4 * Support for extracting WIM images, or files or directories contained in a WIM
9 * Copyright (C) 2012-2016 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.
48 #include "wimlib/apply.h"
49 #include "wimlib/assert.h"
50 #include "wimlib/blob_table.h"
51 #include "wimlib/dentry.h"
52 #include "wimlib/encoding.h"
53 #include "wimlib/endianness.h"
54 #include "wimlib/error.h"
55 #include "wimlib/metadata.h"
56 #include "wimlib/object_id.h"
57 #include "wimlib/pathlist.h"
58 #include "wimlib/paths.h"
59 #include "wimlib/pattern.h"
60 #include "wimlib/reparse.h"
61 #include "wimlib/resource.h"
62 #include "wimlib/security.h"
63 #include "wimlib/unix_data.h"
64 #include "wimlib/wim.h"
65 #include "wimlib/win32.h" /* for realpath() equivalent */
66 #include "wimlib/xml.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 | \
90 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K | \
91 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K | \
92 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K | \
93 WIMLIB_EXTRACT_FLAG_COMPACT_LZX \
96 /* Send WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE or
97 * WIMLIB_PROGRESS_MSG_EXTRACT_METADATA. */
99 do_file_extract_progress(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
101 ctx->count_until_file_progress = 500; /* Arbitrary value to limit calls */
102 return extract_progress(ctx, msg);
106 start_file_phase(struct apply_ctx *ctx, u64 end_file_count, enum wimlib_progress_msg msg)
108 ctx->progress.extract.current_file_count = 0;
109 ctx->progress.extract.end_file_count = end_file_count;
110 return do_file_extract_progress(ctx, msg);
114 start_file_structure_phase(struct apply_ctx *ctx, u64 end_file_count)
116 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
120 start_file_metadata_phase(struct apply_ctx *ctx, u64 end_file_count)
122 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
126 end_file_phase(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
128 ctx->progress.extract.current_file_count = ctx->progress.extract.end_file_count;
129 return do_file_extract_progress(ctx, msg);
133 end_file_structure_phase(struct apply_ctx *ctx)
135 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
139 end_file_metadata_phase(struct apply_ctx *ctx)
141 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
144 /* Are all bytes in the specified buffer zero? */
146 is_all_zeroes(const u8 *p, const size_t size)
148 const u8 * const end = p + size;
150 for (; (uintptr_t)p % WORDBYTES && p != end; p++)
154 for (; end - p >= WORDBYTES; p += WORDBYTES)
155 if (*(const machine_word_t *)p)
158 for (; p != end; p++)
166 * Sparse regions should be detected at the granularity of the filesystem block
167 * size. For now just assume 4096 bytes, which is the default block size on
168 * NTFS and most Linux filesystems.
170 #define SPARSE_UNIT 4096
173 * Detect whether the specified buffer begins with a region of all zero bytes.
174 * Return %true if a zero region was found or %false if a nonzero region was
175 * found, and sets *len_ret to the length of the region. This operates at a
176 * granularity of SPARSE_UNIT bytes, meaning that to extend a zero region, there
177 * must be SPARSE_UNIT zero bytes with no interruption, but to extend a nonzero
178 * region, just one nonzero byte in the next SPARSE_UNIT bytes is sufficient.
180 * Note: besides compression, the WIM format doesn't yet have a way to
181 * efficiently represent zero regions, so that's why we need to detect them
182 * ourselves. Things will still fall apart badly on extremely large sparse
183 * files, but this is a start...
186 detect_sparse_region(const void *data, size_t size, size_t *len_ret)
188 const void *p = data;
189 const void * const end = data + size;
194 size_t n = min(end - p, SPARSE_UNIT);
195 bool z = is_all_zeroes(p, n);
197 if (len != 0 && z != zeroes)
208 #define PWM_FOUND_WIM_HDR (-1)
210 /* Read the header for a blob in a pipable WIM. If @pwm_hdr_ret is not NULL,
211 * also look for a pipable WIM header and return PWM_FOUND_WIM_HDR if found. */
213 read_pwm_blob_header(WIMStruct *pwm, u8 hash_ret[SHA1_HASH_SIZE],
214 struct wim_reshdr *reshdr_ret,
215 struct wim_header_disk *pwm_hdr_ret)
218 struct pwm_blob_hdr blob_hdr;
221 ret = full_read(&pwm->in_fd, &blob_hdr, sizeof(blob_hdr));
225 magic = le64_to_cpu(blob_hdr.magic);
227 if (magic == PWM_MAGIC && pwm_hdr_ret != NULL) {
228 memcpy(pwm_hdr_ret, &blob_hdr, sizeof(blob_hdr));
229 ret = full_read(&pwm->in_fd,
230 (u8 *)pwm_hdr_ret + sizeof(blob_hdr),
231 sizeof(*pwm_hdr_ret) - sizeof(blob_hdr));
234 return PWM_FOUND_WIM_HDR;
237 if (unlikely(magic != PWM_BLOB_MAGIC)) {
238 ERROR("Data read on pipe is invalid (expected blob header)");
239 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
242 copy_hash(hash_ret, blob_hdr.hash);
244 reshdr_ret->size_in_wim = 0; /* Not available */
245 reshdr_ret->flags = le32_to_cpu(blob_hdr.flags);
246 reshdr_ret->offset_in_wim = pwm->in_fd.offset;
247 reshdr_ret->uncompressed_size = le64_to_cpu(blob_hdr.uncompressed_size);
249 if (unlikely(reshdr_ret->uncompressed_size == 0)) {
250 ERROR("Data read on pipe is invalid (resource is of 0 size)");
251 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
257 if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE)
258 ERROR("The pipe ended before all needed data was sent!");
260 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
265 read_blobs_from_pipe(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
268 u8 hash[SHA1_HASH_SIZE];
269 struct wim_reshdr reshdr;
270 struct wim_header_disk pwm_hdr;
271 struct wim_resource_descriptor rdesc;
272 struct blob_descriptor *blob;
274 copy_guid(ctx->progress.extract.guid, ctx->wim->hdr.guid);
275 ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
276 ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
277 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
281 while (ctx->num_blobs_remaining) {
283 ret = read_pwm_blob_header(ctx->wim, hash, &reshdr, &pwm_hdr);
285 if (ret == PWM_FOUND_WIM_HDR) {
286 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
287 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
289 if (part_number == ctx->progress.extract.part_number &&
290 total_parts == ctx->progress.extract.total_parts &&
291 guids_equal(pwm_hdr.guid, ctx->progress.extract.guid))
294 copy_guid(ctx->progress.extract.guid, pwm_hdr.guid);
295 ctx->progress.extract.part_number = part_number;
296 ctx->progress.extract.total_parts = total_parts;
297 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
307 if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)
308 && (blob = lookup_blob(ctx->wim->blob_table, hash))
309 && (blob->out_refcnt))
311 wim_reshdr_to_desc_and_blob(&reshdr, ctx->wim, &rdesc, blob);
312 ret = read_blob_with_sha1(blob, cbs);
313 blob_unset_is_located_in_wim_resource(blob);
316 ctx->num_blobs_remaining--;
318 wim_reshdr_to_desc(&reshdr, ctx->wim, &rdesc);
319 ret = skip_wim_resource(&rdesc);
329 handle_pwm_metadata_resource(WIMStruct *pwm, int image, bool is_needed)
331 struct blob_descriptor *blob;
332 struct wim_reshdr reshdr;
333 struct wim_resource_descriptor *rdesc;
336 ret = WIMLIB_ERR_NOMEM;
337 blob = new_blob_descriptor();
341 ret = read_pwm_blob_header(pwm, blob->hash, &reshdr, NULL);
345 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
346 if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)) {
347 ERROR("Expected metadata resource, but found non-metadata "
352 ret = WIMLIB_ERR_NOMEM;
353 rdesc = MALLOC(sizeof(*rdesc));
357 wim_reshdr_to_desc_and_blob(&reshdr, pwm, rdesc, blob);
360 ret = WIMLIB_ERR_NOMEM;
361 pwm->image_metadata[image - 1] = new_unloaded_image_metadata(blob);
362 if (!pwm->image_metadata[image - 1])
366 /* If the metadata resource is for the image being extracted, then parse
367 * it and save the metadata in memory. Otherwise, skip over it. */
369 ret = select_wim_image(pwm, image);
371 ret = skip_wim_resource(rdesc);
373 free_blob_descriptor(blob);
377 /* Creates a temporary file opened for writing. The open file descriptor is
378 * returned in @fd_ret and its name is returned in @name_ret (dynamically
381 create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
388 name = _wtempnam(NULL, L"wimlib");
390 ERROR_WITH_ERRNO("Failed to create temporary filename");
391 return WIMLIB_ERR_NOMEM;
393 raw_fd = _wopen(name, O_WRONLY | O_CREAT | O_EXCL | O_BINARY |
394 _O_SHORT_LIVED, 0600);
395 if (raw_fd < 0 && errno == EEXIST) {
399 #else /* __WIN32__ */
400 const char *tmpdir = getenv("TMPDIR");
403 name = MALLOC(strlen(tmpdir) + 1 + 6 + 6 + 1);
405 return WIMLIB_ERR_NOMEM;
406 sprintf(name, "%s/wimlibXXXXXX", tmpdir);
407 raw_fd = mkstemp(name);
408 #endif /* !__WIN32__ */
411 ERROR_WITH_ERRNO("Failed to create temporary file "
414 return WIMLIB_ERR_OPEN;
417 filedes_init(fd_ret, raw_fd);
423 begin_extract_blob(struct blob_descriptor *blob, void *_ctx)
425 struct apply_ctx *ctx = _ctx;
427 if (unlikely(blob->out_refcnt > MAX_OPEN_FILES))
428 return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
430 return call_begin_blob(blob, ctx->saved_cbs);
434 extract_chunk(const struct blob_descriptor *blob, u64 offset,
435 const void *chunk, size_t size, void *_ctx)
437 struct apply_ctx *ctx = _ctx;
438 union wimlib_progress_info *progress = &ctx->progress;
439 bool last = (offset + size == blob->size);
442 if (likely(ctx->supported_features.hard_links)) {
443 progress->extract.completed_bytes +=
444 (u64)size * blob->out_refcnt;
446 progress->extract.completed_streams += blob->out_refcnt;
448 const struct blob_extraction_target *targets =
449 blob_extraction_targets(blob);
450 for (u32 i = 0; i < blob->out_refcnt; i++) {
451 const struct wim_inode *inode = targets[i].inode;
452 const struct wim_dentry *dentry;
454 inode_for_each_extraction_alias(dentry, inode) {
455 progress->extract.completed_bytes += size;
457 progress->extract.completed_streams++;
461 if (progress->extract.completed_bytes >= ctx->next_progress) {
463 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
467 set_next_progress(progress->extract.completed_bytes,
468 progress->extract.total_bytes,
469 &ctx->next_progress);
472 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
473 /* Just extracting to temporary file for now. */
474 ret = full_write(&ctx->tmpfile_fd, chunk, size);
476 ERROR_WITH_ERRNO("Error writing data to "
477 "temporary file \"%"TS"\"",
483 return call_continue_blob(blob, offset, chunk, size, ctx->saved_cbs);
486 /* Copy the blob's data from the temporary file to each of its targets.
488 * This is executed only in the very uncommon case that a blob is being
489 * extracted to more than MAX_OPEN_FILES targets! */
491 extract_from_tmpfile(const tchar *tmpfile_name,
492 const struct blob_descriptor *orig_blob,
493 const struct read_blob_callbacks *cbs)
495 struct blob_descriptor tmpfile_blob;
496 const struct blob_extraction_target *targets = blob_extraction_targets(orig_blob);
499 memcpy(&tmpfile_blob, orig_blob, sizeof(struct blob_descriptor));
500 tmpfile_blob.blob_location = BLOB_IN_FILE_ON_DISK;
501 tmpfile_blob.file_on_disk = (tchar *)tmpfile_name;
502 tmpfile_blob.out_refcnt = 1;
504 for (u32 i = 0; i < orig_blob->out_refcnt; i++) {
505 tmpfile_blob.inline_blob_extraction_targets[0] = targets[i];
506 ret = read_blob_with_cbs(&tmpfile_blob, cbs);
514 end_extract_blob(struct blob_descriptor *blob, int status, void *_ctx)
516 struct apply_ctx *ctx = _ctx;
518 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
519 filedes_close(&ctx->tmpfile_fd);
521 status = extract_from_tmpfile(ctx->tmpfile_name, blob,
523 filedes_invalidate(&ctx->tmpfile_fd);
524 tunlink(ctx->tmpfile_name);
525 FREE(ctx->tmpfile_name);
529 return call_end_blob(blob, status, ctx->saved_cbs);
533 * Read the list of blobs to extract and feed their data into the specified
534 * callback functions.
536 * This handles checksumming each blob.
538 * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
540 * This also works if the WIM is being read from a pipe.
542 * This also will split up blobs that will need to be extracted to more than
543 * MAX_OPEN_FILES locations, as measured by the 'out_refcnt' of each blob.
544 * Therefore, the apply_operations implementation need not worry about running
545 * out of file descriptors, unless it might open more than one file descriptor
546 * per 'blob_extraction_target' (e.g. Win32 currently might because the
547 * destination file system might not support hard links).
550 extract_blob_list(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
552 struct read_blob_callbacks wrapper_cbs = {
553 .begin_blob = begin_extract_blob,
554 .continue_blob = extract_chunk,
555 .end_blob = end_extract_blob,
558 ctx->saved_cbs = cbs;
559 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
560 return read_blobs_from_pipe(ctx, &wrapper_cbs);
562 return read_blob_list(&ctx->blob_list,
563 offsetof(struct blob_descriptor,
565 &wrapper_cbs, VERIFY_BLOB_HASHES);
569 /* Extract a WIM dentry to standard output.
571 * This obviously doesn't make sense in all cases. We return an error if the
572 * dentry does not correspond to a regular file. Otherwise we extract the
573 * unnamed data stream only. */
575 extract_dentry_to_stdout(struct wim_dentry *dentry,
576 const struct blob_table *blob_table)
578 struct wim_inode *inode = dentry->d_inode;
579 struct blob_descriptor *blob;
580 struct filedes _stdout;
582 if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
583 FILE_ATTRIBUTE_DIRECTORY |
584 FILE_ATTRIBUTE_ENCRYPTED))
586 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
587 "extracted to standard output", dentry_full_path(dentry));
588 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
591 blob = inode_get_blob_for_unnamed_data_stream(inode, blob_table);
593 const u8 *hash = inode_get_hash_of_unnamed_data_stream(inode);
594 if (!is_zero_hash(hash))
595 return blob_not_found_error(inode, hash);
599 filedes_init(&_stdout, STDOUT_FILENO);
600 return extract_blob_to_fd(blob, &_stdout);
604 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
605 const struct blob_table *blob_table)
607 for (size_t i = 0; i < num_dentries; i++) {
608 int ret = extract_dentry_to_stdout(dentries[i], blob_table);
615 /**********************************************************************/
618 * Removes duplicate dentries from the array.
620 * Returns the new number of dentries, packed at the front of the array.
623 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
626 for (i = 0; i < num_trees; i++) {
627 if (!trees[i]->d_tmp_flag) {
628 /* Found distinct dentry. */
629 trees[i]->d_tmp_flag = 1;
630 trees[j++] = trees[i];
633 for (i = 0; i < j; i++)
634 trees[i]->d_tmp_flag = 0;
639 * Remove dentries that are descendants of other dentries in the array.
641 * Returns the new number of dentries, packed at the front of the array.
644 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
647 for (i = 0; i < num_trees; i++)
648 trees[i]->d_tmp_flag = 1;
649 for (i = 0; i < num_trees; i++) {
650 struct wim_dentry *d = trees[i];
651 while (!dentry_is_root(d)) {
656 trees[j++] = trees[i];
660 trees[i]->d_tmp_flag = 0;
663 for (i = 0; i < j; i++)
664 trees[i]->d_tmp_flag = 0;
669 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
671 struct list_head *dentry_list = _dentry_list;
672 list_add_tail(&dentry->d_extraction_list_node, dentry_list);
677 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
679 dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
683 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
685 if (will_extract_dentry(dentry)) {
686 list_del(&dentry->d_extraction_list_node);
687 dentry_reset_extraction_list_node(dentry);
693 * Build the preliminary list of dentries to be extracted.
695 * The list maintains the invariant that if d1 and d2 are in the list and d1 is
696 * an ancestor of d2, then d1 appears before d2 in the list.
699 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
700 size_t num_trees, bool add_ancestors)
702 INIT_LIST_HEAD(dentry_list);
704 /* Add the trees recursively. */
705 for (size_t i = 0; i < num_trees; i++)
706 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
708 /* If requested, add ancestors of the trees. */
710 for (size_t i = 0; i < num_trees; i++) {
711 struct wim_dentry *dentry = trees[i];
712 struct wim_dentry *ancestor;
713 struct list_head *place_after;
715 if (dentry_is_root(dentry))
718 place_after = dentry_list;
721 ancestor = ancestor->d_parent;
722 if (will_extract_dentry(ancestor)) {
723 place_after = &ancestor->d_extraction_list_node;
726 } while (!dentry_is_root(ancestor));
730 ancestor = ancestor->d_parent;
731 if (will_extract_dentry(ancestor))
733 list_add(&ancestor->d_extraction_list_node, place_after);
734 } while (!dentry_is_root(ancestor));
740 destroy_dentry_list(struct list_head *dentry_list)
742 struct wim_dentry *dentry, *tmp;
743 struct wim_inode *inode;
745 list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
746 inode = dentry->d_inode;
747 dentry_reset_extraction_list_node(dentry);
748 inode->i_visited = 0;
749 inode->i_can_externally_back = 0;
750 if ((void *)dentry->d_extraction_name != (void *)dentry->d_name)
751 FREE(dentry->d_extraction_name);
752 dentry->d_extraction_name = NULL;
753 dentry->d_extraction_name_nchars = 0;
758 destroy_blob_list(struct list_head *blob_list)
760 struct blob_descriptor *blob;
762 list_for_each_entry(blob, blob_list, extraction_list)
763 if (blob->out_refcnt > ARRAY_LEN(blob->inline_blob_extraction_targets))
764 FREE(blob->blob_extraction_targets);
768 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
770 static const utf16lechar replacement_char = cpu_to_le16('?');
774 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
780 for (i = 0; i < num_chars; i++) {
781 switch (le16_to_cpu(name[i])) {
783 case '\x01'...'\x1F':
796 name[i] = replacement_char;
806 dentry_calculate_extraction_name(struct wim_dentry *dentry,
807 struct apply_ctx *ctx)
811 if (dentry_is_root(dentry))
815 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
816 dentry->d_extraction_name = dentry->d_name;
817 dentry->d_extraction_name_nchars = dentry->d_name_nbytes /
823 if (!ctx->supported_features.case_sensitive_filenames) {
824 struct wim_dentry *other;
825 dentry_for_each_ci_match(other, dentry) {
826 if (will_extract_dentry(other)) {
827 if (ctx->extract_flags &
828 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
829 WARNING("\"%"TS"\" has the same "
830 "case-insensitive name as "
831 "\"%"TS"\"; extracting "
832 "dummy name instead",
833 dentry_full_path(dentry),
834 dentry_full_path(other));
837 WARNING("Not extracting \"%"TS"\": "
838 "has same case-insensitive "
840 dentry_full_path(dentry),
841 dentry_full_path(other));
848 if (file_name_valid(dentry->d_name, dentry->d_name_nbytes / 2, false)) {
850 ret = utf16le_get_tstr(dentry->d_name,
851 dentry->d_name_nbytes,
852 (const tchar **)&dentry->d_extraction_name,
854 dentry->d_extraction_name_nchars = nbytes / sizeof(tchar);
857 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
859 WARNING("\"%"TS"\" has an invalid filename "
860 "that is not supported on this platform; "
861 "extracting dummy name instead",
862 dentry_full_path(dentry));
865 WARNING("Not extracting \"%"TS"\": has an invalid filename "
866 "that is not supported on this platform",
867 dentry_full_path(dentry));
874 utf16lechar utf16_name_copy[dentry->d_name_nbytes / 2];
876 memcpy(utf16_name_copy, dentry->d_name, dentry->d_name_nbytes);
877 file_name_valid(utf16_name_copy, dentry->d_name_nbytes / 2, true);
879 const tchar *tchar_name;
882 ret = utf16le_get_tstr(utf16_name_copy,
883 dentry->d_name_nbytes,
884 &tchar_name, &tchar_nchars);
888 tchar_nchars /= sizeof(tchar);
890 size_t fixed_name_num_chars = tchar_nchars;
891 tchar fixed_name[tchar_nchars + 50];
893 tmemcpy(fixed_name, tchar_name, tchar_nchars);
894 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
895 T(" (invalid filename #%lu)"),
896 ++ctx->invalid_sequence);
898 utf16le_put_tstr(tchar_name);
900 dentry->d_extraction_name = TSTRDUP(fixed_name);
901 if (!dentry->d_extraction_name)
902 return WIMLIB_ERR_NOMEM;
903 dentry->d_extraction_name_nchars = fixed_name_num_chars;
908 for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
913 * Calculate the actual filename component at which each WIM dentry will be
914 * extracted, with special handling for dentries that are unsupported by the
915 * extraction backend or have invalid names.
917 * ctx->supported_features must be filled in.
919 * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
922 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
923 struct apply_ctx *ctx)
925 struct list_head *prev, *cur;
927 /* Can't use list_for_each_entry() because a call to
928 * dentry_calculate_extraction_name() may delete the current dentry and
929 * its children from the list. */
933 struct wim_dentry *dentry;
937 if (cur == dentry_list)
940 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
942 ret = dentry_calculate_extraction_name(dentry, ctx);
946 if (prev->next == cur)
949 ; /* Current dentry and its children (which follow in
950 the list) were deleted. prev stays the same. */
956 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
957 struct blob_table *blob_table)
959 struct wim_inode *inode = dentry->d_inode;
960 struct blob_descriptor *blob;
964 /* Special case: when extracting from a pipe, the WIM blob table is
965 * initially empty, so "resolving" an inode's streams is initially not
966 * possible. However, we still need to keep track of which blobs,
967 * identified by SHA-1 message digests, need to be extracted, so we
968 * "resolve" the inode's streams anyway by allocating a 'struct
969 * blob_descriptor' for each one. */
970 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
972 ret = inode_resolve_streams(inode, blob_table, force);
975 for (unsigned i = 0; i < inode->i_num_streams; i++) {
976 blob = stream_blob_resolved(&inode->i_streams[i]);
978 blob->out_refcnt = 0;
984 * For each dentry to be extracted, resolve all streams in the corresponding
985 * inode and set 'out_refcnt' in all referenced blob_descriptors to 0.
987 * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
990 dentry_list_resolve_streams(struct list_head *dentry_list,
991 struct apply_ctx *ctx)
993 struct wim_dentry *dentry;
996 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
997 ret = dentry_resolve_streams(dentry,
999 ctx->wim->blob_table);
1007 ref_stream(struct wim_inode_stream *strm, struct wim_dentry *dentry,
1008 struct apply_ctx *ctx)
1010 struct wim_inode *inode = dentry->d_inode;
1011 struct blob_descriptor *blob = stream_blob_resolved(strm);
1012 struct blob_extraction_target *targets;
1017 /* Tally the size only for each actual extraction of the stream (not
1018 * additional hard links to the inode). */
1019 if (inode->i_visited && ctx->supported_features.hard_links)
1022 ctx->progress.extract.total_bytes += blob->size;
1023 ctx->progress.extract.total_streams++;
1025 if (inode->i_visited)
1028 /* Add each blob to 'ctx->blob_list' only one time, regardless of how
1029 * many extraction targets it will have. */
1030 if (blob->out_refcnt == 0) {
1031 list_add_tail(&blob->extraction_list, &ctx->blob_list);
1032 ctx->num_blobs_remaining++;
1035 /* Set this stream as an extraction target of 'blob'. */
1037 if (blob->out_refcnt < ARRAY_LEN(blob->inline_blob_extraction_targets)) {
1038 targets = blob->inline_blob_extraction_targets;
1040 struct blob_extraction_target *prev_targets;
1041 size_t alloc_blob_extraction_targets;
1043 if (blob->out_refcnt == ARRAY_LEN(blob->inline_blob_extraction_targets)) {
1044 prev_targets = NULL;
1045 alloc_blob_extraction_targets = ARRAY_LEN(blob->inline_blob_extraction_targets);
1047 prev_targets = blob->blob_extraction_targets;
1048 alloc_blob_extraction_targets = blob->alloc_blob_extraction_targets;
1051 if (blob->out_refcnt == alloc_blob_extraction_targets) {
1052 alloc_blob_extraction_targets *= 2;
1053 targets = REALLOC(prev_targets,
1054 alloc_blob_extraction_targets *
1055 sizeof(targets[0]));
1057 return WIMLIB_ERR_NOMEM;
1058 if (!prev_targets) {
1060 blob->inline_blob_extraction_targets,
1061 sizeof(blob->inline_blob_extraction_targets));
1063 blob->blob_extraction_targets = targets;
1064 blob->alloc_blob_extraction_targets = alloc_blob_extraction_targets;
1066 targets = blob->blob_extraction_targets;
1068 targets[blob->out_refcnt].inode = inode;
1069 targets[blob->out_refcnt].stream = strm;
1075 ref_stream_if_needed(struct wim_dentry *dentry, struct wim_inode *inode,
1076 struct wim_inode_stream *strm, struct apply_ctx *ctx)
1078 bool need_stream = false;
1079 switch (strm->stream_type) {
1080 case STREAM_TYPE_DATA:
1081 if (stream_is_named(strm)) {
1082 /* Named data stream */
1083 if (ctx->supported_features.named_data_streams)
1085 } else if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
1086 FILE_ATTRIBUTE_ENCRYPTED))
1087 && !(inode_is_symlink(inode)
1088 && !ctx->supported_features.reparse_points
1089 && ctx->supported_features.symlink_reparse_points))
1092 * Unnamed data stream. Skip if any of the following is true:
1094 * - file is a directory
1095 * - file is encrypted
1096 * - backend needs to create the file as UNIX symlink
1097 * - backend will extract the stream as externally
1098 * backed from the WIM archive itself
1100 if (ctx->apply_ops->will_back_from_wim) {
1101 int ret = (*ctx->apply_ops->will_back_from_wim)(dentry, ctx);
1102 if (ret > 0) /* Error? */
1104 if (ret < 0) /* Won't externally back? */
1111 case STREAM_TYPE_REPARSE_POINT:
1112 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT);
1113 if (ctx->supported_features.reparse_points ||
1114 (inode_is_symlink(inode) &&
1115 ctx->supported_features.symlink_reparse_points))
1118 case STREAM_TYPE_EFSRPC_RAW_DATA:
1119 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED);
1120 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
1121 if (ctx->supported_features.encrypted_directories)
1124 if (ctx->supported_features.encrypted_files)
1130 return ref_stream(strm, dentry, ctx);
1135 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1137 struct wim_inode *inode = dentry->d_inode;
1138 for (unsigned i = 0; i < inode->i_num_streams; i++) {
1139 int ret = ref_stream_if_needed(dentry, inode,
1140 &inode->i_streams[i], ctx);
1144 inode->i_visited = 1;
1149 * Given a list of dentries to be extracted, build the list of blobs that need
1150 * to be extracted, and for each blob determine the streams to which that blob
1151 * will be extracted.
1153 * This also initializes the extract progress info with byte and blob
1156 * ctx->supported_features must be filled in.
1159 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1161 struct wim_dentry *dentry;
1164 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1165 ret = dentry_ref_streams(dentry, ctx);
1169 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1170 dentry->d_inode->i_visited = 0;
1175 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1177 struct wim_dentry *dentry;
1179 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1180 dentry->d_inode->i_first_extraction_alias = NULL;
1182 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1183 dentry->d_next_extraction_alias = dentry->d_inode->i_first_extraction_alias;
1184 dentry->d_inode->i_first_extraction_alias = dentry;
1189 inode_tally_features(const struct wim_inode *inode,
1190 struct wim_features *features)
1192 if (inode->i_attributes & FILE_ATTRIBUTE_READONLY)
1193 features->readonly_files++;
1194 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1195 features->hidden_files++;
1196 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1197 features->system_files++;
1198 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1199 features->archive_files++;
1200 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1201 features->compressed_files++;
1202 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1203 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1204 features->encrypted_directories++;
1206 features->encrypted_files++;
1208 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1209 features->not_context_indexed_files++;
1210 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1211 features->sparse_files++;
1212 if (inode_has_named_data_stream(inode))
1213 features->named_data_streams++;
1214 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1215 features->reparse_points++;
1216 if (inode_is_symlink(inode))
1217 features->symlink_reparse_points++;
1219 features->other_reparse_points++;
1221 if (inode_has_security_descriptor(inode))
1222 features->security_descriptors++;
1223 if (inode_has_unix_data(inode))
1224 features->unix_data++;
1225 if (inode_has_object_id(inode))
1226 features->object_ids++;
1229 /* Tally features necessary to extract a dentry and the corresponding inode. */
1231 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1233 struct wim_inode *inode = dentry->d_inode;
1235 if (dentry_has_short_name(dentry))
1236 features->short_names++;
1238 if (inode->i_visited) {
1239 features->hard_links++;
1241 inode_tally_features(inode, features);
1242 inode->i_visited = 1;
1246 /* Tally the features necessary to extract the specified dentries. */
1248 dentry_list_get_features(struct list_head *dentry_list,
1249 struct wim_features *features)
1251 struct wim_dentry *dentry;
1253 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1254 dentry_tally_features(dentry, features);
1256 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1257 dentry->d_inode->i_visited = 0;
1261 do_feature_check(const struct wim_features *required_features,
1262 const struct wim_features *supported_features,
1265 /* Encrypted files. */
1266 if (required_features->encrypted_files &&
1267 !supported_features->encrypted_files)
1268 WARNING("Ignoring EFS-encrypted data of %lu files",
1269 required_features->encrypted_files);
1271 /* Named data streams. */
1272 if (required_features->named_data_streams &&
1273 !supported_features->named_data_streams)
1274 WARNING("Ignoring named data streams of %lu files",
1275 required_features->named_data_streams);
1277 /* File attributes. */
1278 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1280 if (required_features->readonly_files &&
1281 !supported_features->readonly_files)
1282 WARNING("Ignoring FILE_ATTRIBUTE_READONLY of %lu files",
1283 required_features->readonly_files);
1285 if (required_features->hidden_files &&
1286 !supported_features->hidden_files)
1287 WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1288 required_features->hidden_files);
1290 if (required_features->system_files &&
1291 !supported_features->system_files)
1292 WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1293 required_features->system_files);
1295 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1296 * We're an archive program, so theoretically we can do what we
1299 if (required_features->compressed_files &&
1300 !supported_features->compressed_files)
1301 WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1302 required_features->compressed_files);
1304 if (required_features->not_context_indexed_files &&
1305 !supported_features->not_context_indexed_files)
1306 WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1307 required_features->not_context_indexed_files);
1309 if (required_features->sparse_files &&
1310 !supported_features->sparse_files)
1311 WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1312 required_features->sparse_files);
1314 if (required_features->encrypted_directories &&
1315 !supported_features->encrypted_directories)
1316 WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1317 required_features->encrypted_directories);
1321 if (required_features->hard_links && !supported_features->hard_links)
1322 WARNING("Extracting %lu hard links as independent files",
1323 required_features->hard_links);
1325 /* Symbolic links and reparse points. */
1326 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1327 required_features->symlink_reparse_points &&
1328 !supported_features->symlink_reparse_points &&
1329 !supported_features->reparse_points)
1331 ERROR("Extraction backend does not support symbolic links!");
1332 return WIMLIB_ERR_UNSUPPORTED;
1334 if (required_features->reparse_points &&
1335 !supported_features->reparse_points)
1337 if (supported_features->symlink_reparse_points) {
1338 if (required_features->other_reparse_points) {
1339 WARNING("Ignoring reparse data of %lu non-symlink/junction files",
1340 required_features->other_reparse_points);
1343 WARNING("Ignoring reparse data of %lu files",
1344 required_features->reparse_points);
1348 /* Security descriptors. */
1349 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1350 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1351 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1352 required_features->security_descriptors &&
1353 !supported_features->security_descriptors)
1355 ERROR("Extraction backend does not support security descriptors!");
1356 return WIMLIB_ERR_UNSUPPORTED;
1358 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1359 required_features->security_descriptors &&
1360 !supported_features->security_descriptors)
1361 WARNING("Ignoring Windows NT security descriptors of %lu files",
1362 required_features->security_descriptors);
1365 if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1366 required_features->unix_data && !supported_features->unix_data)
1368 ERROR("Extraction backend does not support UNIX data!");
1369 return WIMLIB_ERR_UNSUPPORTED;
1372 if (required_features->unix_data &&
1373 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1375 WARNING("Ignoring UNIX metadata of %lu files",
1376 required_features->unix_data);
1380 if (required_features->object_ids && !supported_features->object_ids) {
1381 WARNING("Ignoring object IDs of %lu files",
1382 required_features->object_ids);
1386 if (required_features->short_names &&
1387 !supported_features->short_names)
1389 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1390 ERROR("Extraction backend does not support DOS names!");
1391 return WIMLIB_ERR_UNSUPPORTED;
1393 WARNING("Ignoring DOS names of %lu files",
1394 required_features->short_names);
1398 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1399 !supported_features->timestamps)
1401 ERROR("Extraction backend does not support timestamps!");
1402 return WIMLIB_ERR_UNSUPPORTED;
1408 static const struct apply_operations *
1409 select_apply_operations(int extract_flags)
1412 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1413 return &ntfs_3g_apply_ops;
1416 return &win32_apply_ops;
1418 return &unix_apply_ops;
1423 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1424 const tchar *target, int extract_flags)
1426 const struct apply_operations *ops;
1427 struct apply_ctx *ctx;
1429 LIST_HEAD(dentry_list);
1431 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1432 ret = extract_dentries_to_stdout(trees, num_trees,
1437 num_trees = remove_duplicate_trees(trees, num_trees);
1438 num_trees = remove_contained_trees(trees, num_trees);
1440 ops = select_apply_operations(extract_flags);
1442 if (num_trees > 1 && ops->single_tree_only) {
1443 ERROR("Extracting multiple directory trees "
1444 "at once is not supported in %s extraction mode!",
1446 ret = WIMLIB_ERR_UNSUPPORTED;
1450 ctx = CALLOC(1, ops->context_size);
1452 ret = WIMLIB_ERR_NOMEM;
1457 ctx->target = target;
1458 ctx->target_nchars = tstrlen(target);
1459 ctx->extract_flags = extract_flags;
1460 if (ctx->wim->progfunc) {
1461 ctx->progfunc = ctx->wim->progfunc;
1462 ctx->progctx = ctx->wim->progctx;
1463 ctx->progress.extract.image = wim->current_image;
1464 ctx->progress.extract.extract_flags = (extract_flags &
1465 WIMLIB_EXTRACT_MASK_PUBLIC);
1466 ctx->progress.extract.wimfile_name = wim->filename;
1467 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1468 wim->current_image);
1469 ctx->progress.extract.target = target;
1471 INIT_LIST_HEAD(&ctx->blob_list);
1472 filedes_invalidate(&ctx->tmpfile_fd);
1473 ctx->apply_ops = ops;
1475 ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1479 build_dentry_list(&dentry_list, trees, num_trees,
1481 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1483 dentry_list_get_features(&dentry_list, &ctx->required_features);
1485 ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1486 ctx->extract_flags);
1490 ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1494 if (unlikely(list_empty(&dentry_list))) {
1495 WARNING("There is nothing to extract!");
1499 ret = dentry_list_resolve_streams(&dentry_list, ctx);
1503 dentry_list_build_inode_alias_lists(&dentry_list);
1505 ret = dentry_list_ref_streams(&dentry_list, ctx);
1509 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1510 /* When extracting from a pipe, the number of bytes of data to
1511 * extract can't be determined in the normal way (examining the
1512 * blob table), since at this point all we have is a set of
1513 * SHA-1 message digests of blobs that need to be extracted.
1514 * However, we can get a reasonably accurate estimate by taking
1515 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1516 * data. This does assume that a full image is being extracted,
1517 * but currently there is no API for doing otherwise. (Also,
1518 * subtract <HARDLINKBYTES> from this if hard links are
1519 * supported by the extraction mode.) */
1520 ctx->progress.extract.total_bytes =
1521 xml_get_image_total_bytes(wim->xml_info,
1522 wim->current_image);
1523 if (ctx->supported_features.hard_links) {
1524 ctx->progress.extract.total_bytes -=
1525 xml_get_image_hard_link_bytes(wim->xml_info,
1526 wim->current_image);
1530 ret = extract_progress(ctx,
1531 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1532 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1533 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1537 ret = (*ops->extract)(&dentry_list, ctx);
1541 if (ctx->progress.extract.completed_bytes <
1542 ctx->progress.extract.total_bytes)
1544 ctx->progress.extract.completed_bytes =
1545 ctx->progress.extract.total_bytes;
1546 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1551 ret = extract_progress(ctx,
1552 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1553 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1554 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1556 destroy_blob_list(&ctx->blob_list);
1557 destroy_dentry_list(&dentry_list);
1564 mkdir_if_needed(const tchar *target)
1566 if (!tmkdir(target, 0755))
1569 if (errno == EEXIST)
1573 /* _wmkdir() fails with EACCES if called on a drive root directory. */
1574 if (errno == EACCES)
1578 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
1579 return WIMLIB_ERR_MKDIR;
1582 /* Make sure the extraction flags make sense, and update them if needed. */
1584 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1586 int extract_flags = *extract_flags_p;
1588 /* Check for invalid flag combinations */
1590 if ((extract_flags &
1591 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1592 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1593 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1594 return WIMLIB_ERR_INVALID_PARAM;
1596 if ((extract_flags &
1597 (WIMLIB_EXTRACT_FLAG_RPFIX |
1598 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1599 WIMLIB_EXTRACT_FLAG_NORPFIX))
1600 return WIMLIB_ERR_INVALID_PARAM;
1602 #ifndef WITH_NTFS_3G
1603 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1604 ERROR("wimlib was compiled without support for NTFS-3G, so\n"
1605 " it cannot apply a WIM image directly to an NTFS volume.");
1606 return WIMLIB_ERR_UNSUPPORTED;
1610 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1613 return WIMLIB_ERR_NO_FILENAME;
1615 ERROR("WIMBoot extraction is only supported on Windows!");
1616 return WIMLIB_ERR_UNSUPPORTED;
1620 if (extract_flags & (WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K |
1621 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K |
1622 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K |
1623 WIMLIB_EXTRACT_FLAG_COMPACT_LZX))
1627 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K) != 0);
1628 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K) != 0);
1629 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K) != 0);
1630 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_LZX) != 0);
1632 ERROR("Only one compression format can be specified "
1633 "for compact-mode extraction!");
1634 return WIMLIB_ERR_INVALID_PARAM;
1636 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1637 ERROR("Compact-mode extraction and WIMBoot-mode "
1638 "extraction are mutually exclusive!");
1639 return WIMLIB_ERR_INVALID_PARAM;
1642 ERROR("Compact-mode extraction (System Compression) "
1643 "is only supported on Windows!");
1644 return WIMLIB_ERR_UNSUPPORTED;
1649 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1650 WIMLIB_EXTRACT_FLAG_NORPFIX |
1651 WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1652 WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1654 /* For full-image extraction, do reparse point fixups by default
1655 * if the WIM header says they are enabled. */
1656 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1657 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1660 *extract_flags_p = extract_flags;
1664 struct append_dentry_ctx {
1665 struct wim_dentry **dentries;
1666 size_t num_dentries;
1667 size_t num_alloc_dentries;
1671 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1673 struct append_dentry_ctx *ctx = _ctx;
1675 if (ctx->num_dentries == ctx->num_alloc_dentries) {
1676 struct wim_dentry **new_dentries;
1679 new_length = max(ctx->num_alloc_dentries + 8,
1680 ctx->num_alloc_dentries * 3 / 2);
1681 new_dentries = REALLOC(ctx->dentries,
1682 new_length * sizeof(ctx->dentries[0]));
1683 if (new_dentries == NULL)
1684 return WIMLIB_ERR_NOMEM;
1685 ctx->dentries = new_dentries;
1686 ctx->num_alloc_dentries = new_length;
1688 ctx->dentries[ctx->num_dentries++] = dentry;
1692 /* Append dentries matched by a path which can contain wildcard characters. */
1694 append_matched_dentries(WIMStruct *wim, const tchar *orig_pattern,
1695 int extract_flags, struct append_dentry_ctx *ctx)
1697 const size_t count_before = ctx->num_dentries;
1701 pattern = canonicalize_wim_path(orig_pattern);
1703 return WIMLIB_ERR_NOMEM;
1704 ret = expand_path_pattern(wim_get_current_root_dentry(wim), pattern,
1705 append_dentry_cb, ctx);
1707 if (ret || ctx->num_dentries > count_before)
1709 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB) {
1710 ERROR("No matches for path pattern \"%"TS"\"", orig_pattern);
1711 return WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1713 WARNING("No matches for path pattern \"%"TS"\"", orig_pattern);
1718 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1719 const tchar * const *paths, size_t num_paths,
1723 struct wim_dentry **trees;
1726 if (wim == NULL || target == NULL || target[0] == T('\0') ||
1727 (num_paths != 0 && paths == NULL))
1728 return WIMLIB_ERR_INVALID_PARAM;
1730 ret = check_extract_flags(wim, &extract_flags);
1734 ret = select_wim_image(wim, image);
1738 ret = wim_checksum_unhashed_blobs(wim);
1742 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1743 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1744 (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1746 ret = mkdir_if_needed(target);
1751 if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1753 struct append_dentry_ctx append_dentry_ctx = {
1756 .num_alloc_dentries = 0,
1759 for (size_t i = 0; i < num_paths; i++) {
1760 ret = append_matched_dentries(wim, paths[i],
1762 &append_dentry_ctx);
1764 trees = append_dentry_ctx.dentries;
1765 goto out_free_trees;
1768 trees = append_dentry_ctx.dentries;
1769 num_trees = append_dentry_ctx.num_dentries;
1771 trees = MALLOC(num_paths * sizeof(trees[0]));
1773 return WIMLIB_ERR_NOMEM;
1775 for (size_t i = 0; i < num_paths; i++) {
1777 tchar *path = canonicalize_wim_path(paths[i]);
1779 ret = WIMLIB_ERR_NOMEM;
1780 goto out_free_trees;
1783 trees[i] = get_dentry(wim, path,
1784 WIMLIB_CASE_PLATFORM_DEFAULT);
1786 if (trees[i] == NULL) {
1787 ERROR("Path \"%"TS"\" does not exist "
1789 paths[i], wim->current_image);
1790 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1791 goto out_free_trees;
1794 num_trees = num_paths;
1797 if (num_trees == 0) {
1799 goto out_free_trees;
1802 ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1809 extract_single_image(WIMStruct *wim, int image,
1810 const tchar *target, int extract_flags)
1812 const tchar *path = WIMLIB_WIM_ROOT_PATH;
1813 extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1814 return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1817 static const tchar * const filename_forbidden_chars =
1824 /* This function checks if it is okay to use a WIM image's name as a directory
1827 image_name_ok_as_dir(const tchar *image_name)
1829 return image_name && *image_name &&
1830 !tstrpbrk(image_name, filename_forbidden_chars) &&
1831 tstrcmp(image_name, T(".")) &&
1832 tstrcmp(image_name, T("..")) &&
1833 tstrlen(image_name) <= 128;
1836 /* Extracts all images from the WIM to the directory @target, with the images
1837 * placed in subdirectories named by their image names. */
1839 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1841 size_t output_path_len = tstrlen(target);
1842 tchar buf[output_path_len + 1 + 128 + 1];
1845 const tchar *image_name;
1847 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1848 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1849 return WIMLIB_ERR_INVALID_PARAM;
1852 ret = mkdir_if_needed(target);
1855 tmemcpy(buf, target, output_path_len);
1856 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1857 for (image = 1; image <= wim->hdr.image_count; image++) {
1858 image_name = wimlib_get_image_name(wim, image);
1859 if (image_name_ok_as_dir(image_name)) {
1860 tstrcpy(buf + output_path_len + 1, image_name);
1862 /* Image name is empty or contains forbidden characters.
1863 * Use image number instead. */
1864 tsprintf(buf + output_path_len + 1, T("%d"), image);
1866 ret = extract_single_image(wim, image, buf, extract_flags);
1874 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1877 if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1878 WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1879 WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1880 return WIMLIB_ERR_INVALID_PARAM;
1882 if (image == WIMLIB_ALL_IMAGES)
1883 return extract_all_images(wim, target, extract_flags);
1885 return extract_single_image(wim, image, target, extract_flags);
1889 /****************************************************************************
1891 ****************************************************************************/
1894 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1895 const tchar * const *paths, size_t num_paths,
1898 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1899 return WIMLIB_ERR_INVALID_PARAM;
1901 return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1906 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1907 const tchar *path_list_file, int extract_flags)
1914 ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1916 ERROR("Failed to read path list file \"%"TS"\"",
1921 ret = wimlib_extract_paths(wim, image, target,
1922 (const tchar * const *)paths, num_paths,
1930 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1931 const tchar *image_num_or_name,
1932 const tchar *target,
1934 wimlib_progress_func_t progfunc,
1939 struct filedes *in_fd;
1943 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1944 return WIMLIB_ERR_INVALID_PARAM;
1946 /* Read the WIM header from the pipe and get a WIMStruct to represent
1947 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
1948 * wimlib_open_wim(), getting a WIMStruct in this way will result in an
1949 * empty blob table, no XML data read, and no filename set. */
1950 ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1955 /* Sanity check to make sure this is a pipable WIM. */
1956 if (pwm->hdr.magic != PWM_MAGIC) {
1957 ERROR("The WIM being read from file descriptor %d "
1958 "is not pipable!", pipe_fd);
1959 ret = WIMLIB_ERR_NOT_PIPABLE;
1960 goto out_wimlib_free;
1963 /* Sanity check to make sure the first part of a pipable split WIM is
1964 * sent over the pipe first. */
1965 if (pwm->hdr.part_number != 1) {
1966 ERROR("The first part of the split WIM must be "
1967 "sent over the pipe first.");
1968 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1969 goto out_wimlib_free;
1972 in_fd = &pwm->in_fd;
1973 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1975 /* As mentioned, the WIMStruct we created from the pipe does not have
1976 * XML data yet. Fix this by reading the extra copy of the XML data
1977 * that directly follows the header in pipable WIMs. (Note: see
1978 * write_pipable_wim() for more details about the format of pipable
1981 u8 hash[SHA1_HASH_SIZE];
1983 ret = read_pwm_blob_header(pwm, hash,
1984 &pwm->hdr.xml_data_reshdr, NULL);
1986 goto out_wimlib_free;
1988 if (!(pwm->hdr.xml_data_reshdr.flags & WIM_RESHDR_FLAG_METADATA)) {
1989 ERROR("Expected XML data, but found non-metadata resource.");
1990 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1991 goto out_wimlib_free;
1994 ret = read_wim_xml_data(pwm);
1996 goto out_wimlib_free;
1998 if (xml_get_image_count(pwm->xml_info) != pwm->hdr.image_count) {
1999 ERROR("Image count in XML data is not the same as in WIM header.");
2000 ret = WIMLIB_ERR_IMAGE_COUNT;
2001 goto out_wimlib_free;
2005 /* Get image index (this may use the XML data that was just read to
2006 * resolve an image name). */
2007 if (image_num_or_name) {
2008 image = wimlib_resolve_image(pwm, image_num_or_name);
2009 if (image == WIMLIB_NO_IMAGE) {
2010 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
2012 ret = WIMLIB_ERR_INVALID_IMAGE;
2013 goto out_wimlib_free;
2014 } else if (image == WIMLIB_ALL_IMAGES) {
2015 ERROR("Applying all images from a pipe is not supported!");
2016 ret = WIMLIB_ERR_INVALID_IMAGE;
2017 goto out_wimlib_free;
2020 if (pwm->hdr.image_count != 1) {
2021 ERROR("No image was specified, but the pipable WIM "
2022 "did not contain exactly 1 image");
2023 ret = WIMLIB_ERR_INVALID_IMAGE;
2024 goto out_wimlib_free;
2029 /* Load the needed metadata resource. */
2030 for (i = 1; i <= pwm->hdr.image_count; i++) {
2031 ret = handle_pwm_metadata_resource(pwm, i, i == image);
2033 goto out_wimlib_free;
2035 /* Extract the image. */
2036 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2037 ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
2038 /* Clean up and return. */
2046 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2047 const tchar *target, int extract_flags)
2049 return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
2058 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
2061 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
2062 return WIMLIB_ERR_INVALID_PARAM;
2063 return do_wimlib_extract_image(wim, image, target, extract_flags);