ea9c25228d8c8a4877a59223b6daa78e2777d749
[wimlib] / src / extract.c
1 /*
2  * extract.c
3  *
4  * Support for extracting WIM images, or files or directories contained in a WIM
5  * image.
6  */
7
8 /*
9  * Copyright (C) 2012, 2013, 2014, 2015 Eric Biggers
10  *
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
14  * later version.
15  *
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
19  * details.
20  *
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/.
23  */
24
25 /*
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().
30  *
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.
36  */
37
38 #ifdef HAVE_CONFIG_H
39 #  include "config.h"
40 #endif
41
42 #include <errno.h>
43 #include <fcntl.h>
44 #include <sys/stat.h>
45 #include <unistd.h>
46
47 #include "wimlib/apply.h"
48 #include "wimlib/assert.h"
49 #include "wimlib/blob_table.h"
50 #include "wimlib/dentry.h"
51 #include "wimlib/encoding.h"
52 #include "wimlib/endianness.h"
53 #include "wimlib/error.h"
54 #include "wimlib/metadata.h"
55 #include "wimlib/pathlist.h"
56 #include "wimlib/paths.h"
57 #include "wimlib/reparse.h"
58 #include "wimlib/resource.h"
59 #include "wimlib/security.h"
60 #include "wimlib/unix_data.h"
61 #include "wimlib/wildcard.h"
62 #include "wimlib/wim.h"
63 #include "wimlib/win32.h" /* for realpath() equivalent */
64 #include "wimlib/xml.h"
65
66 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE   0x80000000
67 #define WIMLIB_EXTRACT_FLAG_IMAGEMODE   0x40000000
68
69 /* Keep in sync with wimlib.h  */
70 #define WIMLIB_EXTRACT_MASK_PUBLIC                              \
71         (WIMLIB_EXTRACT_FLAG_NTFS                       |       \
72          WIMLIB_EXTRACT_FLAG_UNIX_DATA                  |       \
73          WIMLIB_EXTRACT_FLAG_NO_ACLS                    |       \
74          WIMLIB_EXTRACT_FLAG_STRICT_ACLS                |       \
75          WIMLIB_EXTRACT_FLAG_RPFIX                      |       \
76          WIMLIB_EXTRACT_FLAG_NORPFIX                    |       \
77          WIMLIB_EXTRACT_FLAG_TO_STDOUT                  |       \
78          WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES  |       \
79          WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS         |       \
80          WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS          |       \
81          WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES         |       \
82          WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS            |       \
83          WIMLIB_EXTRACT_FLAG_GLOB_PATHS                 |       \
84          WIMLIB_EXTRACT_FLAG_STRICT_GLOB                |       \
85          WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES              |       \
86          WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE  |       \
87          WIMLIB_EXTRACT_FLAG_WIMBOOT)
88
89 /* Send WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE or
90  * WIMLIB_PROGRESS_MSG_EXTRACT_METADATA.  */
91 int
92 do_file_extract_progress(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
93 {
94         ctx->count_until_file_progress = 500;  /* Arbitrary value to limit calls  */
95         return extract_progress(ctx, msg);
96 }
97
98 static int
99 start_file_phase(struct apply_ctx *ctx, u64 end_file_count, enum wimlib_progress_msg msg)
100 {
101         ctx->progress.extract.current_file_count = 0;
102         ctx->progress.extract.end_file_count = end_file_count;
103         return do_file_extract_progress(ctx, msg);
104 }
105
106 int
107 start_file_structure_phase(struct apply_ctx *ctx, u64 end_file_count)
108 {
109         return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
110 }
111
112 int
113 start_file_metadata_phase(struct apply_ctx *ctx, u64 end_file_count)
114 {
115         return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
116 }
117
118 static int
119 end_file_phase(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
120 {
121         ctx->progress.extract.current_file_count = ctx->progress.extract.end_file_count;
122         return do_file_extract_progress(ctx, msg);
123 }
124
125 int
126 end_file_structure_phase(struct apply_ctx *ctx)
127 {
128         return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
129 }
130
131 int
132 end_file_metadata_phase(struct apply_ctx *ctx)
133 {
134         return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
135 }
136
137 #define PWM_FOUND_WIM_HDR (-1)
138
139 /* Read the header for a blob in a pipable WIM.  If @pwm_hdr_ret is not NULL,
140  * also look for a pipable WIM header and return PWM_FOUND_WIM_HDR if found.  */
141 static int
142 read_pwm_blob_header(WIMStruct *pwm, u8 hash_ret[SHA1_HASH_SIZE],
143                      struct wim_reshdr *reshdr_ret,
144                      struct wim_header_disk *pwm_hdr_ret)
145 {
146         int ret;
147         struct pwm_blob_hdr blob_hdr;
148         u64 magic;
149
150         ret = full_read(&pwm->in_fd, &blob_hdr, sizeof(blob_hdr));
151         if (unlikely(ret))
152                 goto read_error;
153
154         magic = le64_to_cpu(blob_hdr.magic);
155
156         if (magic == PWM_MAGIC && pwm_hdr_ret != NULL) {
157                 memcpy(pwm_hdr_ret, &blob_hdr, sizeof(blob_hdr));
158                 ret = full_read(&pwm->in_fd,
159                                 (u8 *)pwm_hdr_ret + sizeof(blob_hdr),
160                                 sizeof(*pwm_hdr_ret) - sizeof(blob_hdr));
161                 if (unlikely(ret))
162                         goto read_error;
163                 return PWM_FOUND_WIM_HDR;
164         }
165
166         if (unlikely(magic != PWM_BLOB_MAGIC)) {
167                 ERROR("Data read on pipe is invalid (expected blob header)");
168                 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
169         }
170
171         copy_hash(hash_ret, blob_hdr.hash);
172
173         reshdr_ret->size_in_wim = 0; /* Not available  */
174         reshdr_ret->flags = le32_to_cpu(blob_hdr.flags);
175         reshdr_ret->offset_in_wim = pwm->in_fd.offset;
176         reshdr_ret->uncompressed_size = le64_to_cpu(blob_hdr.uncompressed_size);
177
178         if (unlikely(reshdr_ret->uncompressed_size == 0)) {
179                 ERROR("Data read on pipe is invalid (resource is of 0 size)");
180                 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
181         }
182
183         return 0;
184
185 read_error:
186         if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE)
187                 ERROR("The pipe ended before all needed data was sent!");
188         else
189                 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
190         return ret;
191 }
192
193 static int
194 read_blobs_from_pipe(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
195 {
196         int ret;
197         u8 hash[SHA1_HASH_SIZE];
198         struct wim_reshdr reshdr;
199         struct wim_header_disk pwm_hdr;
200         struct wim_resource_descriptor rdesc;
201         struct blob_descriptor *blob;
202
203         copy_guid(ctx->progress.extract.guid, ctx->wim->hdr.guid);
204         ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
205         ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
206         ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
207         if (ret)
208                 return ret;
209
210         while (ctx->num_blobs_remaining) {
211
212                 ret = read_pwm_blob_header(ctx->wim, hash, &reshdr, &pwm_hdr);
213
214                 if (ret == PWM_FOUND_WIM_HDR) {
215                         u16 part_number = le16_to_cpu(pwm_hdr.part_number);
216                         u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
217
218                         if (part_number == ctx->progress.extract.part_number &&
219                             total_parts == ctx->progress.extract.total_parts &&
220                             guids_equal(pwm_hdr.guid, ctx->progress.extract.guid))
221                                 continue;
222
223                         copy_guid(ctx->progress.extract.guid, pwm_hdr.guid);
224                         ctx->progress.extract.part_number = part_number;
225                         ctx->progress.extract.total_parts = total_parts;
226                         ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
227                         if (ret)
228                                 return ret;
229
230                         continue;
231                 }
232
233                 if (ret)
234                         return ret;
235
236                 if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)
237                     && (blob = lookup_blob(ctx->wim->blob_table, hash))
238                     && (blob->out_refcnt))
239                 {
240                         wim_reshdr_to_desc_and_blob(&reshdr, ctx->wim, &rdesc, blob);
241                         ret = read_blob_with_sha1(blob, cbs);
242                         blob_unset_is_located_in_wim_resource(blob);
243                         if (ret)
244                                 return ret;
245                         ctx->num_blobs_remaining--;
246                 } else {
247                         wim_reshdr_to_desc(&reshdr, ctx->wim, &rdesc);
248                         ret = skip_wim_resource(&rdesc);
249                         if (ret)
250                                 return ret;
251                 }
252         }
253
254         return 0;
255 }
256
257 /* Creates a temporary file opened for writing.  The open file descriptor is
258  * returned in @fd_ret and its name is returned in @name_ret (dynamically
259  * allocated).  */
260 static int
261 create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
262 {
263         tchar *name;
264         int open_flags;
265         int raw_fd;
266
267 retry:
268         name = ttempnam(NULL, T("wimlib"));
269         if (!name) {
270                 ERROR_WITH_ERRNO("Failed to create temporary filename");
271                 return WIMLIB_ERR_NOMEM;
272         }
273
274         open_flags = O_WRONLY | O_CREAT | O_EXCL | O_BINARY;
275 #ifdef __WIN32__
276         open_flags |= _O_SHORT_LIVED;
277 #endif
278         raw_fd = topen(name, open_flags, 0600);
279
280         if (raw_fd < 0) {
281                 if (errno == EEXIST) {
282                         FREE(name);
283                         goto retry;
284                 }
285                 ERROR_WITH_ERRNO("Failed to create temporary file "
286                                  "\"%"TS"\"", name);
287                 FREE(name);
288                 return WIMLIB_ERR_OPEN;
289         }
290
291         filedes_init(fd_ret, raw_fd);
292         *name_ret = name;
293         return 0;
294 }
295
296 static int
297 begin_extract_blob_wrapper(struct blob_descriptor *blob, void *_ctx)
298 {
299         struct apply_ctx *ctx = _ctx;
300
301         ctx->cur_blob = blob;
302         ctx->cur_blob_offset = 0;
303
304         if (unlikely(blob->out_refcnt > MAX_OPEN_FILES))
305                 return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
306
307         return call_begin_blob(blob, ctx->saved_cbs);
308 }
309
310 static int
311 extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx)
312 {
313         struct apply_ctx *ctx = _ctx;
314         union wimlib_progress_info *progress = &ctx->progress;
315         int ret;
316
317         ctx->cur_blob_offset += size;
318
319         if (likely(ctx->supported_features.hard_links)) {
320                 progress->extract.completed_bytes +=
321                         (u64)size * ctx->cur_blob->out_refcnt;
322                 if (ctx->cur_blob_offset == ctx->cur_blob->size)
323                         progress->extract.completed_streams += ctx->cur_blob->out_refcnt;
324         } else {
325                 const struct blob_extraction_target *targets =
326                         blob_extraction_targets(ctx->cur_blob);
327                 for (u32 i = 0; i < ctx->cur_blob->out_refcnt; i++) {
328                         const struct wim_inode *inode = targets[i].inode;
329                         const struct wim_dentry *dentry;
330
331                         inode_for_each_extraction_alias(dentry, inode) {
332                                 progress->extract.completed_bytes += size;
333                                 if (ctx->cur_blob_offset == ctx->cur_blob->size)
334                                         progress->extract.completed_streams++;
335                         }
336                 }
337         }
338         if (progress->extract.completed_bytes >= ctx->next_progress) {
339
340                 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
341                 if (ret)
342                         return ret;
343
344                 if (progress->extract.completed_bytes >=
345                     progress->extract.total_bytes)
346                 {
347                         ctx->next_progress = UINT64_MAX;
348                 } else {
349                         /* Send new message as soon as another 1/128 of the
350                          * total has been extracted.  (Arbitrary number.)  */
351                         ctx->next_progress =
352                                 progress->extract.completed_bytes +
353                                         progress->extract.total_bytes / 128;
354
355                         /* ... Unless that would be more than 5000000 bytes, in
356                          * which case send the next after the next 5000000
357                          * bytes.  (Another arbitrary number.)  */
358                         if (progress->extract.completed_bytes + 5000000 <
359                             ctx->next_progress)
360                                 ctx->next_progress =
361                                         progress->extract.completed_bytes + 5000000;
362
363                         /* ... But always send a message as soon as we're
364                          * completely done.  */
365                         if (progress->extract.total_bytes < ctx->next_progress)
366                                 ctx->next_progress = progress->extract.total_bytes;
367                 }
368         }
369
370         if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
371                 /* Just extracting to temporary file for now.  */
372                 ret = full_write(&ctx->tmpfile_fd, chunk, size);
373                 if (ret) {
374                         ERROR_WITH_ERRNO("Error writing data to "
375                                          "temporary file \"%"TS"\"",
376                                          ctx->tmpfile_name);
377                 }
378                 return ret;
379         }
380
381         return call_consume_chunk(chunk, size, ctx->saved_cbs);
382 }
383
384 /* Copy the blob's data from the temporary file to each of its targets.
385  *
386  * This is executed only in the very uncommon case that a blob is being
387  * extracted to more than MAX_OPEN_FILES targets!  */
388 static int
389 extract_from_tmpfile(const tchar *tmpfile_name,
390                      const struct blob_descriptor *orig_blob,
391                      const struct read_blob_callbacks *cbs)
392 {
393         struct blob_descriptor tmpfile_blob;
394         const struct blob_extraction_target *targets = blob_extraction_targets(orig_blob);
395         int ret;
396
397         memcpy(&tmpfile_blob, orig_blob, sizeof(struct blob_descriptor));
398         tmpfile_blob.blob_location = BLOB_IN_FILE_ON_DISK;
399         tmpfile_blob.file_on_disk = (tchar *)tmpfile_name;
400         tmpfile_blob.out_refcnt = 1;
401
402         for (u32 i = 0; i < orig_blob->out_refcnt; i++) {
403                 tmpfile_blob.inline_blob_extraction_targets[0] = targets[i];
404                 ret = read_blob_with_cbs(&tmpfile_blob, cbs);
405                 if (ret)
406                         return ret;
407         }
408         return 0;
409 }
410
411 static int
412 end_extract_blob_wrapper(struct blob_descriptor *blob, int status, void *_ctx)
413 {
414         struct apply_ctx *ctx = _ctx;
415
416         if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
417                 filedes_close(&ctx->tmpfile_fd);
418                 if (!status)
419                         status = extract_from_tmpfile(ctx->tmpfile_name, blob,
420                                                       ctx->saved_cbs);
421                 filedes_invalidate(&ctx->tmpfile_fd);
422                 tunlink(ctx->tmpfile_name);
423                 FREE(ctx->tmpfile_name);
424                 return status;
425         }
426
427         return call_end_blob(blob, status, ctx->saved_cbs);
428 }
429
430 /*
431  * Read the list of blobs to extract and feed their data into the specified
432  * callback functions.
433  *
434  * This handles checksumming each blob.
435  *
436  * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
437  *
438  * This also works if the WIM is being read from a pipe.
439  *
440  * This also will split up blobs that will need to be extracted to more than
441  * MAX_OPEN_FILES locations, as measured by the 'out_refcnt' of each blob.
442  * Therefore, the apply_operations implementation need not worry about running
443  * out of file descriptors, unless it might open more than one file descriptor
444  * per 'blob_extraction_target' (e.g. Win32 currently might because the
445  * destination file system might not support hard links).
446  */
447 int
448 extract_blob_list(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
449 {
450         struct read_blob_callbacks wrapper_cbs = {
451                 .begin_blob     = begin_extract_blob_wrapper,
452                 .consume_chunk  = extract_chunk_wrapper,
453                 .end_blob       = end_extract_blob_wrapper,
454                 .ctx            = ctx,
455         };
456         ctx->saved_cbs = cbs;
457         if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
458                 return read_blobs_from_pipe(ctx, &wrapper_cbs);
459         } else {
460                 return read_blob_list(&ctx->blob_list,
461                                       offsetof(struct blob_descriptor,
462                                                extraction_list),
463                                       &wrapper_cbs, VERIFY_BLOB_HASHES);
464         }
465 }
466
467 /* Extract a WIM dentry to standard output.
468  *
469  * This obviously doesn't make sense in all cases.  We return an error if the
470  * dentry does not correspond to a regular file.  Otherwise we extract the
471  * unnamed data stream only.  */
472 static int
473 extract_dentry_to_stdout(struct wim_dentry *dentry,
474                          const struct blob_table *blob_table)
475 {
476         struct wim_inode *inode = dentry->d_inode;
477         struct blob_descriptor *blob;
478         struct filedes _stdout;
479
480         if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
481                                    FILE_ATTRIBUTE_DIRECTORY |
482                                    FILE_ATTRIBUTE_ENCRYPTED))
483         {
484                 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
485                       "extracted to standard output", dentry_full_path(dentry));
486                 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
487         }
488
489         blob = inode_get_blob_for_unnamed_data_stream(inode, blob_table);
490         if (!blob) {
491                 const u8 *hash = inode_get_hash_of_unnamed_data_stream(inode);
492                 if (!is_zero_hash(hash))
493                         return blob_not_found_error(inode, hash);
494                 return 0;
495         }
496
497         filedes_init(&_stdout, STDOUT_FILENO);
498         return extract_blob_to_fd(blob, &_stdout);
499 }
500
501 static int
502 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
503                            const struct blob_table *blob_table)
504 {
505         for (size_t i = 0; i < num_dentries; i++) {
506                 int ret = extract_dentry_to_stdout(dentries[i], blob_table);
507                 if (ret)
508                         return ret;
509         }
510         return 0;
511 }
512
513 /**********************************************************************/
514
515 /*
516  * Removes duplicate dentries from the array.
517  *
518  * Returns the new number of dentries, packed at the front of the array.
519  */
520 static size_t
521 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
522 {
523         size_t i, j = 0;
524         for (i = 0; i < num_trees; i++) {
525                 if (!trees[i]->tmp_flag) {
526                         /* Found distinct dentry.  */
527                         trees[i]->tmp_flag = 1;
528                         trees[j++] = trees[i];
529                 }
530         }
531         for (i = 0; i < j; i++)
532                 trees[i]->tmp_flag = 0;
533         return j;
534 }
535
536 /*
537  * Remove dentries that are descendants of other dentries in the array.
538  *
539  * Returns the new number of dentries, packed at the front of the array.
540  */
541 static size_t
542 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
543 {
544         size_t i, j = 0;
545         for (i = 0; i < num_trees; i++)
546                 trees[i]->tmp_flag = 1;
547         for (i = 0; i < num_trees; i++) {
548                 struct wim_dentry *d = trees[i];
549                 while (!dentry_is_root(d)) {
550                         d = d->d_parent;
551                         if (d->tmp_flag)
552                                 goto tree_contained;
553                 }
554                 trees[j++] = trees[i];
555                 continue;
556
557         tree_contained:
558                 trees[i]->tmp_flag = 0;
559         }
560
561         for (i = 0; i < j; i++)
562                 trees[i]->tmp_flag = 0;
563         return j;
564 }
565
566 static int
567 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
568 {
569         struct list_head *dentry_list = _dentry_list;
570         list_add_tail(&dentry->d_extraction_list_node, dentry_list);
571         return 0;
572 }
573
574 static void
575 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
576 {
577         dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
578 }
579
580 static int
581 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
582 {
583         list_del(&dentry->d_extraction_list_node);
584         dentry_reset_extraction_list_node(dentry);
585         return 0;
586 }
587
588 /*
589  * Build the preliminary list of dentries to be extracted.
590  *
591  * The list maintains the invariant that if d1 and d2 are in the list and d1 is
592  * an ancestor of d2, then d1 appears before d2 in the list.
593  */
594 static void
595 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
596                   size_t num_trees, bool add_ancestors)
597 {
598         INIT_LIST_HEAD(dentry_list);
599
600         /* Add the trees recursively.  */
601         for (size_t i = 0; i < num_trees; i++)
602                 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
603
604         /* If requested, add ancestors of the trees.  */
605         if (add_ancestors) {
606                 for (size_t i = 0; i < num_trees; i++) {
607                         struct wim_dentry *dentry = trees[i];
608                         struct wim_dentry *ancestor;
609                         struct list_head *place_after;
610
611                         if (dentry_is_root(dentry))
612                                 continue;
613
614                         place_after = dentry_list;
615                         ancestor = dentry;
616                         do {
617                                 ancestor = ancestor->d_parent;
618                                 if (will_extract_dentry(ancestor)) {
619                                         place_after = &ancestor->d_extraction_list_node;
620                                         break;
621                                 }
622                         } while (!dentry_is_root(ancestor));
623
624                         ancestor = dentry;
625                         do {
626                                 ancestor = ancestor->d_parent;
627                                 if (will_extract_dentry(ancestor))
628                                         break;
629                                 list_add(&ancestor->d_extraction_list_node, place_after);
630                         } while (!dentry_is_root(ancestor));
631                 }
632         }
633 }
634
635 static void
636 destroy_dentry_list(struct list_head *dentry_list)
637 {
638         struct wim_dentry *dentry, *tmp;
639         struct wim_inode *inode;
640
641         list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
642                 inode = dentry->d_inode;
643                 dentry_reset_extraction_list_node(dentry);
644                 inode->i_visited = 0;
645                 inode->i_can_externally_back = 0;
646                 if ((void *)dentry->d_extraction_name != (void *)dentry->file_name)
647                         FREE(dentry->d_extraction_name);
648                 dentry->d_extraction_name = NULL;
649                 dentry->d_extraction_name_nchars = 0;
650         }
651 }
652
653 static void
654 destroy_blob_list(struct list_head *blob_list)
655 {
656         struct blob_descriptor *blob;
657
658         list_for_each_entry(blob, blob_list, extraction_list)
659                 if (blob->out_refcnt > ARRAY_LEN(blob->inline_blob_extraction_targets))
660                         FREE(blob->blob_extraction_targets);
661 }
662
663 #ifdef __WIN32__
664 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
665 #else
666 static const utf16lechar replacement_char = cpu_to_le16('?');
667 #endif
668
669 static bool
670 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
671 {
672         size_t i;
673
674         if (num_chars == 0)
675                 return true;
676         for (i = 0; i < num_chars; i++) {
677                 switch (name[i]) {
678         #ifdef __WIN32__
679                 case cpu_to_le16('\\'):
680                 case cpu_to_le16(':'):
681                 case cpu_to_le16('*'):
682                 case cpu_to_le16('?'):
683                 case cpu_to_le16('"'):
684                 case cpu_to_le16('<'):
685                 case cpu_to_le16('>'):
686                 case cpu_to_le16('|'):
687         #endif
688                 case cpu_to_le16('/'):
689                 case cpu_to_le16('\0'):
690                         if (fix)
691                                 name[i] = replacement_char;
692                         else
693                                 return false;
694                 }
695         }
696
697 #ifdef __WIN32__
698         if (name[num_chars - 1] == cpu_to_le16(' ') ||
699             name[num_chars - 1] == cpu_to_le16('.'))
700         {
701                 if (fix)
702                         name[num_chars - 1] = replacement_char;
703                 else
704                         return false;
705         }
706 #endif
707         return true;
708 }
709
710 static int
711 dentry_calculate_extraction_name(struct wim_dentry *dentry,
712                                  struct apply_ctx *ctx)
713 {
714         int ret;
715
716         if (dentry_is_root(dentry))
717                 return 0;
718
719 #ifdef WITH_NTFS_3G
720         if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
721                 dentry->d_extraction_name = dentry->file_name;
722                 dentry->d_extraction_name_nchars = dentry->file_name_nbytes /
723                                                    sizeof(utf16lechar);
724                 return 0;
725         }
726 #endif
727
728         if (!ctx->supported_features.case_sensitive_filenames) {
729                 struct wim_dentry *other;
730                 list_for_each_entry(other, &dentry->d_ci_conflict_list,
731                                     d_ci_conflict_list)
732                 {
733                         if (will_extract_dentry(other)) {
734                                 if (ctx->extract_flags &
735                                     WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
736                                         WARNING("\"%"TS"\" has the same "
737                                                 "case-insensitive name as "
738                                                 "\"%"TS"\"; extracting "
739                                                 "dummy name instead",
740                                                 dentry_full_path(dentry),
741                                                 dentry_full_path(other));
742                                         goto out_replace;
743                                 } else {
744                                         WARNING("Not extracting \"%"TS"\": "
745                                                 "has same case-insensitive "
746                                                 "name as \"%"TS"\"",
747                                                 dentry_full_path(dentry),
748                                                 dentry_full_path(other));
749                                         goto skip_dentry;
750                                 }
751                         }
752                 }
753         }
754
755         if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
756                 ret = utf16le_get_tstr(dentry->file_name,
757                                        dentry->file_name_nbytes,
758                                        (const tchar **)&dentry->d_extraction_name,
759                                        &dentry->d_extraction_name_nchars);
760                 dentry->d_extraction_name_nchars /= sizeof(tchar);
761                 return ret;
762         } else {
763                 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
764                 {
765                         WARNING("\"%"TS"\" has an invalid filename "
766                                 "that is not supported on this platform; "
767                                 "extracting dummy name instead",
768                                 dentry_full_path(dentry));
769                         goto out_replace;
770                 } else {
771                         WARNING("Not extracting \"%"TS"\": has an invalid filename "
772                                 "that is not supported on this platform",
773                                 dentry_full_path(dentry));
774                         goto skip_dentry;
775                 }
776         }
777
778 out_replace:
779         {
780                 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
781
782                 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
783                 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
784
785                 const tchar *tchar_name;
786                 size_t tchar_nchars;
787
788                 ret = utf16le_get_tstr(utf16_name_copy,
789                                        dentry->file_name_nbytes,
790                                        &tchar_name, &tchar_nchars);
791                 if (ret)
792                         return ret;
793
794                 tchar_nchars /= sizeof(tchar);
795
796                 size_t fixed_name_num_chars = tchar_nchars;
797                 tchar fixed_name[tchar_nchars + 50];
798
799                 tmemcpy(fixed_name, tchar_name, tchar_nchars);
800                 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
801                                                  T(" (invalid filename #%lu)"),
802                                                  ++ctx->invalid_sequence);
803
804                 utf16le_put_tstr(tchar_name);
805
806                 dentry->d_extraction_name = TSTRDUP(fixed_name);
807                 if (!dentry->d_extraction_name)
808                         return WIMLIB_ERR_NOMEM;
809                 dentry->d_extraction_name_nchars = fixed_name_num_chars;
810         }
811         return 0;
812
813 skip_dentry:
814         for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
815         return 0;
816 }
817
818 /*
819  * Calculate the actual filename component at which each WIM dentry will be
820  * extracted, with special handling for dentries that are unsupported by the
821  * extraction backend or have invalid names.
822  *
823  * ctx->supported_features must be filled in.
824  *
825  * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
826  */
827 static int
828 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
829                                        struct apply_ctx *ctx)
830 {
831         struct list_head *prev, *cur;
832
833         /* Can't use list_for_each_entry() because a call to
834          * dentry_calculate_extraction_name() may delete the current dentry and
835          * its children from the list.  */
836
837         prev = dentry_list;
838         for (;;) {
839                 struct wim_dentry *dentry;
840                 int ret;
841
842                 cur = prev->next;
843                 if (cur == dentry_list)
844                         break;
845
846                 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
847
848                 ret = dentry_calculate_extraction_name(dentry, ctx);
849                 if (ret)
850                         return ret;
851
852                 if (prev->next == cur)
853                         prev = cur;
854                 else
855                         ; /* Current dentry and its children (which follow in
856                              the list) were deleted.  prev stays the same.  */
857         }
858         return 0;
859 }
860
861 static int
862 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
863                        struct blob_table *blob_table)
864 {
865         struct wim_inode *inode = dentry->d_inode;
866         struct blob_descriptor *blob;
867         int ret;
868         bool force = false;
869
870         /* Special case:  when extracting from a pipe, the WIM blob table is
871          * initially empty, so "resolving" an inode's streams is initially not
872          * possible.  However, we still need to keep track of which blobs,
873          * identified by SHA-1 message digests, need to be extracted, so we
874          * "resolve" the inode's streams anyway by allocating a 'struct
875          * blob_descriptor' for each one.  */
876         if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
877                 force = true;
878         ret = inode_resolve_streams(inode, blob_table, force);
879         if (ret)
880                 return ret;
881         for (unsigned i = 0; i < inode->i_num_streams; i++) {
882                 blob = stream_blob_resolved(&inode->i_streams[i]);
883                 if (blob)
884                         blob->out_refcnt = 0;
885         }
886         return 0;
887 }
888
889 /*
890  * For each dentry to be extracted, resolve all streams in the corresponding
891  * inode and set 'out_refcnt' in all referenced blob_descriptors to 0.
892  *
893  * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
894  */
895 static int
896 dentry_list_resolve_streams(struct list_head *dentry_list,
897                             struct apply_ctx *ctx)
898 {
899         struct wim_dentry *dentry;
900         int ret;
901
902         list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
903                 ret = dentry_resolve_streams(dentry,
904                                              ctx->extract_flags,
905                                              ctx->wim->blob_table);
906                 if (ret)
907                         return ret;
908         }
909         return 0;
910 }
911
912 static int
913 ref_stream(struct wim_inode_stream *strm, struct wim_dentry *dentry,
914            struct apply_ctx *ctx)
915 {
916         struct wim_inode *inode = dentry->d_inode;
917         struct blob_descriptor *blob = stream_blob_resolved(strm);
918         struct blob_extraction_target *targets;
919
920         if (!blob)
921                 return 0;
922
923         /* Tally the size only for each actual extraction of the stream (not
924          * additional hard links to the inode).  */
925         if (inode->i_visited && ctx->supported_features.hard_links)
926                 return 0;
927
928         ctx->progress.extract.total_bytes += blob->size;
929         ctx->progress.extract.total_streams++;
930
931         if (inode->i_visited)
932                 return 0;
933
934         /* Add each blob to 'ctx->blob_list' only one time, regardless of how
935          * many extraction targets it will have.  */
936         if (blob->out_refcnt == 0) {
937                 list_add_tail(&blob->extraction_list, &ctx->blob_list);
938                 ctx->num_blobs_remaining++;
939         }
940
941         /* Set this stream as an extraction target of 'blob'.  */
942
943         if (blob->out_refcnt < ARRAY_LEN(blob->inline_blob_extraction_targets)) {
944                 targets = blob->inline_blob_extraction_targets;
945         } else {
946                 struct blob_extraction_target *prev_targets;
947                 size_t alloc_blob_extraction_targets;
948
949                 if (blob->out_refcnt == ARRAY_LEN(blob->inline_blob_extraction_targets)) {
950                         prev_targets = NULL;
951                         alloc_blob_extraction_targets = ARRAY_LEN(blob->inline_blob_extraction_targets);
952                 } else {
953                         prev_targets = blob->blob_extraction_targets;
954                         alloc_blob_extraction_targets = blob->alloc_blob_extraction_targets;
955                 }
956
957                 if (blob->out_refcnt == alloc_blob_extraction_targets) {
958                         alloc_blob_extraction_targets *= 2;
959                         targets = REALLOC(prev_targets,
960                                           alloc_blob_extraction_targets *
961                                           sizeof(targets[0]));
962                         if (!targets)
963                                 return WIMLIB_ERR_NOMEM;
964                         if (!prev_targets) {
965                                 memcpy(targets,
966                                        blob->inline_blob_extraction_targets,
967                                        sizeof(blob->inline_blob_extraction_targets));
968                         }
969                         blob->blob_extraction_targets = targets;
970                         blob->alloc_blob_extraction_targets = alloc_blob_extraction_targets;
971                 }
972                 targets = blob->blob_extraction_targets;
973         }
974         targets[blob->out_refcnt].inode = inode;
975         targets[blob->out_refcnt].stream = strm;
976         blob->out_refcnt++;
977         return 0;
978 }
979
980 static int
981 ref_stream_if_needed(struct wim_dentry *dentry, struct wim_inode *inode,
982                      struct wim_inode_stream *strm, struct apply_ctx *ctx)
983 {
984         bool need_stream = false;
985         switch (strm->stream_type) {
986         case STREAM_TYPE_DATA:
987                 if (stream_is_named(strm)) {
988                         /* Named data stream  */
989                         if (ctx->supported_features.named_data_streams)
990                                 need_stream = true;
991                 } else if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
992                                                     FILE_ATTRIBUTE_ENCRYPTED))
993                            && !(inode_is_symlink(inode)
994                                 && !ctx->supported_features.reparse_points
995                                 && ctx->supported_features.symlink_reparse_points))
996                 {
997                         /*
998                          * Unnamed data stream.  Skip if any of the following is true:
999                          *
1000                          * - file is a directory
1001                          * - file is encrypted
1002                          * - backend needs to create the file as UNIX symlink
1003                          * - backend will extract the stream as externally backed
1004                          */
1005                         if (ctx->apply_ops->will_externally_back) {
1006                                 int ret = (*ctx->apply_ops->will_externally_back)(dentry, ctx);
1007                                 if (ret > 0) /* Error?  */
1008                                         return ret;
1009                                 if (ret < 0) /* Won't externally back?  */
1010                                         need_stream = true;
1011                         } else {
1012                                 need_stream = true;
1013                         }
1014                 }
1015                 break;
1016         case STREAM_TYPE_REPARSE_POINT:
1017                 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT);
1018                 if (ctx->supported_features.reparse_points ||
1019                     (inode_is_symlink(inode) &&
1020                      ctx->supported_features.symlink_reparse_points))
1021                         need_stream = true;
1022                 break;
1023         case STREAM_TYPE_EFSRPC_RAW_DATA:
1024                 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED);
1025                 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
1026                         if (ctx->supported_features.encrypted_directories)
1027                                 need_stream = true;
1028                 } else {
1029                         if (ctx->supported_features.encrypted_files)
1030                                 need_stream = true;
1031                 }
1032                 break;
1033         }
1034         if (need_stream)
1035                 return ref_stream(strm, dentry, ctx);
1036         return 0;
1037 }
1038
1039 static int
1040 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1041 {
1042         struct wim_inode *inode = dentry->d_inode;
1043         for (unsigned i = 0; i < inode->i_num_streams; i++) {
1044                 int ret = ref_stream_if_needed(dentry, inode,
1045                                                &inode->i_streams[i], ctx);
1046                 if (ret)
1047                         return ret;
1048         }
1049         inode->i_visited = 1;
1050         return 0;
1051 }
1052
1053 /*
1054  * Given a list of dentries to be extracted, build the list of blobs that need
1055  * to be extracted, and for each blob determine the streams to which that blob
1056  * will be extracted.
1057  *
1058  * This also initializes the extract progress info with byte and blob
1059  * information.
1060  *
1061  * ctx->supported_features must be filled in.
1062  */
1063 static int
1064 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1065 {
1066         struct wim_dentry *dentry;
1067         int ret;
1068
1069         list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1070                 ret = dentry_ref_streams(dentry, ctx);
1071                 if (ret)
1072                         return ret;
1073         }
1074         list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1075                 dentry->d_inode->i_visited = 0;
1076         return 0;
1077 }
1078
1079 static void
1080 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1081 {
1082         struct wim_dentry *dentry;
1083
1084         list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1085                 dentry->d_inode->i_first_extraction_alias = NULL;
1086
1087         list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1088                 dentry->d_next_extraction_alias = dentry->d_inode->i_first_extraction_alias;
1089                 dentry->d_inode->i_first_extraction_alias = dentry;
1090         }
1091 }
1092
1093 static void
1094 inode_tally_features(const struct wim_inode *inode,
1095                      struct wim_features *features)
1096 {
1097         if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1098                 features->archive_files++;
1099         if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1100                 features->hidden_files++;
1101         if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1102                 features->system_files++;
1103         if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1104                 features->compressed_files++;
1105         if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1106                 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1107                         features->encrypted_directories++;
1108                 else
1109                         features->encrypted_files++;
1110         }
1111         if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1112                 features->not_context_indexed_files++;
1113         if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1114                 features->sparse_files++;
1115         if (inode_has_named_data_stream(inode))
1116                 features->named_data_streams++;
1117         if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1118                 features->reparse_points++;
1119                 if (inode_is_symlink(inode))
1120                         features->symlink_reparse_points++;
1121                 else
1122                         features->other_reparse_points++;
1123         }
1124         if (inode_has_security_descriptor(inode))
1125                 features->security_descriptors++;
1126         if (inode_has_unix_data(inode))
1127                 features->unix_data++;
1128 }
1129
1130 /* Tally features necessary to extract a dentry and the corresponding inode.  */
1131 static void
1132 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1133 {
1134         struct wim_inode *inode = dentry->d_inode;
1135
1136         if (dentry_has_short_name(dentry))
1137                 features->short_names++;
1138
1139         if (inode->i_visited) {
1140                 features->hard_links++;
1141         } else {
1142                 inode_tally_features(inode, features);
1143                 inode->i_visited = 1;
1144         }
1145 }
1146
1147 /* Tally the features necessary to extract the specified dentries.  */
1148 static void
1149 dentry_list_get_features(struct list_head *dentry_list,
1150                          struct wim_features *features)
1151 {
1152         struct wim_dentry *dentry;
1153
1154         list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1155                 dentry_tally_features(dentry, features);
1156
1157         list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1158                 dentry->d_inode->i_visited = 0;
1159 }
1160
1161 static int
1162 do_feature_check(const struct wim_features *required_features,
1163                  const struct wim_features *supported_features,
1164                  int extract_flags)
1165 {
1166         /* Encrypted files.  */
1167         if (required_features->encrypted_files &&
1168             !supported_features->encrypted_files)
1169                 WARNING("Ignoring EFS-encrypted data of %lu files",
1170                         required_features->encrypted_files);
1171
1172         /* Named data streams.  */
1173         if (required_features->named_data_streams &&
1174             !supported_features->named_data_streams)
1175                 WARNING("Ignoring named data streams of %lu files",
1176                         required_features->named_data_streams);
1177
1178         /* File attributes.  */
1179         if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1180                 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1181                  * We're an archive program, so theoretically we can do what we
1182                  * want with it.  */
1183
1184                 if (required_features->hidden_files &&
1185                     !supported_features->hidden_files)
1186                         WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1187                                 required_features->hidden_files);
1188
1189                 if (required_features->system_files &&
1190                     !supported_features->system_files)
1191                         WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1192                                 required_features->system_files);
1193
1194                 if (required_features->compressed_files &&
1195                     !supported_features->compressed_files)
1196                         WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1197                                 required_features->compressed_files);
1198
1199                 if (required_features->not_context_indexed_files &&
1200                     !supported_features->not_context_indexed_files)
1201                         WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1202                                 required_features->not_context_indexed_files);
1203
1204                 if (required_features->sparse_files &&
1205                     !supported_features->sparse_files)
1206                         WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1207                                 required_features->sparse_files);
1208
1209                 if (required_features->encrypted_directories &&
1210                     !supported_features->encrypted_directories)
1211                         WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1212                                 required_features->encrypted_directories);
1213         }
1214
1215         /* Hard links.  */
1216         if (required_features->hard_links && !supported_features->hard_links)
1217                 WARNING("Extracting %lu hard links as independent files",
1218                         required_features->hard_links);
1219
1220         /* Symbolic links and reparse points.  */
1221         if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1222             required_features->symlink_reparse_points &&
1223             !supported_features->symlink_reparse_points &&
1224             !supported_features->reparse_points)
1225         {
1226                 ERROR("Extraction backend does not support symbolic links!");
1227                 return WIMLIB_ERR_UNSUPPORTED;
1228         }
1229         if (required_features->reparse_points &&
1230             !supported_features->reparse_points)
1231         {
1232                 if (supported_features->symlink_reparse_points) {
1233                         if (required_features->other_reparse_points) {
1234                                 WARNING("Ignoring reparse data of %lu non-symlink/junction files",
1235                                         required_features->other_reparse_points);
1236                         }
1237                 } else {
1238                         WARNING("Ignoring reparse data of %lu files",
1239                                 required_features->reparse_points);
1240                 }
1241         }
1242
1243         /* Security descriptors.  */
1244         if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1245                                WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1246              == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1247             required_features->security_descriptors &&
1248             !supported_features->security_descriptors)
1249         {
1250                 ERROR("Extraction backend does not support security descriptors!");
1251                 return WIMLIB_ERR_UNSUPPORTED;
1252         }
1253         if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1254             required_features->security_descriptors &&
1255             !supported_features->security_descriptors)
1256                 WARNING("Ignoring Windows NT security descriptors of %lu files",
1257                         required_features->security_descriptors);
1258
1259         /* UNIX data.  */
1260         if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1261             required_features->unix_data && !supported_features->unix_data)
1262         {
1263                 ERROR("Extraction backend does not support UNIX data!");
1264                 return WIMLIB_ERR_UNSUPPORTED;
1265         }
1266
1267         if (required_features->unix_data &&
1268             !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1269         {
1270                 WARNING("Ignoring UNIX metadata of %lu files",
1271                         required_features->unix_data);
1272         }
1273
1274         /* DOS Names.  */
1275         if (required_features->short_names &&
1276             !supported_features->short_names)
1277         {
1278                 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1279                         ERROR("Extraction backend does not support DOS names!");
1280                         return WIMLIB_ERR_UNSUPPORTED;
1281                 }
1282                 WARNING("Ignoring DOS names of %lu files",
1283                         required_features->short_names);
1284         }
1285
1286         /* Timestamps.  */
1287         if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1288             !supported_features->timestamps)
1289         {
1290                 ERROR("Extraction backend does not support timestamps!");
1291                 return WIMLIB_ERR_UNSUPPORTED;
1292         }
1293
1294         return 0;
1295 }
1296
1297 static const struct apply_operations *
1298 select_apply_operations(int extract_flags)
1299 {
1300 #ifdef WITH_NTFS_3G
1301         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1302                 return &ntfs_3g_apply_ops;
1303 #endif
1304 #ifdef __WIN32__
1305         return &win32_apply_ops;
1306 #else
1307         return &unix_apply_ops;
1308 #endif
1309 }
1310
1311 static int
1312 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1313               const tchar *target, int extract_flags)
1314 {
1315         const struct apply_operations *ops;
1316         struct apply_ctx *ctx;
1317         int ret;
1318         LIST_HEAD(dentry_list);
1319
1320         if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1321                 ret = extract_dentries_to_stdout(trees, num_trees,
1322                                                  wim->blob_table);
1323                 goto out;
1324         }
1325
1326         num_trees = remove_duplicate_trees(trees, num_trees);
1327         num_trees = remove_contained_trees(trees, num_trees);
1328
1329         ops = select_apply_operations(extract_flags);
1330
1331         if (num_trees > 1 && ops->single_tree_only) {
1332                 ERROR("Extracting multiple directory trees "
1333                       "at once is not supported in %s extraction mode!",
1334                       ops->name);
1335                 ret = WIMLIB_ERR_UNSUPPORTED;
1336                 goto out;
1337         }
1338
1339         ctx = CALLOC(1, ops->context_size);
1340         if (!ctx) {
1341                 ret = WIMLIB_ERR_NOMEM;
1342                 goto out;
1343         }
1344
1345         ctx->wim = wim;
1346         ctx->target = target;
1347         ctx->target_nchars = tstrlen(target);
1348         ctx->extract_flags = extract_flags;
1349         if (ctx->wim->progfunc) {
1350                 ctx->progfunc = ctx->wim->progfunc;
1351                 ctx->progctx = ctx->wim->progctx;
1352                 ctx->progress.extract.image = wim->current_image;
1353                 ctx->progress.extract.extract_flags = (extract_flags &
1354                                                        WIMLIB_EXTRACT_MASK_PUBLIC);
1355                 ctx->progress.extract.wimfile_name = wim->filename;
1356                 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1357                                                                          wim->current_image);
1358                 ctx->progress.extract.target = target;
1359         }
1360         INIT_LIST_HEAD(&ctx->blob_list);
1361         filedes_invalidate(&ctx->tmpfile_fd);
1362         ctx->apply_ops = ops;
1363
1364         ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1365         if (ret)
1366                 goto out_cleanup;
1367
1368         build_dentry_list(&dentry_list, trees, num_trees,
1369                           !(extract_flags &
1370                             WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1371
1372         dentry_list_get_features(&dentry_list, &ctx->required_features);
1373
1374         ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1375                                ctx->extract_flags);
1376         if (ret)
1377                 goto out_cleanup;
1378
1379         ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1380         if (ret)
1381                 goto out_cleanup;
1382
1383         if (unlikely(list_empty(&dentry_list))) {
1384                 WARNING("There is nothing to extract!");
1385                 goto out_cleanup;
1386         }
1387
1388         ret = dentry_list_resolve_streams(&dentry_list, ctx);
1389         if (ret)
1390                 goto out_cleanup;
1391
1392         dentry_list_build_inode_alias_lists(&dentry_list);
1393
1394         ret = dentry_list_ref_streams(&dentry_list, ctx);
1395         if (ret)
1396                 goto out_cleanup;
1397
1398         if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1399                 /* When extracting from a pipe, the number of bytes of data to
1400                  * extract can't be determined in the normal way (examining the
1401                  * blob table), since at this point all we have is a set of
1402                  * SHA-1 message digests of blobs that need to be extracted.
1403                  * However, we can get a reasonably accurate estimate by taking
1404                  * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1405                  * data.  This does assume that a full image is being extracted,
1406                  * but currently there is no API for doing otherwise.  (Also,
1407                  * subtract <HARDLINKBYTES> from this if hard links are
1408                  * supported by the extraction mode.)  */
1409                 ctx->progress.extract.total_bytes =
1410                         wim_info_get_image_total_bytes(wim->wim_info,
1411                                                        wim->current_image);
1412                 if (ctx->supported_features.hard_links) {
1413                         ctx->progress.extract.total_bytes -=
1414                                 wim_info_get_image_hard_link_bytes(wim->wim_info,
1415                                                                    wim->current_image);
1416                 }
1417         }
1418
1419         ret = extract_progress(ctx,
1420                                ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1421                                        WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1422                                        WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1423         if (ret)
1424                 goto out_cleanup;
1425
1426         ret = (*ops->extract)(&dentry_list, ctx);
1427         if (ret)
1428                 goto out_cleanup;
1429
1430         if (ctx->progress.extract.completed_bytes <
1431             ctx->progress.extract.total_bytes)
1432         {
1433                 ctx->progress.extract.completed_bytes =
1434                         ctx->progress.extract.total_bytes;
1435                 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1436                 if (ret)
1437                         goto out_cleanup;
1438         }
1439
1440         ret = extract_progress(ctx,
1441                                ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1442                                        WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1443                                        WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1444 out_cleanup:
1445         destroy_blob_list(&ctx->blob_list);
1446         destroy_dentry_list(&dentry_list);
1447         FREE(ctx);
1448 out:
1449         return ret;
1450 }
1451
1452 static int
1453 mkdir_if_needed(const tchar *target)
1454 {
1455         if (!tmkdir(target, 0755))
1456                 return 0;
1457
1458         if (errno == EEXIST)
1459                 return 0;
1460
1461 #ifdef __WIN32__
1462         /* _wmkdir() fails with EACCES if called on a drive root directory.  */
1463         if (errno == EACCES)
1464                 return 0;
1465 #endif
1466
1467         ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
1468         return WIMLIB_ERR_MKDIR;
1469 }
1470
1471 /* Make sure the extraction flags make sense, and update them if needed.  */
1472 static int
1473 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1474 {
1475         int extract_flags = *extract_flags_p;
1476
1477         /* Check for invalid flag combinations  */
1478
1479         if ((extract_flags &
1480              (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1481               WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1482                                                     WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1483                 return WIMLIB_ERR_INVALID_PARAM;
1484
1485         if ((extract_flags &
1486              (WIMLIB_EXTRACT_FLAG_RPFIX |
1487               WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1488                                                 WIMLIB_EXTRACT_FLAG_NORPFIX))
1489                 return WIMLIB_ERR_INVALID_PARAM;
1490
1491 #ifndef WITH_NTFS_3G
1492         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1493                 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
1494                       "        it cannot apply a WIM image directly to an NTFS volume.");
1495                 return WIMLIB_ERR_UNSUPPORTED;
1496         }
1497 #endif
1498
1499         if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1500 #ifdef __WIN32__
1501                 if (!wim->filename)
1502                         return WIMLIB_ERR_NO_FILENAME;
1503 #else
1504                 ERROR("WIMBoot extraction is only supported on Windows!");
1505                 return WIMLIB_ERR_UNSUPPORTED;
1506 #endif
1507         }
1508
1509
1510         if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1511                               WIMLIB_EXTRACT_FLAG_NORPFIX |
1512                               WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1513                                         WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1514         {
1515                 /* For full-image extraction, do reparse point fixups by default
1516                  * if the WIM header says they are enabled.  */
1517                 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1518                         extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1519         }
1520
1521         *extract_flags_p = extract_flags;
1522         return 0;
1523 }
1524
1525 static u32
1526 get_wildcard_flags(int extract_flags)
1527 {
1528         u32 wildcard_flags = 0;
1529
1530         if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB)
1531                 wildcard_flags |= WILDCARD_FLAG_ERROR_IF_NO_MATCH;
1532         else
1533                 wildcard_flags |= WILDCARD_FLAG_WARN_IF_NO_MATCH;
1534
1535         if (default_ignore_case)
1536                 wildcard_flags |= WILDCARD_FLAG_CASE_INSENSITIVE;
1537
1538         return wildcard_flags;
1539 }
1540
1541 struct append_dentry_ctx {
1542         struct wim_dentry **dentries;
1543         size_t num_dentries;
1544         size_t num_alloc_dentries;
1545 };
1546
1547 static int
1548 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1549 {
1550         struct append_dentry_ctx *ctx = _ctx;
1551
1552         if (ctx->num_dentries == ctx->num_alloc_dentries) {
1553                 struct wim_dentry **new_dentries;
1554                 size_t new_length;
1555
1556                 new_length = max(ctx->num_alloc_dentries + 8,
1557                                  ctx->num_alloc_dentries * 3 / 2);
1558                 new_dentries = REALLOC(ctx->dentries,
1559                                        new_length * sizeof(ctx->dentries[0]));
1560                 if (new_dentries == NULL)
1561                         return WIMLIB_ERR_NOMEM;
1562                 ctx->dentries = new_dentries;
1563                 ctx->num_alloc_dentries = new_length;
1564         }
1565         ctx->dentries[ctx->num_dentries++] = dentry;
1566         return 0;
1567 }
1568
1569 static int
1570 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1571                         const tchar * const *paths, size_t num_paths,
1572                         int extract_flags)
1573 {
1574         int ret;
1575         struct wim_dentry **trees;
1576         size_t num_trees;
1577
1578         if (wim == NULL || target == NULL || target[0] == T('\0') ||
1579             (num_paths != 0 && paths == NULL))
1580                 return WIMLIB_ERR_INVALID_PARAM;
1581
1582         ret = check_extract_flags(wim, &extract_flags);
1583         if (ret)
1584                 return ret;
1585
1586         ret = select_wim_image(wim, image);
1587         if (ret)
1588                 return ret;
1589
1590         ret = wim_checksum_unhashed_blobs(wim);
1591         if (ret)
1592                 return ret;
1593
1594         if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1595                               WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1596             (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1597         {
1598                 ret = mkdir_if_needed(target);
1599                 if (ret)
1600                         return ret;
1601         }
1602
1603         if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1604
1605                 struct append_dentry_ctx append_dentry_ctx = {
1606                         .dentries = NULL,
1607                         .num_dentries = 0,
1608                         .num_alloc_dentries = 0,
1609                 };
1610
1611                 u32 wildcard_flags = get_wildcard_flags(extract_flags);
1612
1613                 for (size_t i = 0; i < num_paths; i++) {
1614                         tchar *path = canonicalize_wim_path(paths[i]);
1615                         if (path == NULL) {
1616                                 ret = WIMLIB_ERR_NOMEM;
1617                                 trees = append_dentry_ctx.dentries;
1618                                 goto out_free_trees;
1619                         }
1620                         ret = expand_wildcard(wim, path,
1621                                               append_dentry_cb,
1622                                               &append_dentry_ctx,
1623                                               wildcard_flags);
1624                         FREE(path);
1625                         if (ret) {
1626                                 trees = append_dentry_ctx.dentries;
1627                                 goto out_free_trees;
1628                         }
1629                 }
1630                 trees = append_dentry_ctx.dentries;
1631                 num_trees = append_dentry_ctx.num_dentries;
1632         } else {
1633                 trees = MALLOC(num_paths * sizeof(trees[0]));
1634                 if (trees == NULL)
1635                         return WIMLIB_ERR_NOMEM;
1636
1637                 for (size_t i = 0; i < num_paths; i++) {
1638
1639                         tchar *path = canonicalize_wim_path(paths[i]);
1640                         if (path == NULL) {
1641                                 ret = WIMLIB_ERR_NOMEM;
1642                                 goto out_free_trees;
1643                         }
1644
1645                         trees[i] = get_dentry(wim, path,
1646                                               WIMLIB_CASE_PLATFORM_DEFAULT);
1647                         FREE(path);
1648                         if (trees[i] == NULL) {
1649                                   ERROR("Path \"%"TS"\" does not exist "
1650                                         "in WIM image %d",
1651                                         paths[i], wim->current_image);
1652                                   ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1653                                   goto out_free_trees;
1654                         }
1655                 }
1656                 num_trees = num_paths;
1657         }
1658
1659         if (num_trees == 0) {
1660                 ret = 0;
1661                 goto out_free_trees;
1662         }
1663
1664         ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1665 out_free_trees:
1666         FREE(trees);
1667         return ret;
1668 }
1669
1670 static int
1671 extract_single_image(WIMStruct *wim, int image,
1672                      const tchar *target, int extract_flags)
1673 {
1674         const tchar *path = WIMLIB_WIM_ROOT_PATH;
1675         extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1676         return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1677 }
1678
1679 static const tchar * const filename_forbidden_chars =
1680 T(
1681 #ifdef __WIN32__
1682 "<>:\"/\\|?*"
1683 #else
1684 "/"
1685 #endif
1686 );
1687
1688 /* This function checks if it is okay to use a WIM image's name as a directory
1689  * name.  */
1690 static bool
1691 image_name_ok_as_dir(const tchar *image_name)
1692 {
1693         return image_name && *image_name &&
1694                 !tstrpbrk(image_name, filename_forbidden_chars) &&
1695                 tstrcmp(image_name, T(".")) &&
1696                 tstrcmp(image_name, T(".."));
1697 }
1698
1699 /* Extracts all images from the WIM to the directory @target, with the images
1700  * placed in subdirectories named by their image names. */
1701 static int
1702 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1703 {
1704         size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
1705         size_t output_path_len = tstrlen(target);
1706         tchar buf[output_path_len + 1 + image_name_max_len + 1];
1707         int ret;
1708         int image;
1709         const tchar *image_name;
1710
1711         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1712                 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1713                 return WIMLIB_ERR_INVALID_PARAM;
1714         }
1715
1716         ret = mkdir_if_needed(target);
1717         if (ret)
1718                 return ret;
1719         tmemcpy(buf, target, output_path_len);
1720         buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1721         for (image = 1; image <= wim->hdr.image_count; image++) {
1722                 image_name = wimlib_get_image_name(wim, image);
1723                 if (image_name_ok_as_dir(image_name)) {
1724                         tstrcpy(buf + output_path_len + 1, image_name);
1725                 } else {
1726                         /* Image name is empty or contains forbidden characters.
1727                          * Use image number instead. */
1728                         tsprintf(buf + output_path_len + 1, T("%d"), image);
1729                 }
1730                 ret = extract_single_image(wim, image, buf, extract_flags);
1731                 if (ret)
1732                         return ret;
1733         }
1734         return 0;
1735 }
1736
1737 static int
1738 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1739                         int extract_flags)
1740 {
1741         if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1742                              WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1743                              WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1744                 return WIMLIB_ERR_INVALID_PARAM;
1745
1746         if (image == WIMLIB_ALL_IMAGES)
1747                 return extract_all_images(wim, target, extract_flags);
1748         else
1749                 return extract_single_image(wim, image, target, extract_flags);
1750 }
1751
1752
1753 /****************************************************************************
1754  *                          Extraction API                                  *
1755  ****************************************************************************/
1756
1757 WIMLIBAPI int
1758 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1759                      const tchar * const *paths, size_t num_paths,
1760                      int extract_flags)
1761 {
1762         if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1763                 return WIMLIB_ERR_INVALID_PARAM;
1764
1765         return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1766                                        extract_flags);
1767 }
1768
1769 WIMLIBAPI int
1770 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1771                         const tchar *path_list_file, int extract_flags)
1772 {
1773         int ret;
1774         tchar **paths;
1775         size_t num_paths;
1776         void *mem;
1777
1778         ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1779         if (ret) {
1780                 ERROR("Failed to read path list file \"%"TS"\"",
1781                       path_list_file);
1782                 return ret;
1783         }
1784
1785         ret = wimlib_extract_paths(wim, image, target,
1786                                    (const tchar * const *)paths, num_paths,
1787                                    extract_flags);
1788         FREE(paths);
1789         FREE(mem);
1790         return ret;
1791 }
1792
1793 WIMLIBAPI int
1794 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1795                                              const tchar *image_num_or_name,
1796                                              const tchar *target,
1797                                              int extract_flags,
1798                                              wimlib_progress_func_t progfunc,
1799                                              void *progctx)
1800 {
1801         int ret;
1802         WIMStruct *pwm;
1803         struct filedes *in_fd;
1804         int image;
1805         unsigned i;
1806
1807         if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1808                 return WIMLIB_ERR_INVALID_PARAM;
1809
1810         /* Read the WIM header from the pipe and get a WIMStruct to represent
1811          * the pipable WIM.  Caveats:  Unlike getting a WIMStruct with
1812          * wimlib_open_wim(), getting a WIMStruct in this way will result in an
1813          * empty blob table, no XML data read, and no filename set.  */
1814         ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1815                                     progfunc, progctx);
1816         if (ret)
1817                 return ret;
1818
1819         /* Sanity check to make sure this is a pipable WIM.  */
1820         if (pwm->hdr.magic != PWM_MAGIC) {
1821                 ERROR("The WIM being read from file descriptor %d "
1822                       "is not pipable!", pipe_fd);
1823                 ret = WIMLIB_ERR_NOT_PIPABLE;
1824                 goto out_wimlib_free;
1825         }
1826
1827         /* Sanity check to make sure the first part of a pipable split WIM is
1828          * sent over the pipe first.  */
1829         if (pwm->hdr.part_number != 1) {
1830                 ERROR("The first part of the split WIM must be "
1831                       "sent over the pipe first.");
1832                 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1833                 goto out_wimlib_free;
1834         }
1835
1836         in_fd = &pwm->in_fd;
1837         wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1838
1839         /* As mentioned, the WIMStruct we created from the pipe does not have
1840          * XML data yet.  Fix this by reading the extra copy of the XML data
1841          * that directly follows the header in pipable WIMs.  (Note: see
1842          * write_pipable_wim() for more details about the format of pipable
1843          * WIMs.)  */
1844         {
1845                 u8 hash[SHA1_HASH_SIZE];
1846
1847                 ret = read_pwm_blob_header(pwm, hash,
1848                                            &pwm->hdr.xml_data_reshdr, NULL);
1849                 if (ret)
1850                         goto out_wimlib_free;
1851
1852                 if (!(pwm->hdr.xml_data_reshdr.flags & WIM_RESHDR_FLAG_METADATA)) {
1853                         ERROR("Expected XML data, but found non-metadata resource.");
1854                         ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1855                         goto out_wimlib_free;
1856                 }
1857
1858                 ret = read_wim_xml_data(pwm);
1859                 if (ret)
1860                         goto out_wimlib_free;
1861
1862                 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
1863                         ERROR("Image count in XML data is not the same as in WIM header.");
1864                         ret = WIMLIB_ERR_IMAGE_COUNT;
1865                         goto out_wimlib_free;
1866                 }
1867         }
1868
1869         /* Get image index (this may use the XML data that was just read to
1870          * resolve an image name).  */
1871         if (image_num_or_name) {
1872                 image = wimlib_resolve_image(pwm, image_num_or_name);
1873                 if (image == WIMLIB_NO_IMAGE) {
1874                         ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
1875                               image_num_or_name);
1876                         ret = WIMLIB_ERR_INVALID_IMAGE;
1877                         goto out_wimlib_free;
1878                 } else if (image == WIMLIB_ALL_IMAGES) {
1879                         ERROR("Applying all images from a pipe is not supported!");
1880                         ret = WIMLIB_ERR_INVALID_IMAGE;
1881                         goto out_wimlib_free;
1882                 }
1883         } else {
1884                 if (pwm->hdr.image_count != 1) {
1885                         ERROR("No image was specified, but the pipable WIM "
1886                               "did not contain exactly 1 image");
1887                         ret = WIMLIB_ERR_INVALID_IMAGE;
1888                         goto out_wimlib_free;
1889                 }
1890                 image = 1;
1891         }
1892
1893         /* Load the needed metadata resource.  */
1894         for (i = 1; i <= pwm->hdr.image_count; i++) {
1895                 struct wim_image_metadata *imd;
1896                 struct wim_reshdr reshdr;
1897                 struct wim_resource_descriptor *metadata_rdesc;
1898
1899                 imd = pwm->image_metadata[i - 1];
1900
1901                 ret = WIMLIB_ERR_NOMEM;
1902                 imd->metadata_blob = new_blob_descriptor();
1903                 if (!imd->metadata_blob)
1904                         goto out_wimlib_free;
1905
1906                 imd->metadata_blob->is_metadata = 1;
1907
1908                 ret = read_pwm_blob_header(pwm, imd->metadata_blob->hash,
1909                                            &reshdr, NULL);
1910                 if (ret)
1911                         goto out_wimlib_free;
1912
1913                 if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)) {
1914                         ERROR("Expected metadata resource, but found "
1915                               "non-metadata resource");
1916                         ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1917                         goto out_wimlib_free;
1918                 }
1919
1920                 ret = WIMLIB_ERR_NOMEM;
1921                 metadata_rdesc = MALLOC(sizeof(struct wim_resource_descriptor));
1922                 if (!metadata_rdesc)
1923                         goto out_wimlib_free;
1924                 wim_reshdr_to_desc_and_blob(&reshdr, pwm, metadata_rdesc,
1925                                             imd->metadata_blob);
1926
1927                 if (i == image) {
1928                         /* Metadata resource is for the image being extracted.
1929                          * Parse it and save the metadata in memory.  */
1930                         ret = read_metadata_resource(imd);
1931                         if (ret)
1932                                 goto out_wimlib_free;
1933                         imd->modified = 1;
1934                 } else {
1935                         /* Metadata resource is not for the image being
1936                          * extracted.  Skip over it.  */
1937                         ret = skip_wim_resource(metadata_rdesc);
1938                         if (ret)
1939                                 goto out_wimlib_free;
1940                 }
1941         }
1942         /* Extract the image.  */
1943         extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
1944         ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
1945         /* Clean up and return.  */
1946 out_wimlib_free:
1947         wimlib_free(pwm);
1948         return ret;
1949 }
1950
1951
1952 WIMLIBAPI int
1953 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
1954                                const tchar *target, int extract_flags)
1955 {
1956         return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
1957                                                             image_num_or_name,
1958                                                             target,
1959                                                             extract_flags,
1960                                                             NULL,
1961                                                             NULL);
1962 }
1963
1964 WIMLIBAPI int
1965 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1966                      int extract_flags)
1967 {
1968         if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1969                 return WIMLIB_ERR_INVALID_PARAM;
1970         return do_wimlib_extract_image(wim, image, target, extract_flags);
1971 }