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