Update progress functions
[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 Eric Biggers
10  *
11  * This file is part of wimlib, a library for working with WIM files.
12  *
13  * wimlib is free software; you can redistribute it and/or modify it under the
14  * terms of the GNU General Public License as published by the Free
15  * Software Foundation; either version 3 of the License, or (at your option)
16  * any later version.
17  *
18  * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
19  * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
20  * A PARTICULAR PURPOSE. See the GNU General Public License for more
21  * details.
22  *
23  * You should have received a copy of the GNU General Public License
24  * along with wimlib; if not, see http://www.gnu.org/licenses/.
25  */
26
27 /*
28  * This file provides the API functions wimlib_extract_image(),
29  * wimlib_extract_image_from_pipe(), wimlib_extract_paths(), and
30  * wimlib_extract_pathlist().  Internally, all end up calling
31  * do_wimlib_extract_paths() and extract_trees().
32  *
33  * Although wimlib supports multiple extraction modes/backends (NTFS-3g, UNIX,
34  * Win32), this file does not itself have code to extract files or directories
35  * to any specific target; instead, it handles generic functionality and relies
36  * on lower-level callback functions declared in `struct apply_operations' to do
37  * the actual extraction.
38  */
39
40 #ifdef HAVE_CONFIG_H
41 #  include "config.h"
42 #endif
43
44 #include "wimlib/apply.h"
45 #include "wimlib/dentry.h"
46 #include "wimlib/encoding.h"
47 #include "wimlib/endianness.h"
48 #include "wimlib/error.h"
49 #include "wimlib/lookup_table.h"
50 #include "wimlib/metadata.h"
51 #include "wimlib/pathlist.h"
52 #include "wimlib/paths.h"
53 #include "wimlib/reparse.h"
54 #include "wimlib/resource.h"
55 #include "wimlib/security.h"
56 #ifdef __WIN32__
57 #  include "wimlib/win32.h" /* for realpath() equivalent */
58 #endif
59 #include "wimlib/xml.h"
60 #include "wimlib/wildcard.h"
61 #include "wimlib/wim.h"
62
63 #include <errno.h>
64 #include <fcntl.h>
65 #include <stdlib.h>
66 #include <sys/stat.h>
67 #include <unistd.h>
68
69 #define WIMLIB_EXTRACT_FLAG_MULTI_IMAGE 0x80000000
70 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE   0x40000000
71 #define WIMLIB_EXTRACT_FLAG_IMAGEMODE   0x20000000
72
73 /* Keep in sync with wimlib.h  */
74 #define WIMLIB_EXTRACT_MASK_PUBLIC                              \
75         (WIMLIB_EXTRACT_FLAG_NTFS                       |       \
76          WIMLIB_EXTRACT_FLAG_UNIX_DATA                  |       \
77          WIMLIB_EXTRACT_FLAG_NO_ACLS                    |       \
78          WIMLIB_EXTRACT_FLAG_STRICT_ACLS                |       \
79          WIMLIB_EXTRACT_FLAG_RPFIX                      |       \
80          WIMLIB_EXTRACT_FLAG_NORPFIX                    |       \
81          WIMLIB_EXTRACT_FLAG_TO_STDOUT                  |       \
82          WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES  |       \
83          WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS         |       \
84          WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS          |       \
85          WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES         |       \
86          WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS            |       \
87          WIMLIB_EXTRACT_FLAG_GLOB_PATHS                 |       \
88          WIMLIB_EXTRACT_FLAG_STRICT_GLOB                |       \
89          WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES              |       \
90          WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE  |       \
91          WIMLIB_EXTRACT_FLAG_WIMBOOT)
92
93 /* Check whether the extraction of a dentry should be skipped completely.  */
94 static bool
95 dentry_is_supported(struct wim_dentry *dentry,
96                     const struct wim_features *supported_features)
97 {
98         struct wim_inode *inode = dentry->d_inode;
99
100         if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
101                 return supported_features->reparse_points ||
102                         (inode_is_symlink(inode) &&
103                          supported_features->symlink_reparse_points);
104         }
105         if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
106                 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
107                         return supported_features->encrypted_directories != 0;
108                 else
109                         return supported_features->encrypted_files != 0;
110         }
111         return true;
112 }
113
114
115 #define PWM_ALLOW_WIM_HDR 0x00001
116 #define PWM_SILENT_EOF    0x00002
117
118 /* Read the header from a stream in a pipable WIM.  */
119 static int
120 read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte,
121                        struct wim_resource_spec *rspec,
122                        int flags, struct wim_header_disk *hdr_ret)
123 {
124         union {
125                 struct pwm_stream_hdr stream_hdr;
126                 struct wim_header_disk pwm_hdr;
127         } buf;
128         struct wim_reshdr reshdr;
129         int ret;
130
131         ret = full_read(&pwm->in_fd, &buf.stream_hdr, sizeof(buf.stream_hdr));
132         if (ret)
133                 goto read_error;
134
135         if ((flags & PWM_ALLOW_WIM_HDR) && buf.stream_hdr.magic == PWM_MAGIC) {
136                 BUILD_BUG_ON(sizeof(buf.pwm_hdr) < sizeof(buf.stream_hdr));
137                 ret = full_read(&pwm->in_fd, &buf.stream_hdr + 1,
138                                 sizeof(buf.pwm_hdr) - sizeof(buf.stream_hdr));
139
140                 if (ret)
141                         goto read_error;
142                 lte->resource_location = RESOURCE_NONEXISTENT;
143                 memcpy(hdr_ret, &buf.pwm_hdr, sizeof(buf.pwm_hdr));
144                 return 0;
145         }
146
147         if (le64_to_cpu(buf.stream_hdr.magic) != PWM_STREAM_MAGIC) {
148                 ERROR("Data read on pipe is invalid (expected stream header).");
149                 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
150         }
151
152         copy_hash(lte->hash, buf.stream_hdr.hash);
153
154         reshdr.size_in_wim = 0;
155         reshdr.flags = le32_to_cpu(buf.stream_hdr.flags);
156         reshdr.offset_in_wim = pwm->in_fd.offset;
157         reshdr.uncompressed_size = le64_to_cpu(buf.stream_hdr.uncompressed_size);
158         wim_res_hdr_to_spec(&reshdr, pwm, rspec);
159         lte_bind_wim_resource_spec(lte, rspec);
160         lte->flags = rspec->flags;
161         lte->size = rspec->uncompressed_size;
162         lte->offset_in_res = 0;
163         return 0;
164
165 read_error:
166         if (ret != WIMLIB_ERR_UNEXPECTED_END_OF_FILE || !(flags & PWM_SILENT_EOF))
167                 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
168         return ret;
169 }
170
171 static int
172 load_streams_from_pipe(struct apply_ctx *ctx,
173                        const struct read_stream_list_callbacks *cbs)
174 {
175         struct wim_lookup_table_entry *found_lte = NULL;
176         struct wim_resource_spec *rspec = NULL;
177         struct wim_lookup_table *lookup_table;
178         int ret;
179
180         ret = WIMLIB_ERR_NOMEM;
181         found_lte = new_lookup_table_entry();
182         if (!found_lte)
183                 goto out;
184
185         rspec = MALLOC(sizeof(struct wim_resource_spec));
186         if (!rspec)
187                 goto out;
188
189         lookup_table = ctx->wim->lookup_table;
190         memcpy(ctx->progress.extract.guid, ctx->wim->hdr.guid, WIM_GUID_LEN);
191         ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
192         ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
193         ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
194         if (ret)
195                 goto out;
196
197         while (ctx->num_streams_remaining) {
198                 struct wim_header_disk pwm_hdr;
199                 struct wim_lookup_table_entry *needed_lte;
200
201                 if (found_lte->resource_location != RESOURCE_NONEXISTENT)
202                         lte_unbind_wim_resource_spec(found_lte);
203                 ret = read_pwm_stream_header(ctx->wim, found_lte, rspec,
204                                              PWM_ALLOW_WIM_HDR, &pwm_hdr);
205                 if (ret)
206                         goto out;
207
208                 if ((found_lte->resource_location != RESOURCE_NONEXISTENT)
209                     && !(found_lte->flags & WIM_RESHDR_FLAG_METADATA)
210                     && (needed_lte = lookup_stream(lookup_table, found_lte->hash))
211                     && (needed_lte->out_refcnt))
212                 {
213                         needed_lte->offset_in_res = found_lte->offset_in_res;
214                         needed_lte->flags = found_lte->flags;
215                         needed_lte->size = found_lte->size;
216
217                         lte_unbind_wim_resource_spec(found_lte);
218                         lte_bind_wim_resource_spec(needed_lte, rspec);
219
220                         ret = (*cbs->begin_stream)(needed_lte, 0,
221                                                    cbs->begin_stream_ctx);
222                         if (ret) {
223                                 lte_unbind_wim_resource_spec(needed_lte);
224                                 goto out;
225                         }
226
227                         ret = extract_stream(needed_lte, needed_lte->size,
228                                              cbs->consume_chunk,
229                                              cbs->consume_chunk_ctx);
230
231                         ret = (*cbs->end_stream)(needed_lte, ret,
232                                                  cbs->end_stream_ctx);
233                         lte_unbind_wim_resource_spec(needed_lte);
234                         if (ret)
235                                 goto out;
236                         ctx->num_streams_remaining--;
237                 } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) {
238                         ret = skip_wim_stream(found_lte);
239                         if (ret)
240                                 goto out;
241                 } else {
242                         u16 part_number = le16_to_cpu(pwm_hdr.part_number);
243                         u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
244
245                         if (part_number != ctx->progress.extract.part_number ||
246                             total_parts != ctx->progress.extract.total_parts ||
247                             memcmp(pwm_hdr.guid, ctx->progress.extract.guid,
248                                    WIM_GUID_LEN))
249                         {
250                                 ctx->progress.extract.part_number = part_number;
251                                 ctx->progress.extract.total_parts = total_parts;
252                                 memcpy(ctx->progress.extract.guid,
253                                        pwm_hdr.guid, WIM_GUID_LEN);
254                                 ret = extract_progress(ctx,
255                                                        WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
256                                 if (ret)
257                                         goto out;
258                         }
259                 }
260         }
261         ret = 0;
262 out:
263         if (found_lte->resource_location != RESOURCE_IN_WIM)
264                 FREE(rspec);
265         free_lookup_table_entry(found_lte);
266         return ret;
267 }
268
269 static int
270 begin_extract_stream_with_progress(struct wim_lookup_table_entry *lte,
271                                    u32 flags, void *_ctx)
272 {
273         struct apply_ctx *ctx = _ctx;
274
275         ctx->cur_stream = lte;
276
277         return (*ctx->saved_cbs->begin_stream)(lte, flags,
278                                                ctx->saved_cbs->begin_stream_ctx);
279 }
280
281 static int
282 consume_chunk_with_progress(const void *chunk, size_t size, void *_ctx)
283 {
284         struct apply_ctx *ctx = _ctx;
285         union wimlib_progress_info *progress = &ctx->progress;
286         int ret;
287
288         if (likely(ctx->supported_features.hard_links)) {
289                 progress->extract.completed_bytes +=
290                         (u64)size * ctx->cur_stream->out_refcnt;
291         } else {
292                 const struct stream_owner *owners = stream_owners(ctx->cur_stream);
293                 for (u32 i = 0; i < ctx->cur_stream->out_refcnt; i++) {
294                         const struct wim_inode *inode = owners[i].inode;
295                         const struct wim_dentry *dentry;
296
297                         list_for_each_entry(dentry,
298                                             &inode->i_extraction_aliases,
299                                             d_extraction_alias_node)
300                         {
301                                 progress->extract.completed_bytes += size;
302                         }
303                 }
304         }
305         if (progress->extract.completed_bytes >= ctx->next_progress) {
306
307                 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
308                 if (ret)
309                         return ret;
310
311                 if (progress->extract.completed_bytes >=
312                     progress->extract.total_bytes)
313                 {
314                         ctx->next_progress = UINT64_MAX;
315                 } else {
316                         ctx->next_progress += progress->extract.total_bytes / 128;
317                         if (ctx->next_progress > progress->extract.total_bytes)
318                                 ctx->next_progress = progress->extract.total_bytes;
319                 }
320         }
321         return (*ctx->saved_cbs->consume_chunk)(chunk, size,
322                                                 ctx->saved_cbs->consume_chunk_ctx);
323 }
324
325 /*
326  * Read the list of single-instance streams to extract and feed their data into
327  * the specified callback functions.
328  *
329  * This handles checksumming each stream.
330  *
331  * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
332  *
333  * This also works if the WIM is being read from a pipe, whereas attempting to
334  * read streams directly (e.g. with read_full_stream_into_buf()) will not.
335  */
336 int
337 extract_stream_list(struct apply_ctx *ctx,
338                     const struct read_stream_list_callbacks *cbs)
339 {
340         struct read_stream_list_callbacks wrapper_cbs = {
341                 .begin_stream      = begin_extract_stream_with_progress,
342                 .begin_stream_ctx  = ctx,
343                 .consume_chunk     = consume_chunk_with_progress,
344                 .consume_chunk_ctx = ctx,
345                 .end_stream        = cbs->end_stream,
346                 .end_stream_ctx    = cbs->end_stream_ctx,
347         };
348         if (ctx->progfunc) {
349                 ctx->saved_cbs = cbs;
350                 cbs = &wrapper_cbs;
351         }
352         if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
353                 return load_streams_from_pipe(ctx, cbs);
354         } else {
355                 return read_stream_list(&ctx->stream_list,
356                                         offsetof(struct wim_lookup_table_entry,
357                                                  extraction_list),
358                                         cbs, VERIFY_STREAM_HASHES);
359         }
360 }
361
362 /* Extract a WIM dentry to standard output.
363  *
364  * This obviously doesn't make sense in all cases.  We return an error if the
365  * dentry does not correspond to a regular file.  Otherwise we extract the
366  * unnamed data stream only.  */
367 static int
368 extract_dentry_to_stdout(struct wim_dentry *dentry,
369                          const struct wim_lookup_table *lookup_table)
370 {
371         struct wim_inode *inode = dentry->d_inode;
372         struct wim_lookup_table_entry *lte;
373         struct filedes _stdout;
374
375         if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
376                                    FILE_ATTRIBUTE_DIRECTORY))
377         {
378                 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
379                       "extracted to standard output", dentry_full_path(dentry));
380                 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
381         }
382
383         lte = inode_unnamed_lte(inode, lookup_table);
384         if (!lte) {
385                 const u8 *hash = inode_unnamed_stream_hash(inode);
386                 if (!is_zero_hash(hash))
387                         return stream_not_found_error(inode, hash);
388                 return 0;
389         }
390
391         filedes_init(&_stdout, STDOUT_FILENO);
392         return extract_full_stream_to_fd(lte, &_stdout);
393 }
394
395 static int
396 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
397                            const struct wim_lookup_table *lookup_table)
398 {
399         for (size_t i = 0; i < num_dentries; i++) {
400                 int ret = extract_dentry_to_stdout(dentries[i], lookup_table);
401                 if (ret)
402                         return ret;
403         }
404         return 0;
405 }
406
407 /**********************************************************************/
408
409 /*
410  * Removes duplicate dentries from the array.
411  *
412  * Returns the new number of dentries, packed at the front of the array.
413  */
414 static size_t
415 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
416 {
417         size_t i, j = 0;
418         for (i = 0; i < num_trees; i++) {
419                 if (!trees[i]->tmp_flag) {
420                         /* Found distinct dentry.  */
421                         trees[i]->tmp_flag = 1;
422                         trees[j++] = trees[i];
423                 }
424         }
425         for (i = 0; i < j; i++)
426                 trees[i]->tmp_flag = 0;
427         return j;
428 }
429
430 /*
431  * Remove dentries that are descendants of other dentries in the array.
432  *
433  * Returns the new number of dentries, packed at the front of the array.
434  */
435 static size_t
436 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
437 {
438         size_t i, j = 0;
439         for (i = 0; i < num_trees; i++)
440                 trees[i]->tmp_flag = 1;
441         for (i = 0; i < num_trees; i++) {
442                 struct wim_dentry *d = trees[i];
443                 while (!dentry_is_root(d)) {
444                         d = d->parent;
445                         if (d->tmp_flag)
446                                 goto tree_contained;
447                 }
448                 trees[j++] = trees[i];
449                 continue;
450
451         tree_contained:
452                 trees[i]->tmp_flag = 0;
453         }
454
455         for (i = 0; i < j; i++)
456                 trees[i]->tmp_flag = 0;
457         return j;
458 }
459
460 static int
461 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
462 {
463         struct list_head *dentry_list = _dentry_list;
464         list_add_tail(&dentry->d_extraction_list_node, dentry_list);
465         return 0;
466 }
467
468 static void
469 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
470 {
471         dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
472 }
473
474 static int
475 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
476 {
477         list_del(&dentry->d_extraction_list_node);
478         dentry_reset_extraction_list_node(dentry);
479         return 0;
480 }
481
482 /*
483  * Build the preliminary list of dentries to be extracted.
484  *
485  * The list maintains the invariant that if d1 and d2 are in the list and d1 is
486  * an ancestor of d2, then d1 appears before d2 in the list.
487  */
488 static void
489 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
490                   size_t num_trees, bool add_ancestors)
491 {
492         INIT_LIST_HEAD(dentry_list);
493
494         /* Add the trees recursively.  */
495         for (size_t i = 0; i < num_trees; i++)
496                 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
497
498         /* If requested, add ancestors of the trees.  */
499         if (add_ancestors) {
500                 for (size_t i = 0; i < num_trees; i++) {
501                         struct wim_dentry *dentry = trees[i];
502                         struct wim_dentry *ancestor;
503                         struct list_head *place_after;
504
505                         if (dentry_is_root(dentry))
506                                 continue;
507
508                         place_after = dentry_list;
509                         ancestor = dentry;
510                         do {
511                                 ancestor = ancestor->parent;
512                                 if (will_extract_dentry(ancestor)) {
513                                         place_after = &ancestor->d_extraction_list_node;
514                                         break;
515                                 }
516                         } while (!dentry_is_root(ancestor));
517
518                         ancestor = dentry;
519                         do {
520                                 ancestor = ancestor->parent;
521                                 if (will_extract_dentry(ancestor))
522                                         break;
523                                 list_add(&ancestor->d_extraction_list_node, place_after);
524                         } while (!dentry_is_root(ancestor));
525                 }
526         }
527 }
528
529 static void
530 destroy_dentry_list(struct list_head *dentry_list)
531 {
532         struct wim_dentry *dentry, *tmp;
533         struct wim_inode *inode;
534
535         list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
536                 inode = dentry->d_inode;
537                 dentry_reset_extraction_list_node(dentry);
538                 inode->i_visited = 0;
539                 if ((void *)dentry->d_extraction_name != (void *)dentry->file_name)
540                         FREE(dentry->d_extraction_name);
541                 dentry->d_extraction_name = NULL;
542                 dentry->d_extraction_name_nchars = 0;
543         }
544 }
545
546 static void
547 destroy_stream_list(struct list_head *stream_list)
548 {
549         struct wim_lookup_table_entry *lte;
550
551         list_for_each_entry(lte, stream_list, extraction_list)
552                 if (lte->out_refcnt > ARRAY_LEN(lte->inline_stream_owners))
553                         FREE(lte->stream_owners);
554 }
555
556 #ifdef __WIN32__
557 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
558 #else
559 static const utf16lechar replacement_char = cpu_to_le16('?');
560 #endif
561
562 static bool
563 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
564 {
565         size_t i;
566
567         if (num_chars == 0)
568                 return true;
569         for (i = 0; i < num_chars; i++) {
570                 switch (name[i]) {
571         #ifdef __WIN32__
572                 case cpu_to_le16('\\'):
573                 case cpu_to_le16(':'):
574                 case cpu_to_le16('*'):
575                 case cpu_to_le16('?'):
576                 case cpu_to_le16('"'):
577                 case cpu_to_le16('<'):
578                 case cpu_to_le16('>'):
579                 case cpu_to_le16('|'):
580         #endif
581                 case cpu_to_le16('/'):
582                 case cpu_to_le16('\0'):
583                         if (fix)
584                                 name[i] = replacement_char;
585                         else
586                                 return false;
587                 }
588         }
589
590 #ifdef __WIN32__
591         if (name[num_chars - 1] == cpu_to_le16(' ') ||
592             name[num_chars - 1] == cpu_to_le16('.'))
593         {
594                 if (fix)
595                         name[num_chars - 1] = replacement_char;
596                 else
597                         return false;
598         }
599 #endif
600         return true;
601 }
602
603 static int
604 dentry_calculate_extraction_name(struct wim_dentry *dentry,
605                                  struct apply_ctx *ctx)
606 {
607         int ret;
608
609         if (!dentry_is_supported(dentry, &ctx->supported_features))
610                 goto skip_dentry;
611
612         if (dentry_is_root(dentry))
613                 return 0;
614
615         if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
616                 dentry->d_extraction_name = dentry->file_name;
617                 dentry->d_extraction_name_nchars = dentry->file_name_nbytes /
618                                                    sizeof(utf16lechar);
619                 return 0;
620         }
621
622         if (!ctx->supported_features.case_sensitive_filenames) {
623                 struct wim_dentry *other;
624                 list_for_each_entry(other, &dentry->d_ci_conflict_list,
625                                     d_ci_conflict_list)
626                 {
627                         if (will_extract_dentry(other)) {
628                                 if (ctx->extract_flags &
629                                     WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
630                                         WARNING("\"%"TS"\" has the same "
631                                                 "case-insensitive name as "
632                                                 "\"%"TS"\"; extracting "
633                                                 "dummy name instead",
634                                                 dentry_full_path(dentry),
635                                                 dentry_full_path(other));
636                                         goto out_replace;
637                                 } else {
638                                         WARNING("Not extracting \"%"TS"\": "
639                                                 "has same case-insensitive "
640                                                 "name as \"%"TS"\"",
641                                                 dentry_full_path(dentry),
642                                                 dentry_full_path(other));
643                                         goto skip_dentry;
644                                 }
645                         }
646                 }
647         }
648
649         if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
650                 ret = utf16le_get_tstr(dentry->file_name,
651                                        dentry->file_name_nbytes,
652                                        (const tchar **)&dentry->d_extraction_name,
653                                        &dentry->d_extraction_name_nchars);
654                 dentry->d_extraction_name_nchars /= sizeof(tchar);
655                 return ret;
656         } else {
657                 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
658                 {
659                         WARNING("\"%"TS"\" has an invalid filename "
660                                 "that is not supported on this platform; "
661                                 "extracting dummy name instead",
662                                 dentry_full_path(dentry));
663                         goto out_replace;
664                 } else {
665                         WARNING("Not extracting \"%"TS"\": has an invalid filename "
666                                 "that is not supported on this platform",
667                                 dentry_full_path(dentry));
668                         goto skip_dentry;
669                 }
670         }
671
672 out_replace:
673         {
674                 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
675
676                 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
677                 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
678
679                 const tchar *tchar_name;
680                 size_t tchar_nchars;
681
682                 ret = utf16le_get_tstr(utf16_name_copy,
683                                        dentry->file_name_nbytes,
684                                        &tchar_name, &tchar_nchars);
685                 if (ret)
686                         return ret;
687
688                 tchar_nchars /= sizeof(tchar);
689
690                 size_t fixed_name_num_chars = tchar_nchars;
691                 tchar fixed_name[tchar_nchars + 50];
692
693                 tmemcpy(fixed_name, tchar_name, tchar_nchars);
694                 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
695                                                  T(" (invalid filename #%lu)"),
696                                                  ++ctx->invalid_sequence);
697
698                 utf16le_put_tstr(tchar_name);
699
700                 dentry->d_extraction_name = memdup(fixed_name,
701                                                    2 * fixed_name_num_chars + 2);
702                 if (!dentry->d_extraction_name)
703                         return WIMLIB_ERR_NOMEM;
704                 dentry->d_extraction_name_nchars = fixed_name_num_chars;
705         }
706         return 0;
707
708 skip_dentry:
709         for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
710         return 0;
711 }
712
713 /*
714  * Calculate the actual filename component at which each WIM dentry will be
715  * extracted, with special handling for dentries that are unsupported by the
716  * extraction backend or have invalid names.
717  *
718  * ctx->supported_features must be filled in.
719  *
720  * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
721  */
722 static int
723 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
724                                        struct apply_ctx *ctx)
725 {
726         struct list_head *prev, *cur;
727
728         /* Can't use list_for_each_entry() because a call to
729          * dentry_calculate_extraction_name() may delete the current dentry and
730          * its children from the list.  */
731
732         prev = dentry_list;
733         for (;;) {
734                 struct wim_dentry *dentry;
735                 int ret;
736
737                 cur = prev->next;
738                 if (cur == dentry_list)
739                         break;
740
741                 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
742
743                 ret = dentry_calculate_extraction_name(dentry, ctx);
744                 if (ret)
745                         return ret;
746
747                 if (prev->next == cur)
748                         prev = cur;
749                 else
750                         ; /* Current dentry and its children (which follow in
751                              the list) were deleted.  prev stays the same.  */
752         }
753         return 0;
754 }
755
756 static int
757 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
758                        struct wim_lookup_table *lookup_table)
759 {
760         struct wim_inode *inode = dentry->d_inode;
761         struct wim_lookup_table_entry *lte;
762         int ret;
763         bool force = false;
764
765         /* Special case:  when extracting from a pipe, the WIM lookup table is
766          * initially empty, so "resolving" an inode's streams is initially not
767          * possible.  However, we still need to keep track of which streams,
768          * identified by SHA1 message digests, need to be extracted, so we
769          * "resolve" the inode's streams anyway by allocating new entries.  */
770         if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
771                 force = true;
772         ret = inode_resolve_streams(inode, lookup_table, force);
773         if (ret)
774                 return ret;
775         for (u32 i = 0; i <= inode->i_num_ads; i++) {
776                 lte = inode_stream_lte_resolved(inode, i);
777                 if (lte)
778                         lte->out_refcnt = 0;
779         }
780         return 0;
781 }
782
783 /*
784  * For each dentry to be extracted, resolve all streams in the corresponding
785  * inode and set 'out_refcnt' in each to 0.
786  *
787  * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
788  */
789 static int
790 dentry_list_resolve_streams(struct list_head *dentry_list,
791                             struct apply_ctx *ctx)
792 {
793         struct wim_dentry *dentry;
794         int ret;
795
796         list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
797                 ret = dentry_resolve_streams(dentry,
798                                              ctx->extract_flags,
799                                              ctx->wim->lookup_table);
800                 if (ret)
801                         return ret;
802         }
803         return 0;
804 }
805
806 static int
807 ref_stream(struct wim_lookup_table_entry *lte, u32 stream_idx,
808            struct wim_dentry *dentry, struct apply_ctx *ctx)
809 {
810         struct wim_inode *inode = dentry->d_inode;
811         struct stream_owner *stream_owners;
812
813         if (!lte)
814                 return 0;
815
816         /* Tally the size only for each extraction of the stream (not hard
817          * links).  */
818         if (inode->i_visited && ctx->supported_features.hard_links)
819                 return 0;
820
821         ctx->progress.extract.total_bytes += lte->size;
822         ctx->progress.extract.num_streams++;
823
824         if (inode->i_visited)
825                 return 0;
826
827         /* Add stream to the dentry_list only one time, even if it's going
828          * to be extracted to multiple inodes.  */
829         if (lte->out_refcnt == 0) {
830                 list_add_tail(&lte->extraction_list, &ctx->stream_list);
831                 ctx->num_streams_remaining++;
832         }
833
834         /* If inode not yet been visited, append it to the stream_owners array.  */
835         if (lte->out_refcnt < ARRAY_LEN(lte->inline_stream_owners)) {
836                 stream_owners = lte->inline_stream_owners;
837         } else {
838                 struct stream_owner *prev_stream_owners;
839                 size_t alloc_stream_owners;
840
841                 if (lte->out_refcnt == ARRAY_LEN(lte->inline_stream_owners)) {
842                         prev_stream_owners = NULL;
843                         alloc_stream_owners = ARRAY_LEN(lte->inline_stream_owners);
844                 } else {
845                         prev_stream_owners = lte->stream_owners;
846                         alloc_stream_owners = lte->alloc_stream_owners;
847                 }
848
849                 if (lte->out_refcnt == alloc_stream_owners) {
850                         alloc_stream_owners *= 2;
851                         stream_owners = REALLOC(prev_stream_owners,
852                                                alloc_stream_owners *
853                                                 sizeof(stream_owners[0]));
854                         if (!stream_owners)
855                                 return WIMLIB_ERR_NOMEM;
856                         if (!prev_stream_owners) {
857                                 memcpy(stream_owners,
858                                        lte->inline_stream_owners,
859                                        sizeof(lte->inline_stream_owners));
860                         }
861                         lte->stream_owners = stream_owners;
862                         lte->alloc_stream_owners = alloc_stream_owners;
863                 }
864                 stream_owners = lte->stream_owners;
865         }
866         stream_owners[lte->out_refcnt].inode = inode;
867         if (stream_idx == 0) {
868                 stream_owners[lte->out_refcnt].stream_name = NULL;
869         } else {
870                 stream_owners[lte->out_refcnt].stream_name =
871                         inode->i_ads_entries[stream_idx - 1].stream_name;
872         }
873         lte->out_refcnt++;
874         return 0;
875 }
876
877 static int
878 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
879 {
880         struct wim_inode *inode = dentry->d_inode;
881         int ret;
882
883         /* The unnamed data stream will always be extracted, except in an
884          * unlikely case.  */
885         if (!inode_is_encrypted_directory(inode)) {
886                 u16 stream_idx;
887                 struct wim_lookup_table_entry *stream;
888
889                 stream = inode_unnamed_stream_resolved(inode, &stream_idx);
890                 ret = ref_stream(stream, stream_idx, dentry, ctx);
891                 if (ret)
892                         return ret;
893         }
894
895         /* Named data streams will be extracted only if supported in the current
896          * extraction mode and volume, and to avoid complications, if not doing
897          * a linked extraction.  */
898         if (ctx->supported_features.named_data_streams) {
899                 for (u16 i = 0; i < inode->i_num_ads; i++) {
900                         if (!ads_entry_is_named_stream(&inode->i_ads_entries[i]))
901                                 continue;
902                         ret = ref_stream(inode->i_ads_entries[i].lte, i + 1,
903                                          dentry, ctx);
904                         if (ret)
905                                 return ret;
906                 }
907         }
908         inode->i_visited = 1;
909         return 0;
910 }
911
912 /*
913  * For each dentry to be extracted, iterate through the data streams of the
914  * corresponding inode.  For each such stream that is not to be ignored due to
915  * the supported features or extraction flags, add it to the list of streams to
916  * be extracted (ctx->stream_list) if not already done so.
917  *
918  * Also builds a mapping from each stream to the inodes referencing it.
919  *
920  * This also initializes the extract progress info with byte and stream
921  * information.
922  *
923  * ctx->supported_features must be filled in.
924  *
925  * Possible error codes: WIMLIB_ERR_NOMEM.
926  */
927 static int
928 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
929 {
930         struct wim_dentry *dentry;
931         int ret;
932
933         list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
934                 ret = dentry_ref_streams(dentry, ctx);
935                 if (ret)
936                         return ret;
937         }
938         list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
939                 dentry->d_inode->i_visited = 0;
940         return 0;
941 }
942
943 static void
944 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
945 {
946         struct wim_dentry *dentry;
947         struct wim_inode *inode;
948
949         list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
950                 inode = dentry->d_inode;
951                 if (!inode->i_visited)
952                         INIT_LIST_HEAD(&inode->i_extraction_aliases);
953                 list_add_tail(&dentry->d_extraction_alias_node,
954                               &inode->i_extraction_aliases);
955                 inode->i_visited = 1;
956         }
957         list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
958                 dentry->d_inode->i_visited = 0;
959 }
960
961 static void
962 inode_tally_features(const struct wim_inode *inode,
963                      struct wim_features *features)
964 {
965         if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
966                 features->archive_files++;
967         if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
968                 features->hidden_files++;
969         if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
970                 features->system_files++;
971         if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
972                 features->compressed_files++;
973         if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
974                 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
975                         features->encrypted_directories++;
976                 else
977                         features->encrypted_files++;
978         }
979         if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
980                 features->not_context_indexed_files++;
981         if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
982                 features->sparse_files++;
983         if (inode_has_named_stream(inode))
984                 features->named_data_streams++;
985         if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
986                 features->reparse_points++;
987                 if (inode_is_symlink(inode))
988                         features->symlink_reparse_points++;
989                 else
990                         features->other_reparse_points++;
991         }
992         if (inode->i_security_id != -1)
993                 features->security_descriptors++;
994         if (inode_has_unix_data(inode))
995                 features->unix_data++;
996 }
997
998 /* Tally features necessary to extract a dentry and the corresponding inode.  */
999 static void
1000 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1001 {
1002         struct wim_inode *inode = dentry->d_inode;
1003
1004         if (dentry_has_short_name(dentry))
1005                 features->short_names++;
1006
1007         if (inode->i_visited) {
1008                 features->hard_links++;
1009         } else {
1010                 inode_tally_features(inode, features);
1011                 inode->i_visited = 1;
1012         }
1013 }
1014
1015 /* Tally the features necessary to extract the specified dentries.  */
1016 static void
1017 dentry_list_get_features(struct list_head *dentry_list,
1018                          struct wim_features *features)
1019 {
1020         struct wim_dentry *dentry;
1021
1022         list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1023                 dentry_tally_features(dentry, features);
1024
1025         list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1026                 dentry->d_inode->i_visited = 0;
1027 }
1028
1029 static int
1030 do_feature_check(const struct wim_features *required_features,
1031                  const struct wim_features *supported_features,
1032                  int extract_flags)
1033 {
1034         /* File attributes.  */
1035         if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1036                 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1037                  * We're an archive program, so theoretically we can do what we
1038                  * want with it.  */
1039
1040                 if (required_features->hidden_files &&
1041                     !supported_features->hidden_files)
1042                         WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1043                                 required_features->hidden_files);
1044
1045                 if (required_features->system_files &&
1046                     !supported_features->system_files)
1047                         WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1048                                 required_features->system_files);
1049
1050                 if (required_features->compressed_files &&
1051                     !supported_features->compressed_files)
1052                         WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1053                                 required_features->compressed_files);
1054
1055                 if (required_features->not_context_indexed_files &&
1056                     !supported_features->not_context_indexed_files)
1057                         WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1058                                 required_features->not_context_indexed_files);
1059
1060                 if (required_features->sparse_files &&
1061                     !supported_features->sparse_files)
1062                         WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1063                                 required_features->sparse_files);
1064
1065                 if (required_features->encrypted_directories &&
1066                     !supported_features->encrypted_directories)
1067                         WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1068                                 required_features->encrypted_directories);
1069         }
1070
1071         /* Encrypted files.  */
1072         if (required_features->encrypted_files &&
1073             !supported_features->encrypted_files)
1074                 WARNING("Ignoring %lu encrypted files",
1075                         required_features->encrypted_files);
1076
1077         /* Named data streams.  */
1078         if (required_features->named_data_streams &&
1079             (!supported_features->named_data_streams))
1080                 WARNING("Ignoring named data streams of %lu files",
1081                         required_features->named_data_streams);
1082
1083         /* Hard links.  */
1084         if (required_features->hard_links && !supported_features->hard_links)
1085                 WARNING("Extracting %lu hard links as independent files",
1086                         required_features->hard_links);
1087
1088         /* Symbolic links and reparse points.  */
1089         if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1090             required_features->symlink_reparse_points &&
1091             !supported_features->symlink_reparse_points &&
1092             !supported_features->reparse_points)
1093         {
1094                 ERROR("Extraction backend does not support symbolic links!");
1095                 return WIMLIB_ERR_UNSUPPORTED;
1096         }
1097         if (required_features->reparse_points &&
1098             !supported_features->reparse_points)
1099         {
1100                 if (supported_features->symlink_reparse_points) {
1101                         if (required_features->other_reparse_points) {
1102                                 WARNING("Ignoring %lu non-symlink/junction "
1103                                         "reparse point files",
1104                                         required_features->other_reparse_points);
1105                         }
1106                 } else {
1107                         WARNING("Ignoring %lu reparse point files",
1108                                 required_features->reparse_points);
1109                 }
1110         }
1111
1112         /* Security descriptors.  */
1113         if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1114                                WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1115              == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1116             required_features->security_descriptors &&
1117             !supported_features->security_descriptors)
1118         {
1119                 ERROR("Extraction backend does not support security descriptors!");
1120                 return WIMLIB_ERR_UNSUPPORTED;
1121         }
1122         if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1123             required_features->security_descriptors &&
1124             !supported_features->security_descriptors)
1125                 WARNING("Ignoring Windows NT security descriptors of %lu files",
1126                         required_features->security_descriptors);
1127
1128         /* UNIX data.  */
1129         if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
1130             required_features->unix_data && !supported_features->unix_data)
1131         {
1132                 ERROR("Extraction backend does not support UNIX data!");
1133                 return WIMLIB_ERR_UNSUPPORTED;
1134         }
1135
1136         /* DOS Names.  */
1137         if (required_features->short_names &&
1138             !supported_features->short_names)
1139         {
1140                 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1141                         ERROR("Extraction backend does not support DOS names!");
1142                         return WIMLIB_ERR_UNSUPPORTED;
1143                 }
1144                 WARNING("Ignoring DOS names of %lu files",
1145                         required_features->short_names);
1146         }
1147
1148         /* Timestamps.  */
1149         if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1150             !supported_features->timestamps)
1151         {
1152                 ERROR("Extraction backend does not support timestamps!");
1153                 return WIMLIB_ERR_UNSUPPORTED;
1154         }
1155
1156         return 0;
1157 }
1158
1159 static const struct apply_operations *
1160 select_apply_operations(int extract_flags)
1161 {
1162 #ifdef WITH_NTFS_3G
1163         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1164                 return &ntfs_3g_apply_ops;
1165 #endif
1166 #ifdef __WIN32__
1167         return &win32_apply_ops;
1168 #else
1169         return &unix_apply_ops;
1170 #endif
1171 }
1172
1173 static int
1174 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1175               const tchar *target, int extract_flags)
1176 {
1177         const struct apply_operations *ops;
1178         struct apply_ctx *ctx;
1179         int ret;
1180         LIST_HEAD(dentry_list);
1181
1182         if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1183                 ret = extract_dentries_to_stdout(trees, num_trees,
1184                                                  wim->lookup_table);
1185                 goto out;
1186         }
1187
1188         num_trees = remove_duplicate_trees(trees, num_trees);
1189         num_trees = remove_contained_trees(trees, num_trees);
1190
1191         ops = select_apply_operations(extract_flags);
1192
1193         if (num_trees > 1 && ops->single_tree_only) {
1194                 ERROR("Extracting multiple directory trees "
1195                       "at once is not supported in %s extraction mode!",
1196                       ops->name);
1197                 ret = WIMLIB_ERR_UNSUPPORTED;
1198                 goto out;
1199         }
1200
1201         ctx = CALLOC(1, ops->context_size);
1202         if (!ctx) {
1203                 ret = WIMLIB_ERR_NOMEM;
1204                 goto out;
1205         }
1206
1207         ctx->wim = wim;
1208         ctx->target = target;
1209         ctx->target_nchars = tstrlen(target);
1210         ctx->extract_flags = extract_flags;
1211         if (ctx->wim->progfunc) {
1212                 ctx->progfunc = ctx->wim->progfunc;
1213                 ctx->progctx = ctx->wim->progctx;
1214                 ctx->progress.extract.image = wim->current_image;
1215                 ctx->progress.extract.extract_flags = (extract_flags &
1216                                                        WIMLIB_EXTRACT_MASK_PUBLIC);
1217                 ctx->progress.extract.wimfile_name = wim->filename;
1218                 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1219                                                                          wim->current_image);
1220                 ctx->progress.extract.target = target;
1221         }
1222         INIT_LIST_HEAD(&ctx->stream_list);
1223
1224         ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1225         if (ret)
1226                 goto out_cleanup;
1227
1228         build_dentry_list(&dentry_list, trees, num_trees,
1229                           !(extract_flags &
1230                             WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1231
1232         dentry_list_get_features(&dentry_list, &ctx->required_features);
1233
1234         ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1235                                ctx->extract_flags);
1236         if (ret)
1237                 goto out_cleanup;
1238
1239         ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1240         if (ret)
1241                 goto out_cleanup;
1242
1243         ret = dentry_list_resolve_streams(&dentry_list, ctx);
1244         if (ret)
1245                 goto out_cleanup;
1246
1247         ret = dentry_list_ref_streams(&dentry_list, ctx);
1248         if (ret)
1249                 goto out_cleanup;
1250
1251         dentry_list_build_inode_alias_lists(&dentry_list);
1252
1253         if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1254                 /* When extracting from a pipe, the number of bytes of data to
1255                  * extract can't be determined in the normal way (examining the
1256                  * lookup table), since at this point all we have is a set of
1257                  * SHA1 message digests of streams that need to be extracted.
1258                  * However, we can get a reasonably accurate estimate by taking
1259                  * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1260                  * data.  This does assume that a full image is being extracted,
1261                  * but currently there is no API for doing otherwise.  (Also,
1262                  * subtract <HARDLINKBYTES> from this if hard links are
1263                  * supported by the extraction mode.)  */
1264                 ctx->progress.extract.total_bytes =
1265                         wim_info_get_image_total_bytes(wim->wim_info,
1266                                                        wim->current_image);
1267                 if (ctx->supported_features.hard_links) {
1268                         ctx->progress.extract.total_bytes -=
1269                                 wim_info_get_image_hard_link_bytes(wim->wim_info,
1270                                                                    wim->current_image);
1271                 }
1272         }
1273
1274         ret = extract_progress(ctx,
1275                                ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1276                                        WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1277                                        WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1278         if (ret)
1279                 goto out_cleanup;
1280
1281         ret = (*ops->extract)(&dentry_list, ctx);
1282         if (ret)
1283                 goto out_cleanup;
1284
1285         if (ctx->progress.extract.completed_bytes <
1286             ctx->progress.extract.total_bytes)
1287         {
1288                 ctx->progress.extract.completed_bytes =
1289                         ctx->progress.extract.total_bytes;
1290                 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1291                 if (ret)
1292                         goto out_cleanup;
1293         }
1294
1295         ret = extract_progress(ctx,
1296                                ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1297                                        WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1298                                        WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1299 out_cleanup:
1300         destroy_stream_list(&ctx->stream_list);
1301         destroy_dentry_list(&dentry_list);
1302         FREE(ctx);
1303 out:
1304         return ret;
1305 }
1306
1307 static int
1308 mkdir_if_needed(const tchar *target)
1309 {
1310         struct stat stbuf;
1311         if (tstat(target, &stbuf)) {
1312                 if (errno == ENOENT) {
1313                         if (tmkdir(target, 0755)) {
1314                                 ERROR_WITH_ERRNO("Failed to create directory "
1315                                                  "\"%"TS"\"", target);
1316                                 return WIMLIB_ERR_MKDIR;
1317                         }
1318                 } else {
1319                         ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target);
1320                         return WIMLIB_ERR_STAT;
1321                 }
1322         } else if (!S_ISDIR(stbuf.st_mode)) {
1323                 ERROR("\"%"TS"\" is not a directory", target);
1324                 return WIMLIB_ERR_NOTDIR;
1325         }
1326         return 0;
1327 }
1328
1329 /* Make sure the extraction flags make sense, and update them if needed.  */
1330 static int
1331 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1332 {
1333         int extract_flags = *extract_flags_p;
1334
1335         /* Check for invalid flag combinations  */
1336
1337         if ((extract_flags &
1338              (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1339               WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1340                                                     WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1341                 return WIMLIB_ERR_INVALID_PARAM;
1342
1343         if ((extract_flags &
1344              (WIMLIB_EXTRACT_FLAG_RPFIX |
1345               WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1346                                                 WIMLIB_EXTRACT_FLAG_NORPFIX))
1347                 return WIMLIB_ERR_INVALID_PARAM;
1348
1349 #ifndef WITH_NTFS_3G
1350         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1351                 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
1352                       "        it cannot apply a WIM image directly to an NTFS volume.");
1353                 return WIMLIB_ERR_UNSUPPORTED;
1354         }
1355 #endif
1356
1357 #ifndef __WIN32__
1358         if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1359                 ERROR("WIMBoot extraction is only supported on Windows!");
1360                 return WIMLIB_ERR_UNSUPPORTED;
1361         }
1362 #endif
1363
1364         if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1365                               WIMLIB_EXTRACT_FLAG_NORPFIX |
1366                               WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1367                                         WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1368         {
1369                 /* For full-image extraction, do reparse point fixups by default
1370                  * if the WIM header says they are enabled.  */
1371                 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1372                         extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1373         }
1374
1375         *extract_flags_p = extract_flags;
1376         return 0;
1377 }
1378
1379 static u32
1380 get_wildcard_flags(int extract_flags)
1381 {
1382         u32 wildcard_flags = 0;
1383
1384         if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB)
1385                 wildcard_flags |= WILDCARD_FLAG_ERROR_IF_NO_MATCH;
1386         else
1387                 wildcard_flags |= WILDCARD_FLAG_WARN_IF_NO_MATCH;
1388
1389         if (default_ignore_case)
1390                 wildcard_flags |= WILDCARD_FLAG_CASE_INSENSITIVE;
1391
1392         return wildcard_flags;
1393 }
1394
1395 struct append_dentry_ctx {
1396         struct wim_dentry **dentries;
1397         size_t num_dentries;
1398         size_t num_alloc_dentries;
1399 };
1400
1401 static int
1402 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1403 {
1404         struct append_dentry_ctx *ctx = _ctx;
1405
1406         if (ctx->num_dentries == ctx->num_alloc_dentries) {
1407                 struct wim_dentry **new_dentries;
1408                 size_t new_length;
1409
1410                 new_length = max(ctx->num_alloc_dentries + 8,
1411                                  ctx->num_alloc_dentries * 3 / 2);
1412                 new_dentries = REALLOC(ctx->dentries,
1413                                        new_length * sizeof(ctx->dentries[0]));
1414                 if (new_dentries == NULL)
1415                         return WIMLIB_ERR_NOMEM;
1416                 ctx->dentries = new_dentries;
1417                 ctx->num_alloc_dentries = new_length;
1418         }
1419         ctx->dentries[ctx->num_dentries++] = dentry;
1420         return 0;
1421 }
1422
1423 static int
1424 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1425                         const tchar * const *paths, size_t num_paths,
1426                         int extract_flags)
1427 {
1428         int ret;
1429         struct wim_dentry **trees;
1430         size_t num_trees;
1431
1432         if (wim == NULL || target == NULL || target[0] == T('\0') ||
1433             (num_paths != 0 && paths == NULL))
1434                 return WIMLIB_ERR_INVALID_PARAM;
1435
1436         ret = check_extract_flags(wim, &extract_flags);
1437         if (ret)
1438                 return ret;
1439
1440         ret = select_wim_image(wim, image);
1441         if (ret)
1442                 return ret;
1443
1444         ret = wim_checksum_unhashed_streams(wim);
1445         if (ret)
1446                 return ret;
1447
1448         if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1449                               WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1450             (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1451         {
1452                 ret = mkdir_if_needed(target);
1453                 if (ret)
1454                         return ret;
1455         }
1456
1457         if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1458
1459                 struct append_dentry_ctx append_dentry_ctx = {
1460                         .dentries = NULL,
1461                         .num_dentries = 0,
1462                         .num_alloc_dentries = 0,
1463                 };
1464
1465                 u32 wildcard_flags = get_wildcard_flags(extract_flags);
1466
1467                 for (size_t i = 0; i < num_paths; i++) {
1468                         tchar *path = canonicalize_wim_path(paths[i]);
1469                         if (path == NULL) {
1470                                 ret = WIMLIB_ERR_NOMEM;
1471                                 trees = append_dentry_ctx.dentries;
1472                                 goto out_free_trees;
1473                         }
1474                         ret = expand_wildcard(wim, path,
1475                                               append_dentry_cb,
1476                                               &append_dentry_ctx,
1477                                               wildcard_flags);
1478                         FREE(path);
1479                         if (ret) {
1480                                 trees = append_dentry_ctx.dentries;
1481                                 goto out_free_trees;
1482                         }
1483                 }
1484                 trees = append_dentry_ctx.dentries;
1485                 num_trees = append_dentry_ctx.num_dentries;
1486         } else {
1487                 trees = MALLOC(num_paths * sizeof(trees[0]));
1488                 if (trees == NULL)
1489                         return WIMLIB_ERR_NOMEM;
1490
1491                 for (size_t i = 0; i < num_paths; i++) {
1492
1493                         tchar *path = canonicalize_wim_path(paths[i]);
1494                         if (path == NULL) {
1495                                 ret = WIMLIB_ERR_NOMEM;
1496                                 goto out_free_trees;
1497                         }
1498
1499                         trees[i] = get_dentry(wim, path,
1500                                               WIMLIB_CASE_PLATFORM_DEFAULT);
1501                         FREE(path);
1502                         if (trees[i] == NULL) {
1503                                   ERROR("Path \"%"TS"\" does not exist "
1504                                         "in WIM image %d",
1505                                         paths[i], wim->current_image);
1506                                   ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1507                                   goto out_free_trees;
1508                         }
1509                 }
1510                 num_trees = num_paths;
1511         }
1512
1513         if (num_trees == 0) {
1514                 ret = 0;
1515                 goto out_free_trees;
1516         }
1517
1518         ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1519 out_free_trees:
1520         FREE(trees);
1521         return ret;
1522 }
1523
1524 static int
1525 extract_single_image(WIMStruct *wim, int image,
1526                      const tchar *target, int extract_flags)
1527 {
1528         const tchar *path = WIMLIB_WIM_ROOT_PATH;
1529         extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1530         return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1531 }
1532
1533 static const tchar * const filename_forbidden_chars =
1534 T(
1535 #ifdef __WIN32__
1536 "<>:\"/\\|?*"
1537 #else
1538 "/"
1539 #endif
1540 );
1541
1542 /* This function checks if it is okay to use a WIM image's name as a directory
1543  * name.  */
1544 static bool
1545 image_name_ok_as_dir(const tchar *image_name)
1546 {
1547         return image_name && *image_name &&
1548                 !tstrpbrk(image_name, filename_forbidden_chars) &&
1549                 tstrcmp(image_name, T(".")) &&
1550                 tstrcmp(image_name, T(".."));
1551 }
1552
1553 /* Extracts all images from the WIM to the directory @target, with the images
1554  * placed in subdirectories named by their image names. */
1555 static int
1556 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1557 {
1558         size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
1559         size_t output_path_len = tstrlen(target);
1560         tchar buf[output_path_len + 1 + image_name_max_len + 1];
1561         int ret;
1562         int image;
1563         const tchar *image_name;
1564
1565         extract_flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
1566
1567         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1568                 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1569                 return WIMLIB_ERR_INVALID_PARAM;
1570         }
1571
1572         ret = mkdir_if_needed(target);
1573         if (ret)
1574                 return ret;
1575         tmemcpy(buf, target, output_path_len);
1576         buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1577         for (image = 1; image <= wim->hdr.image_count; image++) {
1578                 image_name = wimlib_get_image_name(wim, image);
1579                 if (image_name_ok_as_dir(image_name)) {
1580                         tstrcpy(buf + output_path_len + 1, image_name);
1581                 } else {
1582                         /* Image name is empty or contains forbidden characters.
1583                          * Use image number instead. */
1584                         tsprintf(buf + output_path_len + 1, T("%d"), image);
1585                 }
1586                 ret = extract_single_image(wim, image, buf, extract_flags);
1587                 if (ret)
1588                         return ret;
1589         }
1590         return 0;
1591 }
1592
1593 static int
1594 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1595                         int extract_flags)
1596 {
1597         if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1598                              WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1599                              WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1600                 return WIMLIB_ERR_INVALID_PARAM;
1601
1602         if (image == WIMLIB_ALL_IMAGES)
1603                 return extract_all_images(wim, target, extract_flags);
1604         else
1605                 return extract_single_image(wim, image, target, extract_flags);
1606 }
1607
1608
1609 /****************************************************************************
1610  *                          Extraction API                                  *
1611  ****************************************************************************/
1612
1613 WIMLIBAPI int
1614 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1615                      const tchar * const *paths, size_t num_paths,
1616                      int extract_flags)
1617 {
1618         if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1619                 return WIMLIB_ERR_INVALID_PARAM;
1620
1621         return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1622                                        extract_flags);
1623 }
1624
1625 WIMLIBAPI int
1626 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1627                         const tchar *path_list_file, int extract_flags)
1628 {
1629         int ret;
1630         tchar **paths;
1631         size_t num_paths;
1632         void *mem;
1633
1634         ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1635         if (ret) {
1636                 ERROR("Failed to read path list file \"%"TS"\"",
1637                       path_list_file);
1638                 return ret;
1639         }
1640
1641         ret = wimlib_extract_paths(wim, image, target,
1642                                    (const tchar * const *)paths, num_paths,
1643                                    extract_flags);
1644         FREE(paths);
1645         FREE(mem);
1646         return ret;
1647 }
1648
1649 WIMLIBAPI int
1650 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1651                                              const tchar *image_num_or_name,
1652                                              const tchar *target,
1653                                              int extract_flags,
1654                                              wimlib_progress_func_t progfunc,
1655                                              void *progctx)
1656 {
1657         int ret;
1658         WIMStruct *pwm;
1659         struct filedes *in_fd;
1660         int image;
1661         unsigned i;
1662
1663         if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1664                 return WIMLIB_ERR_INVALID_PARAM;
1665
1666         /* Read the WIM header from the pipe and get a WIMStruct to represent
1667          * the pipable WIM.  Caveats:  Unlike getting a WIMStruct with
1668          * wimlib_open_wim(), getting a WIMStruct in this way will result in
1669          * an empty lookup table, no XML data read, and no filename set.  */
1670         ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1671                                     progfunc, progctx);
1672         if (ret)
1673                 return ret;
1674
1675         /* Sanity check to make sure this is a pipable WIM.  */
1676         if (pwm->hdr.magic != PWM_MAGIC) {
1677                 ERROR("The WIM being read from file descriptor %d "
1678                       "is not pipable!", pipe_fd);
1679                 ret = WIMLIB_ERR_NOT_PIPABLE;
1680                 goto out_wimlib_free;
1681         }
1682
1683         /* Sanity check to make sure the first part of a pipable split WIM is
1684          * sent over the pipe first.  */
1685         if (pwm->hdr.part_number != 1) {
1686                 ERROR("The first part of the split WIM must be "
1687                       "sent over the pipe first.");
1688                 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1689                 goto out_wimlib_free;
1690         }
1691
1692         in_fd = &pwm->in_fd;
1693         wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1694
1695         /* As mentioned, the WIMStruct we created from the pipe does not have
1696          * XML data yet.  Fix this by reading the extra copy of the XML data
1697          * that directly follows the header in pipable WIMs.  (Note: see
1698          * write_pipable_wim() for more details about the format of pipable
1699          * WIMs.)  */
1700         {
1701                 struct wim_lookup_table_entry xml_lte;
1702                 struct wim_resource_spec xml_rspec;
1703                 ret = read_pwm_stream_header(pwm, &xml_lte, &xml_rspec, 0, NULL);
1704                 if (ret)
1705                         goto out_wimlib_free;
1706
1707                 if (!(xml_lte.flags & WIM_RESHDR_FLAG_METADATA))
1708                 {
1709                         ERROR("Expected XML data, but found non-metadata "
1710                               "stream.");
1711                         ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1712                         goto out_wimlib_free;
1713                 }
1714
1715                 wim_res_spec_to_hdr(&xml_rspec, &pwm->hdr.xml_data_reshdr);
1716
1717                 ret = read_wim_xml_data(pwm);
1718                 if (ret)
1719                         goto out_wimlib_free;
1720
1721                 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
1722                         ERROR("Image count in XML data is not the same as in WIM header.");
1723                         ret = WIMLIB_ERR_IMAGE_COUNT;
1724                         goto out_wimlib_free;
1725                 }
1726         }
1727
1728         /* Get image index (this may use the XML data that was just read to
1729          * resolve an image name).  */
1730         if (image_num_or_name) {
1731                 image = wimlib_resolve_image(pwm, image_num_or_name);
1732                 if (image == WIMLIB_NO_IMAGE) {
1733                         ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
1734                               image_num_or_name);
1735                         ret = WIMLIB_ERR_INVALID_IMAGE;
1736                         goto out_wimlib_free;
1737                 } else if (image == WIMLIB_ALL_IMAGES) {
1738                         ERROR("Applying all images from a pipe is not supported!");
1739                         ret = WIMLIB_ERR_INVALID_IMAGE;
1740                         goto out_wimlib_free;
1741                 }
1742         } else {
1743                 if (pwm->hdr.image_count != 1) {
1744                         ERROR("No image was specified, but the pipable WIM "
1745                               "did not contain exactly 1 image");
1746                         ret = WIMLIB_ERR_INVALID_IMAGE;
1747                         goto out_wimlib_free;
1748                 }
1749                 image = 1;
1750         }
1751
1752         /* Load the needed metadata resource.  */
1753         for (i = 1; i <= pwm->hdr.image_count; i++) {
1754                 struct wim_lookup_table_entry *metadata_lte;
1755                 struct wim_image_metadata *imd;
1756                 struct wim_resource_spec *metadata_rspec;
1757
1758                 metadata_lte = new_lookup_table_entry();
1759                 if (metadata_lte == NULL) {
1760                         ret = WIMLIB_ERR_NOMEM;
1761                         goto out_wimlib_free;
1762                 }
1763                 metadata_rspec = MALLOC(sizeof(struct wim_resource_spec));
1764                 if (metadata_rspec == NULL) {
1765                         ret = WIMLIB_ERR_NOMEM;
1766                         free_lookup_table_entry(metadata_lte);
1767                         goto out_wimlib_free;
1768                 }
1769
1770                 ret = read_pwm_stream_header(pwm, metadata_lte, metadata_rspec, 0, NULL);
1771                 imd = pwm->image_metadata[i - 1];
1772                 imd->metadata_lte = metadata_lte;
1773                 if (ret) {
1774                         FREE(metadata_rspec);
1775                         goto out_wimlib_free;
1776                 }
1777
1778                 if (!(metadata_lte->flags & WIM_RESHDR_FLAG_METADATA)) {
1779                         ERROR("Expected metadata resource, but found "
1780                               "non-metadata stream.");
1781                         ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1782                         goto out_wimlib_free;
1783                 }
1784
1785                 if (i == image) {
1786                         /* Metadata resource is for the image being extracted.
1787                          * Parse it and save the metadata in memory.  */
1788                         ret = read_metadata_resource(pwm, imd);
1789                         if (ret)
1790                                 goto out_wimlib_free;
1791                         imd->modified = 1;
1792                 } else {
1793                         /* Metadata resource is not for the image being
1794                          * extracted.  Skip over it.  */
1795                         ret = skip_wim_stream(metadata_lte);
1796                         if (ret)
1797                                 goto out_wimlib_free;
1798                 }
1799         }
1800         /* Extract the image.  */
1801         extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
1802         ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
1803         /* Clean up and return.  */
1804 out_wimlib_free:
1805         wimlib_free(pwm);
1806         return ret;
1807 }
1808
1809
1810 WIMLIBAPI int
1811 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
1812                                const tchar *target, int extract_flags)
1813 {
1814         return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
1815                                                             image_num_or_name,
1816                                                             target,
1817                                                             extract_flags,
1818                                                             NULL,
1819                                                             NULL);
1820 }
1821
1822 WIMLIBAPI int
1823 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1824                      int extract_flags)
1825 {
1826         if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1827                 return WIMLIB_ERR_INVALID_PARAM;
1828         return do_wimlib_extract_image(wim, image, target, extract_flags);
1829 }