f9c4ee9894d48445e8392a63251fca9f64b844b5
[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 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 #ifdef HAVE_CONFIG_H
28 #  include "config.h"
29 #endif
30
31 #ifdef __WIN32__
32 #  include "wimlib/win32_common.h" /* For GetFullPathName() */
33 #endif
34
35 #include "wimlib/apply.h"
36 #include "wimlib/dentry.h"
37 #include "wimlib/encoding.h"
38 #include "wimlib/endianness.h"
39 #include "wimlib/error.h"
40 #include "wimlib/lookup_table.h"
41 #include "wimlib/paths.h"
42 #include "wimlib/resource.h"
43 #include "wimlib/swm.h"
44 #ifdef __WIN32__
45 #  include "wimlib/win32.h" /* for realpath() equivalent */
46 #endif
47 #include "wimlib/xml.h"
48
49 #include <errno.h>
50 #include <limits.h>
51 #ifdef WITH_NTFS_3G
52 #  include <ntfs-3g/volume.h> /* for ntfs_mount(), ntfs_umount() */
53 #endif
54 #include <stdlib.h>
55 #include <sys/stat.h>
56 #include <unistd.h>
57
58 #define MAX_EXTRACT_LONG_PATH_WARNINGS 5
59
60 static int
61 do_apply_op(struct wim_dentry *dentry, struct apply_args *args,
62             int (*apply_dentry_func)(const tchar *, size_t,
63                                      struct wim_dentry *, struct apply_args *))
64 {
65         tchar *p;
66         size_t extraction_path_nchars;
67         struct wim_dentry *d;
68         LIST_HEAD(ancestor_list);
69         const tchar *target;
70         size_t target_nchars;
71
72 #ifdef __WIN32__
73         if (args->target_lowlevel_path) {
74                 target = args->target_lowlevel_path;
75                 target_nchars = args->target_lowlevel_path_nchars;
76         } else
77 #endif
78         {
79                 target = args->target;
80                 target_nchars = args->target_nchars;
81         }
82
83         extraction_path_nchars = target_nchars;
84
85         for (d = dentry; d != args->extract_root; d = d->parent) {
86                 if (d->not_extracted)
87                         return 0;
88                 extraction_path_nchars += d->extraction_name_nchars + 1;
89                 list_add(&d->tmp_list, &ancestor_list);
90         }
91
92         tchar extraction_path[extraction_path_nchars + 1];
93         p = tmempcpy(extraction_path, target, target_nchars);
94
95
96         list_for_each_entry(d, &ancestor_list, tmp_list) {
97                 *p++ = OS_PREFERRED_PATH_SEPARATOR;
98                 p = tmempcpy(p, d->extraction_name, d->extraction_name_nchars);
99         }
100         *p = T('\0');
101
102 #ifdef __WIN32__
103         /* Warn the user if the path exceeds MAX_PATH */
104
105         /* + 1 for '\0', -4 for \\?\.  */
106         if (extraction_path_nchars + 1 - 4 > MAX_PATH) {
107                 if (dentry->needs_extraction &&
108                     args->num_long_paths < MAX_EXTRACT_LONG_PATH_WARNINGS)
109                 {
110                         WARNING("Path \"%ls\" exceeds MAX_PATH and will not be accessible "
111                                 "to most Windows software", extraction_path);
112                         if (++args->num_long_paths == MAX_EXTRACT_LONG_PATH_WARNINGS)
113                                 WARNING("Suppressing further warnings about long paths");
114                 }
115         }
116 #endif
117         return (*apply_dentry_func)(extraction_path, extraction_path_nchars,
118                                     dentry, args);
119 }
120
121
122 /* Extracts a file, directory, or symbolic link from the WIM archive. */
123 static int
124 apply_dentry_normal(struct wim_dentry *dentry, void *arg)
125 {
126 #ifdef __WIN32__
127         return do_apply_op(dentry, arg, win32_do_apply_dentry);
128 #else
129         return do_apply_op(dentry, arg, unix_do_apply_dentry);
130 #endif
131 }
132
133
134 /* Apply timestamps to an extracted file or directory */
135 static int
136 apply_dentry_timestamps_normal(struct wim_dentry *dentry, void *arg)
137 {
138 #ifdef __WIN32__
139         return do_apply_op(dentry, arg, win32_do_apply_dentry_timestamps);
140 #else
141         return do_apply_op(dentry, arg, unix_do_apply_dentry_timestamps);
142 #endif
143 }
144
145 static bool
146 dentry_is_dot_or_dotdot(const struct wim_dentry *dentry)
147 {
148         const utf16lechar *file_name = dentry->file_name;
149         return file_name != NULL &&
150                 file_name[0] == cpu_to_le16('.') &&
151                 (file_name[1] == cpu_to_le16('\0') ||
152                  (file_name[1] == cpu_to_le16('.') &&
153                   file_name[2] == cpu_to_le16('\0')));
154 }
155
156 /* Extract a dentry if it hasn't already been extracted and either
157  * WIMLIB_EXTRACT_FLAG_NO_STREAMS is not specified, or the dentry is a directory
158  * and/or has no unnamed stream. */
159 static int
160 maybe_apply_dentry(struct wim_dentry *dentry, void *arg)
161 {
162         struct apply_args *args = arg;
163         int ret;
164
165         if (!dentry->needs_extraction)
166                 return 0;
167
168         if (args->extract_flags & WIMLIB_EXTRACT_FLAG_NO_STREAMS &&
169             !dentry_is_directory(dentry) &&
170             inode_unnamed_lte_resolved(dentry->d_inode) != NULL)
171                 return 0;
172
173         if ((args->extract_flags & WIMLIB_EXTRACT_FLAG_VERBOSE) &&
174              args->progress_func) {
175                 ret = calculate_dentry_full_path(dentry);
176                 if (ret)
177                         return ret;
178                 args->progress.extract.cur_path = dentry->_full_path;
179                 args->progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY,
180                                     &args->progress);
181         }
182         ret = args->apply_dentry(dentry, args);
183         if (ret == 0)
184                 dentry->needs_extraction = 0;
185         return ret;
186 }
187
188 static void
189 calculate_bytes_to_extract(struct list_head *stream_list,
190                            int extract_flags,
191                            union wimlib_progress_info *progress)
192 {
193         struct wim_lookup_table_entry *lte;
194         u64 total_bytes = 0;
195         u64 num_streams = 0;
196
197         /* For each stream to be extracted... */
198         list_for_each_entry(lte, stream_list, extraction_list) {
199                 if (extract_flags &
200                     (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK))
201                 {
202                         /* In the symlink or hard link extraction mode, each
203                          * stream will be extracted one time regardless of how
204                          * many dentries share the stream. */
205                         wimlib_assert(!(extract_flags & WIMLIB_EXTRACT_FLAG_NTFS));
206                         if (!lte->extracted_file) {
207                                 num_streams++;
208                                 total_bytes += wim_resource_size(lte);
209                         }
210                 } else {
211                         num_streams += lte->out_refcnt;
212                         total_bytes += lte->out_refcnt * wim_resource_size(lte);
213                 }
214         }
215         progress->extract.num_streams = num_streams;
216         progress->extract.total_bytes = total_bytes;
217         progress->extract.completed_bytes = 0;
218 }
219
220 static void
221 maybe_add_stream_for_extraction(struct wim_lookup_table_entry *lte,
222                                 struct list_head *stream_list)
223 {
224         if (++lte->out_refcnt == 1) {
225                 INIT_LIST_HEAD(&lte->lte_dentry_list);
226                 list_add_tail(&lte->extraction_list, stream_list);
227         }
228 }
229
230 struct find_streams_ctx {
231         struct list_head stream_list;
232         int extract_flags;
233 };
234
235 static int
236 dentry_find_streams_to_extract(struct wim_dentry *dentry, void *_ctx)
237 {
238         struct find_streams_ctx *ctx = _ctx;
239         struct wim_inode *inode = dentry->d_inode;
240         struct wim_lookup_table_entry *lte;
241         bool dentry_added = false;
242         struct list_head *stream_list = &ctx->stream_list;
243         int extract_flags = ctx->extract_flags;
244
245         if (!dentry->needs_extraction)
246                 return 0;
247
248         lte = inode_unnamed_lte_resolved(inode);
249         if (lte) {
250                 if (!inode->i_visited)
251                         maybe_add_stream_for_extraction(lte, stream_list);
252                 list_add_tail(&dentry->extraction_stream_list, &lte->lte_dentry_list);
253                 dentry_added = true;
254         }
255
256         /* Determine whether to include alternate data stream entries or not.
257          *
258          * UNIX:  Include them if extracting using NTFS-3g.
259          *
260          * Windows: Include them undconditionally, although if the filesystem is
261          * not NTFS we won't actually be able to extract them. */
262 #if defined(WITH_NTFS_3G)
263         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
264 #elif defined(__WIN32__)
265         if (1)
266 #else
267         if (0)
268 #endif
269         {
270                 for (unsigned i = 0; i < inode->i_num_ads; i++) {
271                         if (inode->i_ads_entries[i].stream_name_nbytes != 0) {
272                                 lte = inode->i_ads_entries[i].lte;
273                                 if (lte) {
274                                         if (!inode->i_visited) {
275                                                 maybe_add_stream_for_extraction(lte,
276                                                                                 stream_list);
277                                         }
278                                         if (!dentry_added) {
279                                                 list_add_tail(&dentry->extraction_stream_list,
280                                                               &lte->lte_dentry_list);
281                                                 dentry_added = true;
282                                         }
283                                 }
284                         }
285                 }
286         }
287         inode->i_visited = 1;
288         return 0;
289 }
290
291 static int
292 dentry_resolve_and_zero_lte_refcnt(struct wim_dentry *dentry, void *_lookup_table)
293 {
294         struct wim_inode *inode = dentry->d_inode;
295         struct wim_lookup_table *lookup_table = _lookup_table;
296         struct wim_lookup_table_entry *lte;
297         int ret;
298
299         ret = inode_resolve_ltes(inode, lookup_table);
300         if (ret)
301                 return ret;
302         for (unsigned i = 0; i <= inode->i_num_ads; i++) {
303                 lte = inode_stream_lte_resolved(inode, i);
304                 if (lte)
305                         lte->out_refcnt = 0;
306         }
307         return 0;
308 }
309
310 static int
311 find_streams_for_extraction(struct wim_dentry *root,
312                             struct list_head *stream_list,
313                             struct wim_lookup_table *lookup_table,
314                             int extract_flags)
315 {
316         struct find_streams_ctx ctx;
317         int ret;
318
319         INIT_LIST_HEAD(&ctx.stream_list);
320         ctx.extract_flags = extract_flags;
321         ret = for_dentry_in_tree(root, dentry_resolve_and_zero_lte_refcnt, lookup_table);
322         if (ret)
323                 return ret;
324         for_dentry_in_tree(root, dentry_find_streams_to_extract, &ctx);
325         list_transfer(&ctx.stream_list, stream_list);
326         return 0;
327 }
328
329 struct apply_operations {
330         int (*apply_dentry)(struct wim_dentry *dentry, void *arg);
331         int (*apply_dentry_timestamps)(struct wim_dentry *dentry, void *arg);
332 };
333
334 static const struct apply_operations normal_apply_operations = {
335         .apply_dentry = apply_dentry_normal,
336         .apply_dentry_timestamps = apply_dentry_timestamps_normal,
337 };
338
339 #ifdef WITH_NTFS_3G
340 static const struct apply_operations ntfs_apply_operations = {
341         .apply_dentry = apply_dentry_ntfs,
342         .apply_dentry_timestamps = apply_dentry_timestamps_ntfs,
343 };
344 #endif
345
346 static int
347 apply_stream_list(struct list_head *stream_list,
348                   struct apply_args *args,
349                   const struct apply_operations *ops,
350                   wimlib_progress_func_t progress_func)
351 {
352         uint64_t bytes_per_progress = args->progress.extract.total_bytes / 100;
353         uint64_t next_progress = bytes_per_progress;
354         struct wim_lookup_table_entry *lte;
355         struct wim_dentry *dentry;
356         int ret;
357
358         /* This complicated loop is essentially looping through the dentries,
359          * although dentries may be visited more than once (if a dentry contains
360          * two different nonempty streams) or not at all (if a dentry contains
361          * no non-empty streams).
362          *
363          * The outer loop is over the distinct streams to be extracted so that
364          * sequential reading of the WIM can be implemented. */
365
366         /* For each distinct stream to be extracted */
367         list_for_each_entry(lte, stream_list, extraction_list) {
368                 /* For each dentry to be extracted that is a name for an inode
369                  * containing the stream */
370                 list_for_each_entry(dentry, &lte->lte_dentry_list, extraction_stream_list) {
371                         /* Extract the dentry if it was not already
372                          * extracted */
373                         ret = maybe_apply_dentry(dentry, args);
374                         if (ret)
375                                 return ret;
376                         if (progress_func &&
377                             args->progress.extract.completed_bytes >= next_progress)
378                         {
379                                 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS,
380                                               &args->progress);
381                                 if (args->progress.extract.completed_bytes >=
382                                     args->progress.extract.total_bytes)
383                                 {
384                                         next_progress = ~0ULL;
385                                 } else {
386                                         next_progress =
387                                                 min (args->progress.extract.completed_bytes +
388                                                      bytes_per_progress,
389                                                      args->progress.extract.total_bytes);
390                                 }
391                         }
392                 }
393         }
394         return 0;
395 }
396
397 static int
398 sort_stream_list_by_wim_position(struct list_head *stream_list)
399 {
400         struct list_head *cur;
401         size_t num_streams;
402         struct wim_lookup_table_entry **array;
403         size_t i;
404         size_t array_size;
405
406         num_streams = 0;
407         list_for_each(cur, stream_list)
408                 num_streams++;
409         array_size = num_streams * sizeof(array[0]);
410         array = MALLOC(array_size);
411         if (!array) {
412                 ERROR("Failed to allocate %zu bytes to sort stream entries",
413                       array_size);
414                 return WIMLIB_ERR_NOMEM;
415         }
416         cur = stream_list->next;
417         for (i = 0; i < num_streams; i++) {
418                 array[i] = container_of(cur, struct wim_lookup_table_entry, extraction_list);
419                 cur = cur->next;
420         }
421
422         qsort(array, num_streams, sizeof(array[0]), cmp_streams_by_wim_position);
423
424         INIT_LIST_HEAD(stream_list);
425         for (i = 0; i < num_streams; i++)
426                 list_add_tail(&array[i]->extraction_list, stream_list);
427         FREE(array);
428         return 0;
429 }
430
431 /*
432  * Extract a dentry to standard output.
433  *
434  * This obviously doesn't make sense in all cases.  We return an error if the
435  * dentry does not correspond to a regular file.  Otherwise we extract the
436  * unnamed data stream only.
437  */
438 static int
439 extract_dentry_to_stdout(struct wim_dentry *dentry)
440 {
441         int ret = 0;
442         if (!dentry_is_regular_file(dentry)) {
443                 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
444                       "extracted to standard output", dentry->_full_path);
445                 ret = WIMLIB_ERR_NOT_A_REGULAR_FILE;
446         } else {
447                 struct wim_lookup_table_entry *lte;
448
449                 lte = inode_unnamed_lte_resolved(dentry->d_inode);
450                 if (lte) {
451                         ret = extract_wim_resource_to_fd(lte, STDOUT_FILENO,
452                                                          wim_resource_size(lte));
453                 }
454         }
455         return ret;
456 }
457
458 #ifdef __WIN32__
459 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
460 #else
461 static const utf16lechar replacement_char = cpu_to_le16('?');
462 #endif
463
464 static bool
465 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
466 {
467         size_t i;
468
469         if (num_chars == 0)
470                 return true;
471         for (i = 0; i < num_chars; i++) {
472                 switch (name[i]) {
473         #ifdef __WIN32__
474                 case cpu_to_le16('\\'):
475                 case cpu_to_le16(':'):
476                 case cpu_to_le16('*'):
477                 case cpu_to_le16('?'):
478                 case cpu_to_le16('"'):
479                 case cpu_to_le16('<'):
480                 case cpu_to_le16('>'):
481                 case cpu_to_le16('|'):
482         #endif
483                 case cpu_to_le16('/'):
484                 case cpu_to_le16('\0'):
485                         if (fix)
486                                 name[i] = replacement_char;
487                         else
488                                 return false;
489                 }
490         }
491
492 #ifdef __WIN32__
493         if (name[num_chars - 1] == cpu_to_le16(' ') ||
494             name[num_chars - 1] == cpu_to_le16('.'))
495         {
496                 if (fix)
497                         name[num_chars - 1] = replacement_char;
498                 else
499                         return false;
500         }
501 #endif
502         return true;
503 }
504
505 /*
506  * dentry_calculate_extraction_path-
507  *
508  * Calculate the actual filename component at which a WIM dentry will be
509  * extracted, handling invalid filenames "properly".
510  *
511  * dentry->extraction_name usually will be set the same as dentry->file_name (on
512  * UNIX, converted into the platform's multibyte encoding).  However, if the
513  * file name contains characters that are not valid on the current platform or
514  * has some other format that is not valid, leave dentry->extraction_name as
515  * NULL and clear dentry->needs_extraction to indicate that this dentry should
516  * not be extracted, unless the appropriate flag
517  * WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES is set in the extract flags, in
518  * which case a substitute filename will be created and set instead.
519  *
520  * Conflicts with case-insensitive names on Windows are handled similarly; see
521  * below.
522  */
523 static int
524 dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args)
525 {
526         struct apply_args *args = _args;
527         int ret;
528
529         dentry->needs_extraction = 1;
530
531         if (dentry == args->extract_root)
532                 return 0;
533
534         if (dentry_is_dot_or_dotdot(dentry)) {
535                 /* WIM files shouldn't contain . or .. entries.  But if they are
536                  * there, don't attempt to extract them. */
537                 WARNING("Skipping extraction of unexpected . or .. file \"%"TS"\"",
538                         dentry_full_path(dentry));
539                 goto skip_dentry;
540         }
541
542 #ifdef __WIN32__
543         struct wim_dentry *other;
544         list_for_each_entry(other, &dentry->case_insensitive_conflict_list,
545                             case_insensitive_conflict_list)
546         {
547                 if (other->needs_extraction) {
548                         if (args->extract_flags & WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS)
549                         {
550                                 WARNING("\"%"TS"\" has the same case-insensitive "
551                                         "name as \"%"TS"\"; extracting dummy name instead",
552                                         dentry_full_path(dentry),
553                                         dentry_full_path(other));
554                                 goto out_replace;
555                         } else {
556                                 WARNING("Not extracting \"%"TS"\": has same case-insensitive "
557                                         "name as \"%"TS"\"",
558                                         dentry_full_path(dentry),
559                                         dentry_full_path(other));
560                                 goto skip_dentry;
561                         }
562                 }
563         }
564 #endif
565
566         if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
567 #ifdef __WIN32__
568                 dentry->extraction_name = dentry->file_name;
569                 dentry->extraction_name_nchars = dentry->file_name_nbytes / 2;
570                 return 0;
571 #else
572                 return utf16le_to_tstr(dentry->file_name,
573                                        dentry->file_name_nbytes,
574                                        &dentry->extraction_name,
575                                        &dentry->extraction_name_nchars);
576 #endif
577         } else {
578                 if (args->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
579                 {
580                         WARNING("\"%"TS"\" has an invalid filename "
581                                 "that is not supported on this platform; "
582                                 "extracting dummy name instead",
583                                 dentry_full_path(dentry));
584                         goto out_replace;
585                 } else {
586                         WARNING("Not extracting \"%"TS"\": has an invalid filename "
587                                 "that is not supported on this platform",
588                                 dentry_full_path(dentry));
589                         goto skip_dentry;
590                 }
591         }
592
593 out_replace:
594         {
595                 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
596
597                 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
598                 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
599
600                 tchar *tchar_name;
601                 size_t tchar_nchars;
602         #ifdef __WIN32__
603                 tchar_name = utf16_name_copy;
604                 tchar_nchars = dentry->file_name_nbytes / 2;
605         #else
606                 ret = utf16le_to_tstr(utf16_name_copy,
607                                       dentry->file_name_nbytes,
608                                       &tchar_name, &tchar_nchars);
609                 if (ret)
610                         return ret;
611         #endif
612                 size_t fixed_name_num_chars = tchar_nchars;
613                 tchar fixed_name[tchar_nchars + 50];
614
615                 tmemcpy(fixed_name, tchar_name, tchar_nchars);
616                 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
617                                                  T(" (invalid filename #%lu)"),
618                                                  ++args->invalid_sequence);
619         #ifndef __WIN32__
620                 FREE(tchar_name);
621         #endif
622                 dentry->extraction_name = memdup(fixed_name, 2 * fixed_name_num_chars + 2);
623                 if (!dentry->extraction_name)
624                         return WIMLIB_ERR_NOMEM;
625                 dentry->extraction_name_nchars = fixed_name_num_chars;
626         }
627         return 0;
628 skip_dentry:
629         dentry->needs_extraction = 0;
630         dentry->not_extracted = 1;
631         return 0;
632 }
633
634 static int
635 dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore)
636 {
637         dentry->needs_extraction = 0;
638         dentry->not_extracted = 0;
639         dentry->is_win32_name = 0;
640         dentry->d_inode->i_visited = 0;
641         dentry->d_inode->i_dos_name_extracted = 0;
642         FREE(dentry->d_inode->i_extracted_file);
643         dentry->d_inode->i_extracted_file = NULL;
644         if ((void*)dentry->extraction_name != (void*)dentry->file_name)
645                 FREE(dentry->extraction_name);
646         dentry->extraction_name = NULL;
647         return 0;
648 }
649
650 #define WINDOWS_NT_MAX_PATH 32768
651
652 /*
653  * extract_tree - Extract a file or directory tree from the currently selected
654  *                WIM image.
655  *
656  * @wim:        WIMStruct for the WIM file, with the desired image selected
657  *              (as wim->current_image).
658  * @wim_source_path:
659  *              "Canonical" (i.e. no leading or trailing slashes, path
660  *              separators forwald slashes) path inside the WIM image to
661  *              extract.  An empty string means the full image.
662  * @target:
663  *              Filesystem path to extract the file or directory tree to.
664  *
665  * @extract_flags:
666  *              WIMLIB_EXTRACT_FLAG_*.  Also, the private flag
667  *              WIMLIB_EXTRACT_FLAG_MULTI_IMAGE will be set if this is being
668  *              called through wimlib_extract_image() with WIMLIB_ALL_IMAGES as
669  *              the image.
670  *
671  * @progress_func:
672  *              If non-NULL, progress function for the extraction.  The messages
673  *              we may in this function are:
674  *
675  *              WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN or
676  *                      WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN;
677  *              WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN;
678  *              WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END;
679  *              WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY;
680  *              WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS;
681  *              WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS;
682  *              WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END or
683  *                      WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END.
684  *
685  * Returns 0 on success; nonzero on failure.
686  */
687 static int
688 extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target,
689              int extract_flags, wimlib_progress_func_t progress_func)
690 {
691         int ret;
692         struct list_head stream_list;
693         struct apply_args args;
694         const struct apply_operations *ops;
695         struct wim_dentry *root;
696
697         memset(&args, 0, sizeof(args));
698
699
700         args.w                      = wim;
701         args.target                 = target;
702         args.target_nchars          = tstrlen(target);
703         args.extract_flags          = extract_flags;
704         args.progress_func          = progress_func;
705
706 #ifdef __WIN32__
707         /* Work around defective behavior in Windows where paths longer than 260
708          * characters are not supported by default; instead they need to be
709          * turned into absolute paths and prefixed with "\\?\".  */
710         args.target_lowlevel_path = MALLOC(WINDOWS_NT_MAX_PATH * sizeof(wchar_t));
711         if (!args.target_lowlevel_path)
712         {
713                 ret = WIMLIB_ERR_NOMEM;
714                 goto out;
715         }
716         args.target_lowlevel_path_nchars =
717                 GetFullPathName(args.target, WINDOWS_NT_MAX_PATH - 4,
718                                 &args.target_lowlevel_path[4], NULL);
719
720         if (args.target_lowlevel_path_nchars == 0 ||
721             args.target_lowlevel_path_nchars >= WINDOWS_NT_MAX_PATH - 4)
722         {
723                 WARNING("Can't get full path name for \"%ls\"", args.target);
724                 FREE(args.target_lowlevel_path);
725                 args.target_lowlevel_path = NULL;
726         } else {
727                 wmemcpy(args.target_lowlevel_path, L"\\\\?\\", 4);
728                 args.target_lowlevel_path_nchars += 4;
729         }
730 #endif
731
732         if (progress_func) {
733                 args.progress.extract.wimfile_name = wim->filename;
734                 args.progress.extract.image = wim->current_image;
735                 args.progress.extract.extract_flags = (extract_flags &
736                                                        WIMLIB_EXTRACT_MASK_PUBLIC);
737                 args.progress.extract.image_name = wimlib_get_image_name(wim,
738                                                                          wim->current_image);
739                 args.progress.extract.extract_root_wim_source_path = wim_source_path;
740                 args.progress.extract.target = target;
741         }
742
743 #ifdef WITH_NTFS_3G
744         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
745                 args.vol = ntfs_mount(target, 0);
746                 if (!args.vol) {
747                         ERROR_WITH_ERRNO("Failed to mount NTFS volume `%"TS"'",
748                                          target);
749                         ret = WIMLIB_ERR_NTFS_3G;
750                         goto out_free_target_lowlevel_path;
751                 }
752                 ops = &ntfs_apply_operations;
753         } else
754 #endif
755                 ops = &normal_apply_operations;
756
757         root = get_dentry(wim, wim_source_path);
758         if (!root) {
759                 ERROR("Path \"%"TS"\" does not exist in WIM image %d",
760                       wim_source_path, wim->current_image);
761                 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
762                 goto out_ntfs_umount;
763         }
764         args.extract_root = root;
765
766         /* Calculate the actual filename component of each extracted dentry, and
767          * in the process set the dentry->needs_extraction flag on dentries that
768          * will be extracted. */
769         ret = for_dentry_in_tree(root, dentry_calculate_extraction_path, &args);
770         if (ret)
771                 goto out_dentry_reset_needs_extraction;
772
773         /* Build a list of the streams that need to be extracted */
774         ret = find_streams_for_extraction(root,
775                                           &stream_list,
776                                           wim->lookup_table, extract_flags);
777         if (ret)
778                 goto out_dentry_reset_needs_extraction;
779
780         /* Calculate the number of bytes of data that will be extracted */
781         calculate_bytes_to_extract(&stream_list, extract_flags,
782                                    &args.progress);
783
784         if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
785                 ret = extract_dentry_to_stdout(root);
786                 goto out_dentry_reset_needs_extraction;
787         }
788
789         if (progress_func) {
790                 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN :
791                               WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN,
792                               &args.progress);
793         }
794
795         /* If a sequential extraction was specified, sort the streams to be
796          * extracted by their position in the WIM file, so that the WIM file can
797          * be read sequentially. */
798         if (extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) {
799                 ret = sort_stream_list_by_wim_position(&stream_list);
800                 if (ret != 0) {
801                         WARNING("Falling back to non-sequential extraction");
802                         extract_flags &= ~WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
803                 }
804         }
805
806         if (progress_func) {
807                 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
808                               &args.progress);
809         }
810
811         /* Make the directory structure and extract empty files */
812         args.extract_flags |= WIMLIB_EXTRACT_FLAG_NO_STREAMS;
813         args.apply_dentry = ops->apply_dentry;
814         ret = for_dentry_in_tree(root, maybe_apply_dentry, &args);
815         args.extract_flags &= ~WIMLIB_EXTRACT_FLAG_NO_STREAMS;
816         if (ret)
817                 goto out_dentry_reset_needs_extraction;
818
819         if (progress_func) {
820                 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
821                               &args.progress);
822         }
823
824         if (extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) {
825                 args.target_realpath = realpath(target, NULL);
826                 if (!args.target_realpath) {
827                         ret = WIMLIB_ERR_NOMEM;
828                         goto out_dentry_reset_needs_extraction;
829                 }
830                 args.target_realpath_len = tstrlen(args.target_realpath);
831         }
832
833         /* Extract non-empty files */
834         ret = apply_stream_list(&stream_list, &args, ops, progress_func);
835         if (ret)
836                 goto out_free_target_realpath;
837
838         if (progress_func) {
839                 progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS,
840                               &args.progress);
841         }
842
843         /* Apply timestamps */
844         ret = for_dentry_in_tree_depth(root,
845                                        ops->apply_dentry_timestamps, &args);
846         if (ret)
847                 goto out_free_target_realpath;
848
849         if (progress_func) {
850                 progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END :
851                               WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END,
852                               &args.progress);
853         }
854 out_free_target_realpath:
855         FREE(args.target_realpath);
856 out_dentry_reset_needs_extraction:
857         for_dentry_in_tree(root, dentry_reset_needs_extraction, NULL);
858 out_ntfs_umount:
859 #ifdef WITH_NTFS_3G
860         /* Unmount the NTFS volume */
861         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
862                 if (ntfs_umount(args.vol, FALSE) != 0) {
863                         ERROR_WITH_ERRNO("Failed to unmount NTFS volume `%"TS"'",
864                                          args.target);
865                         if (ret == 0)
866                                 ret = WIMLIB_ERR_NTFS_3G;
867                 }
868         }
869 #endif
870 out_free_target_lowlevel_path:
871 #ifdef __WIN32__
872         FREE(args.target_lowlevel_path);
873 #endif
874 out:
875         return ret;
876 }
877
878 /* Validates a single wimlib_extract_command, mostly checking to make sure the
879  * extract flags make sense. */
880 static int
881 check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags)
882 {
883         int extract_flags;
884         bool is_entire_image = (cmd->wim_source_path[0] == T('\0'));
885
886         /* Empty destination path? */
887         if (cmd->fs_dest_path[0] == T('\0'))
888                 return WIMLIB_ERR_INVALID_PARAM;
889
890         extract_flags = cmd->extract_flags;
891
892         /* Specified both symlink and hardlink modes? */
893         if ((extract_flags &
894              (WIMLIB_EXTRACT_FLAG_SYMLINK |
895               WIMLIB_EXTRACT_FLAG_HARDLINK)) == (WIMLIB_EXTRACT_FLAG_SYMLINK |
896                                                  WIMLIB_EXTRACT_FLAG_HARDLINK))
897                 return WIMLIB_ERR_INVALID_PARAM;
898
899 #ifdef __WIN32__
900         /* Wanted UNIX data on Windows? */
901         if (extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
902                 ERROR("Extracting UNIX data is not supported on Windows");
903                 return WIMLIB_ERR_INVALID_PARAM;
904         }
905         /* Wanted linked extraction on Windows?  (XXX This is possible, just not
906          * implemented yet.) */
907         if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
908                              WIMLIB_EXTRACT_FLAG_HARDLINK))
909         {
910                 ERROR("Linked extraction modes are not supported on Windows");
911                 return WIMLIB_ERR_INVALID_PARAM;
912         }
913 #endif
914
915         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
916                 /* NTFS-3g extraction mode requested */
917 #ifdef WITH_NTFS_3G
918                 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
919                                       WIMLIB_EXTRACT_FLAG_HARDLINK))) {
920                         ERROR("Cannot specify symlink or hardlink flags when applying\n"
921                               "        directly to a NTFS volume");
922                         return WIMLIB_ERR_INVALID_PARAM;
923                 }
924                 if (!is_entire_image &&
925                     (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS))
926                 {
927                         ERROR("When applying directly to a NTFS volume you can "
928                               "only extract a full image, not part of one");
929                         return WIMLIB_ERR_INVALID_PARAM;
930                 }
931                 if (extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
932                         ERROR("Cannot restore UNIX-specific data in "
933                               "the NTFS extraction mode");
934                         return WIMLIB_ERR_INVALID_PARAM;
935                 }
936 #else
937                 ERROR("wimlib was compiled without support for NTFS-3g, so");
938                 ERROR("we cannot apply a WIM image directly to a NTFS volume");
939                 return WIMLIB_ERR_UNSUPPORTED;
940 #endif
941         }
942
943         if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
944                               WIMLIB_EXTRACT_FLAG_NORPFIX)) ==
945                 (WIMLIB_EXTRACT_FLAG_RPFIX | WIMLIB_EXTRACT_FLAG_NORPFIX))
946         {
947                 ERROR("Cannot specify RPFIX and NORPFIX flags at the same time!");
948                 return WIMLIB_ERR_INVALID_PARAM;
949         }
950
951         if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
952                               WIMLIB_EXTRACT_FLAG_NORPFIX)) == 0)
953         {
954                 /* Do reparse point fixups by default if the WIM header says
955                  * they are enabled and we are extracting a full image. */
956                 if ((wim_header_flags & WIM_HDR_FLAG_RP_FIX) && is_entire_image)
957                         extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
958         }
959
960         if (!is_entire_image && (extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX)) {
961                 ERROR("Cannot specify --rpfix when not extracting entire image");
962                 return WIMLIB_ERR_INVALID_PARAM;
963         }
964
965         cmd->extract_flags = extract_flags;
966         return 0;
967 }
968
969
970 /* Internal function to execute extraction commands for a WIM image. */
971 static int
972 do_wimlib_extract_files(WIMStruct *wim,
973                         int image,
974                         struct wimlib_extract_command *cmds,
975                         size_t num_cmds,
976                         wimlib_progress_func_t progress_func)
977 {
978         int ret;
979         bool found_link_cmd = false;
980         bool found_nolink_cmd = false;
981
982         /* Select the image from which we are extracting files */
983         ret = select_wim_image(wim, image);
984         if (ret)
985                 return ret;
986
987         /* Make sure there are no streams in the WIM that have not been
988          * checksummed yet. */
989         ret = wim_checksum_unhashed_streams(wim);
990         if (ret)
991                 return ret;
992
993         /* Check for problems with the extraction commands */
994         for (size_t i = 0; i < num_cmds; i++) {
995                 ret = check_extract_command(&cmds[i], wim->hdr.flags);
996                 if (ret)
997                         return ret;
998                 if (cmds[i].extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
999                                              WIMLIB_EXTRACT_FLAG_HARDLINK)) {
1000                         found_link_cmd = true;
1001                 } else {
1002                         found_nolink_cmd = true;
1003                 }
1004                 if (found_link_cmd && found_nolink_cmd) {
1005                         ERROR("Symlink or hardlink extraction mode must "
1006                               "be set on all extraction commands");
1007                         return WIMLIB_ERR_INVALID_PARAM;
1008                 }
1009         }
1010
1011         /* Execute the extraction commands */
1012         for (size_t i = 0; i < num_cmds; i++) {
1013                 ret = extract_tree(wim,
1014                                    cmds[i].wim_source_path,
1015                                    cmds[i].fs_dest_path,
1016                                    cmds[i].extract_flags,
1017                                    progress_func);
1018                 if (ret)
1019                         return ret;
1020         }
1021         return 0;
1022 }
1023
1024 /* Extract files or directories from a WIM image. */
1025 WIMLIBAPI int
1026 wimlib_extract_files(WIMStruct *wim,
1027                      int image,
1028                      const struct wimlib_extract_command *cmds,
1029                      size_t num_cmds,
1030                      int default_extract_flags,
1031                      WIMStruct **additional_swms,
1032                      unsigned num_additional_swms,
1033                      wimlib_progress_func_t progress_func)
1034 {
1035         int ret;
1036         struct wimlib_extract_command *cmds_copy;
1037         int all_flags = 0;
1038
1039         default_extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
1040
1041         ret = verify_swm_set(wim, additional_swms, num_additional_swms);
1042         if (ret)
1043                 goto out;
1044
1045         if (num_cmds == 0)
1046                 goto out;
1047
1048         if (num_additional_swms)
1049                 merge_lookup_tables(wim, additional_swms, num_additional_swms);
1050
1051         cmds_copy = CALLOC(num_cmds, sizeof(cmds[0]));
1052         if (!cmds_copy) {
1053                 ret = WIMLIB_ERR_NOMEM;
1054                 goto out_restore_lookup_table;
1055         }
1056
1057         for (size_t i = 0; i < num_cmds; i++) {
1058                 cmds_copy[i].extract_flags = (default_extract_flags |
1059                                                  cmds[i].extract_flags)
1060                                                 & WIMLIB_EXTRACT_MASK_PUBLIC;
1061                 all_flags |= cmds_copy[i].extract_flags;
1062
1063                 cmds_copy[i].wim_source_path = canonicalize_wim_path(cmds[i].wim_source_path);
1064                 if (!cmds_copy[i].wim_source_path) {
1065                         ret = WIMLIB_ERR_NOMEM;
1066                         goto out_free_cmds_copy;
1067                 }
1068
1069                 cmds_copy[i].fs_dest_path = canonicalize_fs_path(cmds[i].fs_dest_path);
1070                 if (!cmds_copy[i].fs_dest_path) {
1071                         ret = WIMLIB_ERR_NOMEM;
1072                         goto out_free_cmds_copy;
1073                 }
1074
1075         }
1076         ret = do_wimlib_extract_files(wim, image,
1077                                       cmds_copy, num_cmds,
1078                                       progress_func);
1079
1080         if (all_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
1081                          WIMLIB_EXTRACT_FLAG_HARDLINK))
1082         {
1083                 for_lookup_table_entry(wim->lookup_table,
1084                                        lte_free_extracted_file, NULL);
1085         }
1086 out_free_cmds_copy:
1087         for (size_t i = 0; i < num_cmds; i++) {
1088                 FREE(cmds_copy[i].wim_source_path);
1089                 FREE(cmds_copy[i].fs_dest_path);
1090         }
1091         FREE(cmds_copy);
1092 out_restore_lookup_table:
1093         if (num_additional_swms)
1094                 unmerge_lookup_table(wim);
1095 out:
1096         return ret;
1097 }
1098
1099 /*
1100  * Extracts an image from a WIM file.
1101  *
1102  * @wim:                WIMStruct for the WIM file.
1103  *
1104  * @image:              Number of the single image to extract.
1105  *
1106  * @target:             Directory or NTFS volume to extract the image to.
1107  *
1108  * @extract_flags:      Bitwise or of WIMLIB_EXTRACT_FLAG_*.
1109  *
1110  * @progress_func:      If non-NULL, a progress function to be called
1111  *                      periodically.
1112  *
1113  * Returns 0 on success; nonzero on failure.
1114  */
1115 static int
1116 extract_single_image(WIMStruct *wim, int image,
1117                      const tchar *target, int extract_flags,
1118                      wimlib_progress_func_t progress_func)
1119 {
1120         int ret;
1121         tchar *target_copy = canonicalize_fs_path(target);
1122         if (!target_copy)
1123                 return WIMLIB_ERR_NOMEM;
1124         struct wimlib_extract_command cmd = {
1125                 .wim_source_path = T(""),
1126                 .fs_dest_path = target_copy,
1127                 .extract_flags = extract_flags,
1128         };
1129         ret = do_wimlib_extract_files(wim, image, &cmd, 1, progress_func);
1130         FREE(target_copy);
1131         return ret;
1132 }
1133
1134 static const tchar * const filename_forbidden_chars =
1135 T(
1136 #ifdef __WIN32__
1137 "<>:\"/\\|?*"
1138 #else
1139 "/"
1140 #endif
1141 );
1142
1143 /* This function checks if it is okay to use a WIM image's name as a directory
1144  * name.  */
1145 static bool
1146 image_name_ok_as_dir(const tchar *image_name)
1147 {
1148         return image_name && *image_name &&
1149                 !tstrpbrk(image_name, filename_forbidden_chars) &&
1150                 tstrcmp(image_name, T(".")) &&
1151                 tstrcmp(image_name, T(".."));
1152 }
1153
1154 /* Extracts all images from the WIM to the directory @target, with the images
1155  * placed in subdirectories named by their image names. */
1156 static int
1157 extract_all_images(WIMStruct *wim,
1158                    const tchar *target,
1159                    int extract_flags,
1160                    wimlib_progress_func_t progress_func)
1161 {
1162         size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
1163         size_t output_path_len = tstrlen(target);
1164         tchar buf[output_path_len + 1 + image_name_max_len + 1];
1165         int ret;
1166         int image;
1167         const tchar *image_name;
1168         struct stat stbuf;
1169
1170         if (tstat(target, &stbuf)) {
1171                 if (errno == ENOENT)
1172                 {
1173                         if (tmkdir(target, S_IRWXU | S_IRGRP | S_IXGRP |
1174                                            S_IROTH | S_IXOTH))
1175                         {
1176                                 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
1177                                 return WIMLIB_ERR_MKDIR;
1178                         }
1179                 } else {
1180                         ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target);
1181                         return WIMLIB_ERR_STAT;
1182                 }
1183         } else if (!S_ISDIR(stbuf.st_mode)) {
1184                 ERROR("\"%"TS"\" is not a directory", target);
1185                 return WIMLIB_ERR_NOTDIR;
1186         }
1187
1188         tmemcpy(buf, target, output_path_len);
1189         buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1190         for (image = 1; image <= wim->hdr.image_count; image++) {
1191                 image_name = wimlib_get_image_name(wim, image);
1192                 if (image_name_ok_as_dir(image_name)) {
1193                         tstrcpy(buf + output_path_len + 1, image_name);
1194                 } else {
1195                         /* Image name is empty or contains forbidden characters.
1196                          * Use image number instead. */
1197                         tsprintf(buf + output_path_len + 1, T("%d"), image);
1198                 }
1199                 ret = extract_single_image(wim, image, buf, extract_flags,
1200                                            progress_func);
1201                 if (ret)
1202                         return ret;
1203         }
1204         return 0;
1205 }
1206
1207 /* Extracts a single image or all images from a WIM file to a directory or NTFS
1208  * volume. */
1209 WIMLIBAPI int
1210 wimlib_extract_image(WIMStruct *wim,
1211                      int image,
1212                      const tchar *target,
1213                      int extract_flags,
1214                      WIMStruct **additional_swms,
1215                      unsigned num_additional_swms,
1216                      wimlib_progress_func_t progress_func)
1217 {
1218         int ret;
1219
1220         extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
1221
1222         ret = verify_swm_set(wim, additional_swms, num_additional_swms);
1223         if (ret)
1224                 return ret;
1225
1226         if (num_additional_swms)
1227                 merge_lookup_tables(wim, additional_swms, num_additional_swms);
1228
1229         if (image == WIMLIB_ALL_IMAGES) {
1230                 ret = extract_all_images(wim, target,
1231                                          extract_flags | WIMLIB_EXTRACT_FLAG_MULTI_IMAGE,
1232                                          progress_func);
1233         } else {
1234                 ret = extract_single_image(wim, image, target, extract_flags,
1235                                            progress_func);
1236         }
1237
1238         if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
1239                              WIMLIB_EXTRACT_FLAG_HARDLINK))
1240         {
1241                 for_lookup_table_entry(wim->lookup_table,
1242                                        lte_free_extracted_file,
1243                                        NULL);
1244         }
1245         if (num_additional_swms)
1246                 unmerge_lookup_table(wim);
1247         return ret;
1248 }