnt == 0) {
1012 list_add_tail(&blob->extraction_list, &ctx->blob_list);
1013 ctx->num_blobs_remaining++;
1016 /* Set this stream as an extraction target of 'blob'. */
1018 if (blob->out_refcnt < ARRAY_LEN(blob->inline_blob_extraction_targets)) {
1019 targets = blob->inline_blob_extraction_targets;
1021 struct blob_extraction_target *prev_targets;
1022 size_t alloc_blob_extraction_targets;
1024 if (blob->out_refcnt == ARRAY_LEN(blob->inline_blob_extraction_targets)) {
1025 prev_targets = NULL;
1026 alloc_blob_extraction_targets = ARRAY_LEN(blob->inline_blob_extraction_targets);
1028 prev_targets = blob->blob_extraction_targets;
1029 alloc_blob_extraction_targets = blob->alloc_blob_extraction_targets;
1032 if (blob->out_refcnt == alloc_blob_extraction_targets) {
1033 alloc_blob_extraction_targets *= 2;
1034 targets = REALLOC(prev_targets,
1035 alloc_blob_extraction_targets *
1036 sizeof(targets[0]));
1038 return WIMLIB_ERR_NOMEM;
1039 if (!prev_targets) {
1041 blob->inline_blob_extraction_targets,
1042 sizeof(blob->inline_blob_extraction_targets));
1044 blob->blob_extraction_targets = targets;
1045 blob->alloc_blob_extraction_targets = alloc_blob_extraction_targets;
1047 targets = blob->blob_extraction_targets;
1049 targets[blob->out_refcnt].inode = inode;
1050 targets[blob->out_refcnt].stream = strm;
1056 ref_stream_if_needed(struct wim_dentry *dentry, struct wim_inode *inode,
1057 struct wim_inode_stream *strm, struct apply_ctx *ctx)
1059 bool need_stream = false;
1060 switch (strm->stream_type) {
1061 case STREAM_TYPE_DATA:
1062 if (stream_is_named(strm)) {
1063 /* Named data stream */
1064 if (ctx->supported_features.named_data_streams)
1066 } else if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
1067 FILE_ATTRIBUTE_ENCRYPTED))
1068 && !(inode_is_symlink(inode)
1069 && !ctx->supported_features.reparse_points
1070 && ctx->supported_features.symlink_reparse_points))
1073 * Unnamed data stream. Skip if any of the following is true:
1075 * - file is a directory
1076 * - file is encrypted
1077 * - backend needs to create the file as UNIX symlink
1078 * - backend will extract the stream as externally backed
1080 if (ctx->apply_ops->will_externally_back) {
1081 int ret = (*ctx->apply_ops->will_externally_back)(dentry, ctx);
1082 if (ret > 0) /* Error? */
1084 if (ret < 0) /* Won't externally back? */
1091 case STREAM_TYPE_REPARSE_POINT:
1092 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT);
1093 if (ctx->supported_features.reparse_points ||
1094 (inode_is_symlink(inode) &&
1095 ctx->supported_features.symlink_reparse_points))
1098 case STREAM_TYPE_EFSRPC_RAW_DATA:
1099 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED);
1100 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
1101 if (ctx->supported_features.encrypted_directories)
1104 if (ctx->supported_features.encrypted_files)
1110 return ref_stream(strm, dentry, ctx);
1115 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1117 struct wim_inode *inode = dentry->d_inode;
1118 for (unsigned i = 0; i < inode->i_num_streams; i++) {
1119 int ret = ref_stream_if_needed(dentry, inode,
1120 &inode->i_streams[i], ctx);
1124 inode->i_visited = 1;
1129 * Given a list of dentries to be extracted, build the list of blobs that need
1130 * to be extracted, and for each blob determine the streams to which that blob
1131 * will be extracted.
1133 * This also initializes the extract progress info with byte and blob
1136 * ctx->supported_features must be filled in.
1139 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1141 struct wim_dentry *dentry;
1144 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1145 ret = dentry_ref_streams(dentry, ctx);
1149 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1150 dentry->d_inode->i_visited = 0;
1155 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1157 struct wim_dentry *dentry;
1158 struct wim_inode *inode;
1160 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1161 inode = dentry->d_inode;
1162 if (!inode->i_visited)
1163 INIT_LIST_HEAD(&inode->i_extraction_aliases);
1164 list_add_tail(&dentry->d_extraction_alias_node,
1165 &inode->i_extraction_aliases);
1166 inode->i_visited = 1;
1168 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1169 dentry->d_inode->i_visited = 0;
1173 inode_tally_features(const struct wim_inode *inode,
1174 struct wim_features *features)
1176 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1177 features->archive_files++;
1178 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1179 features->hidden_files++;
1180 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1181 features->system_files++;
1182 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1183 features->compressed_files++;
1184 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1185 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1186 features->encrypted_directories++;
1188 features->encrypted_files++;
1190 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1191 features->not_context_indexed_files++;
1192 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1193 features->sparse_files++;
1194 if (inode_has_named_data_stream(inode))
1195 features->named_data_streams++;
1196 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1197 features->reparse_points++;
1198 if (inode_is_symlink(inode))
1199 features->symlink_reparse_points++;
1201 features->other_reparse_points++;
1203 if (inode->i_security_id != -1)
1204 features->security_descriptors++;
1205 if (inode_has_unix_data(inode))
1206 features->unix_data++;
1209 /* Tally features necessary to extract a dentry and the corresponding inode. */
1211 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1213 struct wim_inode *inode = dentry->d_inode;
1215 if (dentry_has_short_name(dentry))
1216 features->short_names++;
1218 if (inode->i_visited) {
1219 features->hard_links++;
1221 inode_tally_features(inode, features);
1222 inode->i_visited = 1;
1226 /* Tally the features necessary to extract the specified dentries. */
1228 dentry_list_get_features(struct list_head *dentry_list,
1229 struct wim_features *features)
1231 struct wim_dentry *dentry;
1233 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1234 dentry_tally_features(dentry, features);
1236 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1237 dentry->d_inode->i_visited = 0;
1241 do_feature_check(const struct wim_features *required_features,
1242 const struct wim_features *supported_features,
1245 /* Encrypted files. */
1246 if (required_features->encrypted_files &&
1247 !supported_features->encrypted_files)
1248 WARNING("Ignoring EFS-encrypted data of %lu files",
1249 required_features->encrypted_files);
1251 /* Named data streams. */
1252 if (required_features->named_data_streams &&
1253 !supported_features->named_data_streams)
1254 WARNING("Ignoring named data streams of %lu files",
1255 required_features->named_data_streams);
1257 /* File attributes. */
1258 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1259 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1260 * We're an archive program, so theoretically we can do what we
1263 if (required_features->hidden_files &&
1264 !supported_features->hidden_files)
1265 WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1266 required_features->hidden_files);
1268 if (required_features->system_files &&
1269 !supported_features->system_files)
1270 WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1271 required_features->system_files);
1273 if (required_features->compressed_files &&
1274 !supported_features->compressed_files)
1275 WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1276 required_features->compressed_files);
1278 if (required_features->not_context_indexed_files &&
1279 !supported_features->not_context_indexed_files)
1280 WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1281 required_features->not_context_indexed_files);
1283 if (required_features->sparse_files &&
1284 !supported_features->sparse_files)
1285 WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1286 required_features->sparse_files);
1288 if (required_features->encrypted_directories &&
1289 !supported_features->encrypted_directories)
1290 WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1291 required_features->encrypted_directories);
1295 if (required_features->hard_links && !supported_features->hard_links)
1296 WARNING("Extracting %lu hard links as independent files",
1297 required_features->hard_links);
1299 /* Symbolic links and reparse points. */
1300 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1301 required_features->symlink_reparse_points &&
1302 !supported_features->symlink_reparse_points &&
1303 !supported_features->reparse_points)
1305 ERROR("Extraction backend does not support symbolic links!");
1306 return WIMLIB_ERR_UNSUPPORTED;
1308 if (required_features->reparse_points &&
1309 !supported_features->reparse_points)
1311 if (supported_features->symlink_reparse_points) {
1312 if (required_features->other_reparse_points) {
1313 WARNING("Ignoring reparse data of %lu non-symlink/junction files",
1314 required_features->other_reparse_points);
1317 WARNING("Ignoring reparse data of %lu files",
1318 required_features->reparse_points);
1322 /* Security descriptors. */
1323 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1324 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1325 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1326 required_features->security_descriptors &&
1327 !supported_features->security_descriptors)
1329 ERROR("Extraction backend does not support security descriptors!");
1330 return WIMLIB_ERR_UNSUPPORTED;
1332 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1333 required_features->security_descriptors &&
1334 !supported_features->security_descriptors)
1335 WARNING("Ignoring Windows NT security descriptors of %lu files",
1336 required_features->security_descriptors);
1339 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1340 required_features->unix_data && !supported_features->unix_data)
1342 ERROR("Extraction backend does not support UNIX data!");
1343 return WIMLIB_ERR_UNSUPPORTED;
1346 if (required_features->unix_data &&
1347 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1349 WARNING("Ignoring UNIX metadata of %lu files",
1350 required_features->unix_data);
1354 if (required_features->short_names &&
1355 !supported_features->short_names)
1357 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1358 ERROR("Extraction backend does not support DOS names!");
1359 return WIMLIB_ERR_UNSUPPORTED;
1361 WARNING("Ignoring DOS names of %lu files",
1362 required_features->short_names);
1366 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1367 !supported_features->timestamps)
1369 ERROR("Extraction backend does not support timestamps!");
1370 return WIMLIB_ERR_UNSUPPORTED;
1376 static const struct apply_operations *
1377 select_apply_operations(int extract_flags)
1380 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1381 return &ntfs_3g_apply_ops;
1384 return &win32_apply_ops;
1386 return &unix_apply_ops;
1391 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1392 const tchar *target, int extract_flags)
1394 const struct apply_operations *ops;
1395 struct apply_ctx *ctx;
1397 LIST_HEAD(dentry_list);
1399 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1400 ret = extract_dentries_to_stdout(trees, num_trees,
1405 num_trees = remove_duplicate_trees(trees, num_trees);
1406 num_trees = remove_contained_trees(trees, num_trees);
1408 ops = select_apply_operations(extract_flags);
1410 if (num_trees > 1 && ops->single_tree_only) {
1411 ERROR("Extracting multiple directory trees "
1412 "at once is not supported in %s extraction mode!",
1414 ret = WIMLIB_ERR_UNSUPPORTED;
1418 ctx = CALLOC(1, ops->context_size);
1420 ret = WIMLIB_ERR_NOMEM;
1425 ctx->target = target;
1426 ctx->target_nchars = tstrlen(target);
1427 ctx->extract_flags = extract_flags;
1428 if (ctx->wim->progfunc) {
1429 ctx->progfunc = ctx->wim->progfunc;
1430 ctx->progctx = ctx->wim->progctx;
1431 ctx->progress.extract.image = wim->current_image;
1432 ctx->progress.extract.extract_flags = (extract_flags &
1433 WIMLIB_EXTRACT_MASK_PUBLIC);
1434 ctx->progress.extract.wimfile_name = wim->filename;
1435 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1436 wim->current_image);
1437 ctx->progress.extract.target = target;
1439 INIT_LIST_HEAD(&ctx->blob_list);
1440 filedes_invalidate(&ctx->tmpfile_fd);
1441 ctx->apply_ops = ops;
1443 ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1447 build_dentry_list(&dentry_list, trees, num_trees,
1449 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1451 dentry_list_get_features(&dentry_list, &ctx->required_features);
1453 ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1454 ctx->extract_flags);
1458 ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1462 if (unlikely(list_empty(&dentry_list))) {
1463 WARNING("There is nothing to extract!");
1467 ret = dentry_list_resolve_streams(&dentry_list, ctx);
1471 dentry_list_build_inode_alias_lists(&dentry_list);
1473 ret = dentry_list_ref_streams(&dentry_list, ctx);
1477 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1478 /* When extracting from a pipe, the number of bytes of data to
1479 * extract can't be determined in the normal way (examining the
1480 * blob table), since at this point all we have is a set of
1481 * SHA-1 message digests of blobs that need to be extracted.
1482 * However, we can get a reasonably accurate estimate by taking
1483 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1484 * data. This does assume that a full image is being extracted,
1485 * but currently there is no API for doing otherwise. (Also,
1486 * subtract <HARDLINKBYTES> from this if hard links are
1487 * supported by the extraction mode.) */
1488 ctx->progress.extract.total_bytes =
1489 wim_info_get_image_total_bytes(wim->wim_info,
1490 wim->current_image);
1491 if (ctx->supported_features.hard_links) {
1492 ctx->progress.extract.total_bytes -=
1493 wim_info_get_image_hard_link_bytes(wim->wim_info,
1494 wim->current_image);
1498 ret = extract_progress(ctx,
1499 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1500 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1501 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1505 ret = (*ops->extract)(&dentry_list, ctx);
1509 if (ctx->progress.extract.completed_bytes <
1510 ctx->progress.extract.total_bytes)
1512 ctx->progress.extract.completed_bytes =
1513 ctx->progress.extract.total_bytes;
1514 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1519 ret = extract_progress(ctx,
1520 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1521 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1522 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1524 destroy_blob_list(&ctx->blob_list);
1525 destroy_dentry_list(&dentry_list);
1532 mkdir_if_needed(const tchar *target)
1534 if (!tmkdir(target, 0755))
1537 if (errno == EEXIST)
1541 /* _wmkdir() fails with EACCES if called on a drive root directory. */
1542 if (errno == EACCES)
1546 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
1547 return WIMLIB_ERR_MKDIR;
1550 /* Make sure the extraction flags make sense, and update them if needed. */
1552 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1554 int extract_flags = *extract_flags_p;
1556 /* Check for invalid flag combinations */
1558 if ((extract_flags &
1559 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1560 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1561 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1562 return WIMLIB_ERR_INVALID_PARAM;
1564 if ((extract_flags &
1565 (WIMLIB_EXTRACT_FLAG_RPFIX |
1566 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1567 WIMLIB_EXTRACT_FLAG_NORPFIX))
1568 return WIMLIB_ERR_INVALID_PARAM;
1570 #ifndef WITH_NTFS_3G
1571 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1572 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
1573 " it cannot apply a WIM image directly to an NTFS volume.");
1574 return WIMLIB_ERR_UNSUPPORTED;
1578 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1581 return WIMLIB_ERR_NO_FILENAME;
1583 ERROR("WIMBoot extraction is only supported on Windows!");
1584 return WIMLIB_ERR_UNSUPPORTED;
1589 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1590 WIMLIB_EXTRACT_FLAG_NORPFIX |
1591 WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1592 WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1594 /* For full-image extraction, do reparse point fixups by default
1595 * if the WIM header says they are enabled. */
1596 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1597 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1600 *extract_flags_p = extract_flags;
1605 get_wildcard_flags(int extract_flags)
1607 u32 wildcard_flags = 0;
1609 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB)
1610 wildcard_flags |= WILDCARD_FLAG_ERROR_IF_NO_MATCH;
1612 wildcard_flags |= WILDCARD_FLAG_WARN_IF_NO_MATCH;
1614 if (default_ignore_case)
1615 wildcard_flags |= WILDCARD_FLAG_CASE_INSENSITIVE;
1617 return wildcard_flags;
1620 struct append_dentry_ctx {
1621 struct wim_dentry **dentries;
1622 size_t num_dentries;
1623 size_t num_alloc_dentries;
1627 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1629 struct append_dentry_ctx *ctx = _ctx;
1631 if (ctx->num_dentries == ctx->num_alloc_dentries) {
1632 struct wim_dentry **new_dentries;
1635 new_length = max(ctx->num_alloc_dentries + 8,
1636 ctx->num_alloc_dentries * 3 / 2);
1637 new_dentries = REALLOC(ctx->dentries,
1638 new_length * sizeof(ctx->dentries[0]));
1639 if (new_dentries == NULL)
1640 return WIMLIB_ERR_NOMEM;
1641 ctx->dentries = new_dentries;
1642 ctx->num_alloc_dentries = new_length;
1644 ctx->dentries[ctx->num_dentries++] = dentry;
1649 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1650 const tchar * const *paths, size_t num_paths,
1654 struct wim_dentry **trees;
1657 if (wim == NULL || target == NULL || target[0] == T('\0') ||
1658 (num_paths != 0 && paths == NULL))
1659 return WIMLIB_ERR_INVALID_PARAM;
1661 ret = check_extract_flags(wim, &extract_flags);
1665 ret = select_wim_image(wim, image);
1669 ret = wim_checksum_unhashed_blobs(wim);
1673 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1674 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1675 (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1677 ret = mkdir_if_needed(target);
1682 if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1684 struct append_dentry_ctx append_dentry_ctx = {
1687 .num_alloc_dentries = 0,
1690 u32 wildcard_flags = get_wildcard_flags(extract_flags);
1692 for (size_t i = 0; i < num_paths; i++) {
1693 tchar *path = canonicalize_wim_path(paths[i]);
1695 ret = WIMLIB_ERR_NOMEM;
1696 trees = append_dentry_ctx.dentries;
1697 goto out_free_trees;
1699 ret = expand_wildcard(wim, path,
1705 trees = append_dentry_ctx.dentries;
1706 goto out_free_trees;
1709 trees = append_dentry_ctx.dentries;
1710 num_trees = append_dentry_ctx.num_dentries;
1712 trees = MALLOC(num_paths * sizeof(trees[0]));
1714 return WIMLIB_ERR_NOMEM;
1716 for (size_t i = 0; i < num_paths; i++) {
1718 tchar *path = canonicalize_wim_path(paths[i]);
1720 ret = WIMLIB_ERR_NOMEM;
1721 goto out_free_trees;
1724 trees[i] = get_dentry(wim, path,
1725 WIMLIB_CASE_PLATFORM_DEFAULT);
1727 if (trees[i] == NULL) {
1728 ERROR("Path \"%"TS"\" does not exist "
1730 paths[i], wim->current_image);
1731 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1732 goto out_free_trees;
1735 num_trees = num_paths;
1738 if (num_trees == 0) {
1740 goto out_free_trees;
1743 ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1750 extract_single_image(WIMStruct *wim, int image,
1751 const tchar *target, int extract_flags)
1753 const tchar *path = WIMLIB_WIM_ROOT_PATH;
1754 extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1755 return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1758 static const tchar * const filename_forbidden_chars =
1767 /* This function checks if it is okay to use a WIM image's name as a directory
1770 image_name_ok_as_dir(const tchar *image_name)
1772 return image_name && *image_name &&
1773 !tstrpbrk(image_name, filename_forbidden_chars) &&
1774 tstrcmp(image_name, T(".")) &&
1775 tstrcmp(image_name, T(".."));
1778 /* Extracts all images from the WIM to the directory @target, with the images
1779 * placed in subdirectories named by their image names. */
1781 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1783 size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
1784 size_t output_path_len = tstrlen(target);
1785 tchar buf[output_path_len + 1 + image_name_max_len + 1];
1788 const tchar *image_name;
1790 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1791 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1792 return WIMLIB_ERR_INVALID_PARAM;
1795 ret = mkdir_if_needed(target);
1798 tmemcpy(buf, target, output_path_len);
1799 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1800 for (image = 1; image <= wim->hdr.image_count; image++) {
1801 image_name = wimlib_get_image_name(wim, image);
1802 if (image_name_ok_as_dir(image_name)) {
1803 tstrcpy(buf + output_path_len + 1, image_name);
1805 /* Image name is empty or contains forbidden characters.
1806 * Use image number instead. */
1807 tsprintf(buf + output_path_len + 1, T("%d"), image);
1809 ret = extract_single_image(wim, image, buf, extract_flags);
1817 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1820 if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1821 WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1822 WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1823 return WIMLIB_ERR_INVALID_PARAM;
1825 if (image == WIMLIB_ALL_IMAGES)
1826 return extract_all_images(wim, target, extract_flags);
1828 return extract_single_image(wim, image, target, extract_flags);
1832 /****************************************************************************
1834 ****************************************************************************/
1837 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1838 const tchar * const *paths, size_t num_paths,
1841 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1842 return WIMLIB_ERR_INVALID_PARAM;
1844 return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1849 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1850 const tchar *path_list_file, int extract_flags)
1857 ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1859 ERROR("Failed to read path list file \"%"TS"\"",
1864 ret = wimlib_extract_paths(wim, image, target,
1865 (const tchar * const *)paths, num_paths,
1873 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1874 const tchar *image_num_or_name,
1875 const tchar *target,
1877 wimlib_progress_func_t progfunc,
1882 struct filedes *in_fd;
1886 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1887 return WIMLIB_ERR_INVALID_PARAM;
1889 /* Read the WIM header from the pipe and get a WIMStruct to represent
1890 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
1891 * wimlib_open_wim(), getting a WIMStruct in this way will result in an
1892 * empty blob table, no XML data read, and no filename set. */
1893 ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1898 /* Sanity check to make sure this is a pipable WIM. */
1899 if (pwm->hdr.magic != PWM_MAGIC) {
1900 ERROR("The WIM being read from file descriptor %d "
1901 "is not pipable!", pipe_fd);
1902 ret = WIMLIB_ERR_NOT_PIPABLE;
1903 goto out_wimlib_free;
1906 /* Sanity check to make sure the first part of a pipable split WIM is
1907 * sent over the pipe first. */
1908 if (pwm->hdr.part_number != 1) {
1909 ERROR("The first part of the split WIM must be "
1910 "sent over the pipe first.");
1911 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1912 goto out_wimlib_free;
1915 in_fd = &pwm->in_fd;
1916 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1918 /* As mentioned, the WIMStruct we created from the pipe does not have
1919 * XML data yet. Fix this by reading the extra copy of the XML data
1920 * that directly follows the header in pipable WIMs. (Note: see
1921 * write_pipable_wim() for more details about the format of pipable
1924 struct blob_descriptor xml_blob;
1925 struct wim_resource_descriptor xml_rdesc;
1926 ret = read_pwm_blob_header(pwm, &xml_blob, &xml_rdesc, 0, NULL);
1928 goto out_wimlib_free;
1930 if (!(xml_blob.flags & WIM_RESHDR_FLAG_METADATA))
1932 ERROR("Expected XML data, but found non-metadata resource.");
1933 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1934 goto out_wimlib_free;
1937 wim_res_desc_to_hdr(&xml_rdesc, &pwm->hdr.xml_data_reshdr);
1939 ret = read_wim_xml_data(pwm);
1941 goto out_wimlib_free;
1943 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
1944 ERROR("Image count in XML data is not the same as in WIM header.");
1945 ret = WIMLIB_ERR_IMAGE_COUNT;
1946 goto out_wimlib_free;
1950 /* Get image index (this may use the XML data that was just read to
1951 * resolve an image name). */
1952 if (image_num_or_name) {
1953 image = wimlib_resolve_image(pwm, image_num_or_name);
1954 if (image == WIMLIB_NO_IMAGE) {
1955 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
1957 ret = WIMLIB_ERR_INVALID_IMAGE;
1958 goto out_wimlib_free;
1959 } else if (image == WIMLIB_ALL_IMAGES) {
1960 ERROR("Applying all images from a pipe is not supported!");
1961 ret = WIMLIB_ERR_INVALID_IMAGE;
1962 goto out_wimlib_free;
1965 if (pwm->hdr.image_count != 1) {
1966 ERROR("No image was specified, but the pipable WIM "
1967 "did not contain exactly 1 image");
1968 ret = WIMLIB_ERR_INVALID_IMAGE;
1969 goto out_wimlib_free;
1974 /* Load the needed metadata resource. */
1975 for (i = 1; i <= pwm->hdr.image_count; i++) {
1976 struct blob_descriptor *metadata_blob;
1977 struct wim_image_metadata *imd;
1978 struct wim_resource_descriptor *metadata_rdesc;
1980 metadata_blob = new_blob_descriptor();
1981 if (metadata_blob == NULL) {
1982 ret = WIMLIB_ERR_NOMEM;
1983 goto out_wimlib_free;
1985 metadata_rdesc = MALLOC(sizeof(struct wim_resource_descriptor));
1986 if (metadata_rdesc == NULL) {
1987 ret = WIMLIB_ERR_NOMEM;
1988 free_blob_descriptor(metadata_blob);
1989 goto out_wimlib_free;
1992 ret = read_pwm_blob_header(pwm, metadata_blob, metadata_rdesc, 0, NULL);
1993 imd = pwm->image_metadata[i - 1];
1994 imd->metadata_blob = metadata_blob;
1996 FREE(metadata_rdesc);
1997 goto out_wimlib_free;
2000 if (!(metadata_blob->flags & WIM_RESHDR_FLAG_METADATA)) {
2001 ERROR("Expected metadata resource, but found "
2002 "non-metadata resource.");
2003 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2004 goto out_wimlib_free;
2008 /* Metadata resource is for the image being extracted.
2009 * Parse it and save the metadata in memory. */
2010 ret = read_metadata_resource(imd);
2012 goto out_wimlib_free;
2015 /* Metadata resource is not for the image being
2016 * extracted. Skip over it. */
2017 ret = skip_wim_resource(metadata_rdesc);
2019 goto out_wimlib_free;
2022 /* Extract the image. */
2023 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2024 ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
2025 /* Clean up and return. */
2033 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2034 const tchar *target, int extract_flags)
2036 return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
2045 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
2048 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
2049 return WIMLIB_ERR_INVALID_PARAM;
2050 return do_wimlib_extract_image(wim, image, target, extract_flags);