7effb4f395d4abab7eff5e7b80b09a38e9b30a18
[wimlib] / src / extract_image.c
1 /*
2  * extract_image.c
3  *
4  * Support for extracting WIM files.
5  */
6
7 /*
8  * Copyright (C) 2012, 2013 Eric Biggers
9  *
10  * This file is part of wimlib, a library for working with WIM files.
11  *
12  * wimlib is free software; you can redistribute it and/or modify it under the
13  * terms of the GNU General Public License as published by the Free
14  * Software Foundation; either version 3 of the License, or (at your option)
15  * any later version.
16  *
17  * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
18  * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
19  * A PARTICULAR PURPOSE. See the GNU General Public License for more
20  * details.
21  *
22  * You should have received a copy of the GNU General Public License
23  * along with wimlib; if not, see http://www.gnu.org/licenses/.
24  */
25
26 #include "config.h"
27
28 #include <dirent.h>
29
30 #ifdef __WIN32__
31 #  include "win32.h"
32 #else
33 #  ifdef HAVE_UTIME_H
34 #    include <utime.h>
35 #  endif
36 #  include "timestamp.h"
37 #  include <sys/time.h>
38 #endif
39
40 #include <errno.h>
41 #include <fcntl.h>
42 #include <stdlib.h>
43 #include <string.h>
44 #include <sys/stat.h>
45 #include <unistd.h>
46
47 #include "dentry.h"
48 #include "lookup_table.h"
49 #include "wimlib_internal.h"
50 #include "xml.h"
51
52 #ifdef WITH_NTFS_3G
53 #  include <ntfs-3g/volume.h>
54 #endif
55
56 #ifdef HAVE_ALLOCA_H
57 #  include <alloca.h>
58 #endif
59
60 #ifndef __WIN32__
61
62 /* Returns the number of components of @path.  */
63 static unsigned
64 get_num_path_components(const char *path)
65 {
66         unsigned num_components = 0;
67         while (*path) {
68                 while (*path == '/')
69                         path++;
70                 if (*path)
71                         num_components++;
72                 while (*path && *path != '/')
73                         path++;
74         }
75         return num_components;
76 }
77
78 static const char *
79 path_next_part(const char *path)
80 {
81         while (*path && *path != '/')
82                 path++;
83         while (*path && *path == '/')
84                 path++;
85         return path;
86 }
87
88 static int
89 extract_regular_file_linked(struct wim_dentry *dentry,
90                             const char *output_path,
91                             struct apply_args *args,
92                             struct wim_lookup_table_entry *lte)
93 {
94         /* This mode overrides the normal hard-link extraction and
95          * instead either symlinks or hardlinks *all* identical files in
96          * the WIM, even if they are in a different image (in the case
97          * of a multi-image extraction) */
98
99         if (args->extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) {
100                 if (link(lte->extracted_file, output_path) != 0) {
101                         ERROR_WITH_ERRNO("Failed to hard link "
102                                          "`%s' to `%s'",
103                                          output_path, lte->extracted_file);
104                         return WIMLIB_ERR_LINK;
105                 }
106         } else {
107                 int num_path_components;
108                 int num_output_dir_path_components;
109                 size_t extracted_file_len;
110                 char *p;
111                 const char *p2;
112                 size_t i;
113
114                 num_path_components = get_num_path_components(dentry->_full_path) - 1;
115                 num_output_dir_path_components = get_num_path_components(args->target);
116
117                 if (args->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE) {
118                         num_path_components++;
119                         num_output_dir_path_components--;
120                 }
121                 extracted_file_len = strlen(lte->extracted_file);
122
123                 char buf[extracted_file_len + 3 * num_path_components + 1];
124                 p = &buf[0];
125
126                 for (i = 0; i < num_path_components; i++) {
127                         *p++ = '.';
128                         *p++ = '.';
129                         *p++ = '/';
130                 }
131                 p2 = lte->extracted_file;
132                 while (*p2 == '/')
133                         p2++;
134                 while (num_output_dir_path_components > 0) {
135                         p2 = path_next_part(p2);
136                         num_output_dir_path_components--;
137                 }
138                 strcpy(p, p2);
139                 if (symlink(buf, output_path) != 0) {
140                         ERROR_WITH_ERRNO("Failed to symlink `%s' to `%s'",
141                                          buf, lte->extracted_file);
142                         return WIMLIB_ERR_LINK;
143                 }
144         }
145         return 0;
146 }
147
148 static int
149 symlink_apply_unix_data(const char *link,
150                         const struct wimlib_unix_data *unix_data)
151 {
152         if (lchown(link, unix_data->uid, unix_data->gid)) {
153                 if (errno == EPERM) {
154                         /* Ignore */
155                         WARNING_WITH_ERRNO("failed to set symlink UNIX owner/group");
156                 } else {
157                         ERROR_WITH_ERRNO("failed to set symlink UNIX owner/group");
158                         return WIMLIB_ERR_INVALID_DENTRY;
159                 }
160         }
161         return 0;
162 }
163
164 static int
165 fd_apply_unix_data(int fd, const struct wimlib_unix_data *unix_data)
166 {
167         if (fchown(fd, unix_data->uid, unix_data->gid)) {
168                 if (errno == EPERM) {
169                         WARNING_WITH_ERRNO("failed to set file UNIX owner/group");
170                         /* Ignore? */
171                 } else {
172                         ERROR_WITH_ERRNO("failed to set file UNIX owner/group");
173                         return WIMLIB_ERR_INVALID_DENTRY;
174                 }
175         }
176
177         if (fchmod(fd, unix_data->mode)) {
178                 if (errno == EPERM) {
179                         WARNING_WITH_ERRNO("failed to set UNIX file mode");
180                         /* Ignore? */
181                 } else {
182                         ERROR_WITH_ERRNO("failed to set UNIX file mode");
183                         return WIMLIB_ERR_INVALID_DENTRY;
184                 }
185         }
186         return 0;
187 }
188
189 static int
190 dir_apply_unix_data(const char *dir, const struct wimlib_unix_data *unix_data)
191 {
192         int dfd = open(dir, O_RDONLY);
193         int ret;
194         if (dfd >= 0) {
195                 ret = fd_apply_unix_data(dfd, unix_data);
196                 if (close(dfd)) {
197                         ERROR_WITH_ERRNO("can't close directory `%s'", dir);
198                         ret = WIMLIB_ERR_MKDIR;
199                 }
200         } else {
201                 ERROR_WITH_ERRNO("can't open directory `%s'", dir);
202                 ret = WIMLIB_ERR_MKDIR;
203         }
204         return ret;
205 }
206
207 static int
208 extract_regular_file_unlinked(struct wim_dentry *dentry,
209                               struct apply_args *args,
210                               const char *output_path,
211                               struct wim_lookup_table_entry *lte)
212 {
213         /* Normal mode of extraction.  Regular files and hard links are
214          * extracted in the way that they appear in the WIM. */
215
216         int out_fd;
217         int ret;
218         struct wim_inode *inode = dentry->d_inode;
219
220         if (!((args->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE)
221                 && (args->extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
222                                      WIMLIB_EXTRACT_FLAG_HARDLINK))))
223         {
224                 /* If the dentry is part of a hard link set of at least 2
225                  * dentries and one of the other dentries has already been
226                  * extracted, make a hard link to the file corresponding to this
227                  * already-extracted directory.  Otherwise, extract the file and
228                  * set the inode->i_extracted_file field so that other dentries
229                  * in the hard link group can link to it. */
230                 if (inode->i_nlink > 1) {
231                         if (inode->i_extracted_file) {
232                                 DEBUG("Extracting hard link `%s' => `%s'",
233                                       output_path, inode->i_extracted_file);
234                                 if (link(inode->i_extracted_file, output_path) != 0) {
235                                         ERROR_WITH_ERRNO("Failed to hard link "
236                                                          "`%s' to `%s'",
237                                                          output_path,
238                                                          inode->i_extracted_file);
239                                         return WIMLIB_ERR_LINK;
240                                 }
241                                 return 0;
242                         }
243                         FREE(inode->i_extracted_file);
244                         inode->i_extracted_file = STRDUP(output_path);
245                         if (!inode->i_extracted_file) {
246                                 ERROR("Failed to allocate memory for filename");
247                                 return WIMLIB_ERR_NOMEM;
248                         }
249                 }
250         }
251
252         /* Extract the contents of the file to @output_path. */
253
254         out_fd = open(output_path, O_WRONLY | O_CREAT | O_TRUNC, 0644);
255         if (out_fd == -1) {
256                 ERROR_WITH_ERRNO("Failed to open the file `%s' for writing",
257                                  output_path);
258                 return WIMLIB_ERR_OPEN;
259         }
260
261         if (!lte) {
262                 /* Empty file with no lookup table entry */
263                 DEBUG("Empty file `%s'.", output_path);
264                 ret = 0;
265                 goto out_extract_unix_data;
266         }
267
268         ret = extract_wim_resource_to_fd(lte, out_fd, wim_resource_size(lte));
269         if (ret) {
270                 ERROR("Failed to extract resource to `%s'", output_path);
271                 goto out;
272         }
273
274 out_extract_unix_data:
275         if (args->extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
276                 struct wimlib_unix_data unix_data;
277                 ret = inode_get_unix_data(inode, &unix_data, NULL);
278                 if (ret > 0)
279                         ;
280                 else if (ret < 0)
281                         ret = 0;
282                 else
283                         ret = fd_apply_unix_data(out_fd, &unix_data);
284                 if (ret)
285                         goto out;
286         }
287         if (lte)
288                 args->progress.extract.completed_bytes += wim_resource_size(lte);
289 out:
290         if (close(out_fd) != 0) {
291                 ERROR_WITH_ERRNO("Failed to close file `%s'", output_path);
292                 if (ret == 0)
293                         ret = WIMLIB_ERR_WRITE;
294         }
295         return ret;
296 }
297
298 static int
299 extract_regular_file(struct wim_dentry *dentry,
300                      struct apply_args *args,
301                      const char *output_path)
302 {
303         struct wim_lookup_table_entry *lte;
304         const struct wim_inode *inode = dentry->d_inode;
305
306         lte = inode_unnamed_lte_resolved(inode);
307
308         if (lte && (args->extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
309                                            WIMLIB_EXTRACT_FLAG_HARDLINK)))
310         {
311                 if (lte->extracted_file) {
312                         return extract_regular_file_linked(dentry, output_path, args, lte);
313                 } else {
314                         lte->extracted_file = STRDUP(output_path);
315                         if (!lte->extracted_file)
316                                 return WIMLIB_ERR_NOMEM;
317                 }
318         }
319         return extract_regular_file_unlinked(dentry, args, output_path, lte);
320 }
321
322 static int
323 extract_symlink(struct wim_dentry *dentry,
324                 struct apply_args *args,
325                 const char *output_path)
326 {
327         char target[4096 + args->target_realpath_len];
328         char *fixed_target;
329
330         ssize_t ret = inode_readlink(dentry->d_inode,
331                                      target + args->target_realpath_len,
332                                      sizeof(target) - args->target_realpath_len - 1,
333                                      args->w, false);
334         struct wim_lookup_table_entry *lte;
335
336         if (ret <= 0) {
337                 ERROR("Could not read the symbolic link from dentry `%s'",
338                       dentry->_full_path);
339                 return WIMLIB_ERR_INVALID_DENTRY;
340         }
341         target[args->target_realpath_len + ret] = '\0';
342         if (target[args->target_realpath_len] == '/' &&
343             args->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX)
344         {
345                 /* Fix absolute symbolic link target to point into the actual
346                  * extraction destination */
347                 memcpy(target, args->target_realpath,
348                        args->target_realpath_len);
349                 fixed_target = target;
350         } else {
351                 /* Keep same link target */
352                 fixed_target = target + args->target_realpath_len;
353         }
354         ret = symlink(fixed_target, output_path);
355         if (ret) {
356                 ERROR_WITH_ERRNO("Failed to symlink `%s' to `%s'",
357                                  output_path, fixed_target);
358                 return WIMLIB_ERR_LINK;
359         }
360         lte = inode_unnamed_lte_resolved(dentry->d_inode);
361         wimlib_assert(lte != NULL);
362         if (args->extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
363                 struct wimlib_unix_data unix_data;
364                 ret = inode_get_unix_data(dentry->d_inode, &unix_data, NULL);
365                 if (ret > 0)
366                         ;
367                 else if (ret < 0)
368                         ret = 0;
369                 else
370                         ret = symlink_apply_unix_data(output_path, &unix_data);
371                 if (ret)
372                         return ret;
373         }
374         args->progress.extract.completed_bytes += wim_resource_size(lte);
375         return 0;
376 }
377
378 #endif /* !__WIN32__ */
379
380 static int
381 extract_directory(struct wim_dentry *dentry,
382                   const tchar *output_path, bool is_root)
383 {
384         int ret;
385         struct stat stbuf;
386
387         ret = tstat(output_path, &stbuf);
388         if (ret == 0) {
389                 if (S_ISDIR(stbuf.st_mode)) {
390                         /*if (!is_root)*/
391                                 /*WARNING("`%s' already exists", output_path);*/
392                         goto dir_exists;
393                 } else {
394                         ERROR("`%"TS"' is not a directory", output_path);
395                         return WIMLIB_ERR_MKDIR;
396                 }
397         } else {
398                 if (errno != ENOENT) {
399                         ERROR_WITH_ERRNO("Failed to stat `%"TS"'", output_path);
400                         return WIMLIB_ERR_STAT;
401                 }
402         }
403
404         if (tmkdir(output_path, S_IRWXU | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH))
405         {
406                 ERROR_WITH_ERRNO("Cannot create directory `%"TS"'", output_path);
407                 return WIMLIB_ERR_MKDIR;
408         }
409 dir_exists:
410         ret = 0;
411 #ifndef __WIN32__
412         if (dentry) {
413                 struct wimlib_unix_data unix_data;
414                 ret = inode_get_unix_data(dentry->d_inode, &unix_data, NULL);
415                 if (ret > 0)
416                         ;
417                 else if (ret < 0)
418                         ret = 0;
419                 else
420                         ret = dir_apply_unix_data(output_path, &unix_data);
421         }
422 #endif
423         return ret;
424 }
425
426 #ifndef __WIN32__
427 static int
428 unix_do_apply_dentry(const char *output_path, size_t output_path_len,
429                      struct wim_dentry *dentry, struct apply_args *args)
430 {
431         const struct wim_inode *inode = dentry->d_inode;
432
433         if (inode_is_symlink(inode))
434                 return extract_symlink(dentry, args, output_path);
435         else if (inode_is_directory(inode))
436                 return extract_directory((args->extract_flags &
437                                            WIMLIB_EXTRACT_FLAG_UNIX_DATA) ? dentry : NULL,
438                                          output_path, false);
439         else
440                 return extract_regular_file(dentry, args, output_path);
441 }
442
443 static int
444 unix_do_apply_dentry_timestamps(const char *output_path,
445                                 size_t output_path_len,
446                                 const struct wim_dentry *dentry,
447                                 struct apply_args *args)
448 {
449         int ret;
450         const struct wim_inode *inode = dentry->d_inode;
451
452 #ifdef HAVE_UTIMENSAT
453         /* Convert the WIM timestamps, which are accurate to 100 nanoseconds,
454          * into `struct timespec's for passing to utimensat(), which is accurate
455          * to 1 nanosecond. */
456
457         struct timespec ts[2];
458         ts[0] = wim_timestamp_to_timespec(inode->i_last_access_time);
459         ts[1] = wim_timestamp_to_timespec(inode->i_last_write_time);
460         ret = utimensat(AT_FDCWD, output_path, ts, AT_SYMLINK_NOFOLLOW);
461         if (ret)
462                 ret = errno;
463 #else
464         ret = ENOSYS;
465 #endif
466
467         if (ret == ENOSYS) {
468                 /* utimensat() not implemented or not available */
469         #ifdef HAVE_LUTIMES
470                 /* Convert the WIM timestamps, which are accurate to 100
471                  * nanoseconds, into `struct timeval's for passing to lutimes(),
472                  * which is accurate to 1 microsecond. */
473                 struct timeval tv[2];
474                 tv[0] = wim_timestamp_to_timeval(inode->i_last_access_time);
475                 tv[1] = wim_timestamp_to_timeval(inode->i_last_write_time);
476                 ret = lutimes(output_path, tv);
477                 if (ret)
478                         ret = errno;
479         #endif
480         }
481
482         if (ret == ENOSYS) {
483                 /* utimensat() and lutimes() both not implemented or not
484                  * available */
485         #ifdef HAVE_UTIME
486                 /* Convert the WIM timestamps, which are accurate to 100
487                  * nanoseconds, into a `struct utimbuf's for passing to
488                  * utime(), which is accurate to 1 second. */
489                 struct utimbuf buf;
490                 buf.actime = wim_timestamp_to_unix(inode->i_last_access_time);
491                 buf.modtime = wim_timestamp_to_unix(inode->i_last_write_time);
492                 ret = utime(output_path, &buf);
493         #endif
494         }
495         if (ret && args->num_utime_warnings < 10) {
496                 WARNING_WITH_ERRNO("Failed to set timestamp on file `%s'",
497                                     output_path);
498                 args->num_utime_warnings++;
499         }
500         return 0;
501 }
502 #endif /* !__WIN32__ */
503
504 /* Extracts a file, directory, or symbolic link from the WIM archive. */
505 static int
506 apply_dentry_normal(struct wim_dentry *dentry, void *arg)
507 {
508         struct apply_args *args = arg;
509         size_t len;
510         tchar *output_path;
511
512         len = tstrlen(args->target);
513         if (dentry_is_root(dentry)) {
514                 output_path = (tchar*)args->target;
515         } else {
516                 output_path = alloca(len * sizeof(tchar) + dentry->full_path_nbytes +
517                                      sizeof(tchar));
518                 memcpy(output_path, args->target, len * sizeof(tchar));
519                 memcpy(output_path + len, dentry->_full_path, dentry->full_path_nbytes);
520                 len += dentry->full_path_nbytes / sizeof(tchar);
521                 output_path[len] = T('\0');
522         }
523 #ifdef __WIN32__
524         return win32_do_apply_dentry(output_path, len, dentry, args);
525 #else
526         return unix_do_apply_dentry(output_path, len, dentry, args);
527 #endif
528 }
529
530
531 /* Apply timestamps to an extracted file or directory */
532 static int
533 apply_dentry_timestamps_normal(struct wim_dentry *dentry, void *arg)
534 {
535         struct apply_args *args = arg;
536         size_t len;
537         tchar *output_path;
538
539         len = tstrlen(args->target);
540         if (dentry_is_root(dentry)) {
541                 output_path = (tchar*)args->target;
542         } else {
543                 output_path = alloca(len * sizeof(tchar) + dentry->full_path_nbytes +
544                                      sizeof(tchar));
545                 memcpy(output_path, args->target, len * sizeof(tchar));
546                 memcpy(output_path + len, dentry->_full_path, dentry->full_path_nbytes);
547                 len += dentry->full_path_nbytes / sizeof(tchar);
548                 output_path[len] = T('\0');
549         }
550
551 #ifdef __WIN32__
552         return win32_do_apply_dentry_timestamps(output_path, len, dentry, args);
553 #else
554         return unix_do_apply_dentry_timestamps(output_path, len, dentry, args);
555 #endif
556 }
557
558 /* Extract a dentry if it hasn't already been extracted, and either the dentry
559  * has no streams or WIMLIB_EXTRACT_FLAG_NO_STREAMS is not specified. */
560 static int
561 maybe_apply_dentry(struct wim_dentry *dentry, void *arg)
562 {
563         struct apply_args *args = arg;
564         int ret;
565
566         if (dentry->is_extracted)
567                 return 0;
568
569         if (args->extract_flags & WIMLIB_EXTRACT_FLAG_NO_STREAMS)
570                 if (inode_unnamed_lte_resolved(dentry->d_inode))
571                         return 0;
572
573         if ((args->extract_flags & WIMLIB_EXTRACT_FLAG_VERBOSE) &&
574              args->progress_func) {
575                 args->progress.extract.cur_path = dentry->_full_path;
576                 args->progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY,
577                                     &args->progress);
578         }
579         ret = args->apply_dentry(dentry, args);
580         if (ret == 0)
581                 dentry->is_extracted = 1;
582         return ret;
583 }
584
585 static void
586 calculate_bytes_to_extract(struct list_head *stream_list,
587                            int extract_flags,
588                            union wimlib_progress_info *progress)
589 {
590         struct wim_lookup_table_entry *lte;
591         u64 total_bytes = 0;
592         u64 num_streams = 0;
593
594         /* For each stream to be extracted... */
595         list_for_each_entry(lte, stream_list, extraction_list) {
596                 if (extract_flags &
597                     (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK))
598                 {
599                         /* In the symlink or hard link extraction mode, each
600                          * stream will be extracted one time regardless of how
601                          * many dentries share the stream. */
602                         wimlib_assert(!(extract_flags & WIMLIB_EXTRACT_FLAG_NTFS));
603                         if (!lte->extracted_file) {
604                                 num_streams++;
605                                 total_bytes += wim_resource_size(lte);
606                         }
607                 } else {
608                         num_streams += lte->out_refcnt;
609                         total_bytes += lte->out_refcnt * wim_resource_size(lte);
610                 }
611         }
612         progress->extract.num_streams = num_streams;
613         progress->extract.total_bytes = total_bytes;
614         progress->extract.completed_bytes = 0;
615 }
616
617 static void
618 maybe_add_stream_for_extraction(struct wim_lookup_table_entry *lte,
619                                 struct list_head *stream_list)
620 {
621         if (++lte->out_refcnt == 1) {
622                 INIT_LIST_HEAD(&lte->inode_list);
623                 list_add_tail(&lte->extraction_list, stream_list);
624         }
625 }
626
627 static void
628 inode_find_streams_for_extraction(struct wim_inode *inode,
629                                   struct list_head *stream_list,
630                                   int extract_flags)
631 {
632         struct wim_lookup_table_entry *lte;
633         bool inode_added = false;
634
635         lte = inode_unnamed_lte_resolved(inode);
636         if (lte) {
637                 maybe_add_stream_for_extraction(lte, stream_list);
638                 list_add_tail(&inode->i_lte_inode_list, &lte->inode_list);
639                 inode_added = true;
640         }
641
642         /* Determine whether to include alternate data stream entries or not.
643          *
644          * UNIX:  Include them if extracting using NTFS-3g.
645          *
646          * Windows: Include them undconditionally, although if the filesystem is
647          * not NTFS we won't actually be able to extract them. */
648 #if defined(WITH_NTFS_3G)
649         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
650 #elif defined(__WIN32__)
651         if (1)
652 #else
653         if (0)
654 #endif
655         {
656                 for (unsigned i = 0; i < inode->i_num_ads; i++) {
657                         if (inode->i_ads_entries[i].stream_name_nbytes != 0) {
658                                 lte = inode->i_ads_entries[i].lte;
659                                 if (lte) {
660                                         maybe_add_stream_for_extraction(lte,
661                                                                         stream_list);
662                                         if (!inode_added) {
663                                                 list_add_tail(&inode->i_lte_inode_list,
664                                                               &lte->inode_list);
665                                                 inode_added = true;
666                                         }
667                                 }
668                         }
669                 }
670         }
671 }
672
673 static void
674 find_streams_for_extraction(struct wim_image_metadata *imd,
675                             struct list_head *stream_list,
676                             struct wim_lookup_table *lookup_table,
677                             int extract_flags)
678 {
679         struct wim_inode *inode;
680         struct wim_dentry *dentry;
681
682         for_lookup_table_entry(lookup_table, lte_zero_out_refcnt, NULL);
683         INIT_LIST_HEAD(stream_list);
684         image_for_each_inode(inode, imd) {
685                 if (!inode->i_resolved)
686                         inode_resolve_ltes(inode, lookup_table);
687                 inode_for_each_dentry(dentry, inode)
688                         dentry->is_extracted = 0;
689                 inode_find_streams_for_extraction(inode, stream_list,
690                                                   extract_flags);
691         }
692 }
693
694 struct apply_operations {
695         int (*apply_dentry)(struct wim_dentry *dentry, void *arg);
696         int (*apply_dentry_timestamps)(struct wim_dentry *dentry, void *arg);
697 };
698
699 static const struct apply_operations normal_apply_operations = {
700         .apply_dentry = apply_dentry_normal,
701         .apply_dentry_timestamps = apply_dentry_timestamps_normal,
702 };
703
704 #ifdef WITH_NTFS_3G
705 static const struct apply_operations ntfs_apply_operations = {
706         .apply_dentry = apply_dentry_ntfs,
707         .apply_dentry_timestamps = apply_dentry_timestamps_ntfs,
708 };
709 #endif
710
711 static int
712 apply_stream_list(struct list_head *stream_list,
713                   struct apply_args *args,
714                   const struct apply_operations *ops,
715                   wimlib_progress_func_t progress_func)
716 {
717         uint64_t bytes_per_progress = args->progress.extract.total_bytes / 100;
718         uint64_t next_progress = bytes_per_progress;
719         struct wim_lookup_table_entry *lte;
720         struct wim_inode *inode;
721         struct wim_dentry *dentry;
722         int ret;
723
724         /* This complicated loop is essentially looping through the dentries,
725          * although dentries may be visited more than once (if a dentry contains
726          * two different nonempty streams) or not at all (if a dentry contains
727          * no non-empty streams).
728          *
729          * The outer loop is over the distinct streams to be extracted so that
730          * sequential reading of the WIM can be implemented. */
731
732         /* For each distinct stream to be extracted */
733         list_for_each_entry(lte, stream_list, extraction_list) {
734                 /* For each inode that contains the stream */
735                 list_for_each_entry(inode, &lte->inode_list, i_lte_inode_list) {
736                         /* For each dentry that points to the inode */
737                         inode_for_each_dentry(dentry, inode) {
738                                 /* Extract the dentry if it was not already
739                                  * extracted */
740                                 ret = maybe_apply_dentry(dentry, args);
741                                 if (ret)
742                                         return ret;
743                                 if (progress_func &&
744                                     args->progress.extract.completed_bytes >= next_progress)
745                                 {
746                                         progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS,
747                                                       &args->progress);
748                                         if (args->progress.extract.completed_bytes >=
749                                             args->progress.extract.total_bytes)
750                                         {
751                                                 next_progress = ~0ULL;
752                                         } else {
753                                                 next_progress =
754                                                         min (args->progress.extract.completed_bytes +
755                                                              bytes_per_progress,
756                                                              args->progress.extract.total_bytes);
757                                         }
758                                 }
759                         }
760                 }
761         }
762         return 0;
763 }
764
765 static int
766 sort_stream_list_by_wim_position(struct list_head *stream_list)
767 {
768         struct list_head *cur;
769         size_t num_streams;
770         struct wim_lookup_table_entry **array;
771         size_t i;
772         size_t array_size;
773
774         num_streams = 0;
775         list_for_each(cur, stream_list)
776                 num_streams++;
777         array_size = num_streams * sizeof(array[0]);
778         array = MALLOC(array_size);
779         if (!array) {
780                 ERROR("Failed to allocate %zu bytes to sort stream entries",
781                       array_size);
782                 return WIMLIB_ERR_NOMEM;
783         }
784         cur = stream_list->next;
785         for (i = 0; i < num_streams; i++) {
786                 array[i] = container_of(cur, struct wim_lookup_table_entry, extraction_list);
787                 cur = cur->next;
788         }
789
790         qsort(array, num_streams, sizeof(array[0]), cmp_streams_by_wim_position);
791
792         INIT_LIST_HEAD(stream_list);
793         for (i = 0; i < num_streams; i++)
794                 list_add_tail(&array[i]->extraction_list, stream_list);
795         FREE(array);
796         return 0;
797 }
798
799
800 /* Extracts the image @image from the WIM @w to the directory or NTFS volume
801  * @target. */
802 static int
803 extract_single_image(WIMStruct *w, int image,
804                      const tchar *target, int extract_flags,
805                      wimlib_progress_func_t progress_func)
806 {
807         int ret;
808         struct list_head stream_list;
809
810         struct apply_args args;
811         const struct apply_operations *ops;
812
813         memset(&args, 0, sizeof(args));
814
815         args.w                  = w;
816         args.target             = target;
817         args.extract_flags      = extract_flags;
818         args.progress_func      = progress_func;
819
820         if (progress_func) {
821                 args.progress.extract.wimfile_name = w->filename;
822                 args.progress.extract.image = image;
823                 args.progress.extract.extract_flags = (extract_flags &
824                                                        WIMLIB_EXTRACT_MASK_PUBLIC);
825                 args.progress.extract.image_name = wimlib_get_image_name(w, image);
826                 args.progress.extract.target = target;
827         }
828
829 #ifdef WITH_NTFS_3G
830         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
831                 args.vol = ntfs_mount(target, 0);
832                 if (!args.vol) {
833                         ERROR_WITH_ERRNO("Failed to mount NTFS volume `%"TS"'",
834                                          target);
835                         return WIMLIB_ERR_NTFS_3G;
836                 }
837                 ops = &ntfs_apply_operations;
838         } else
839 #endif
840                 ops = &normal_apply_operations;
841
842         ret = select_wim_image(w, image);
843         if (ret)
844                 goto out;
845
846         /* Build a list of the streams that need to be extracted */
847         find_streams_for_extraction(wim_get_current_image_metadata(w),
848                                     &stream_list,
849                                     w->lookup_table, extract_flags);
850
851         /* Calculate the number of bytes of data that will be extracted */
852         calculate_bytes_to_extract(&stream_list, extract_flags,
853                                    &args.progress);
854
855         if (progress_func) {
856                 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN,
857                               &args.progress);
858         }
859
860         /* If a sequential extraction was specified, sort the streams to be
861          * extracted by their position in the WIM file, so that the WIM file can
862          * be read sequentially. */
863         if (extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) {
864                 ret = sort_stream_list_by_wim_position(&stream_list);
865                 if (ret != 0) {
866                         WARNING("Falling back to non-sequential extraction");
867                         extract_flags &= ~WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
868                 }
869         }
870
871         if (progress_func) {
872                 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
873                               &args.progress);
874         }
875
876         ret = calculate_dentry_tree_full_paths(wim_root_dentry(w));
877         if (ret)
878                 goto out;
879
880         /* Make the directory structure and extract empty files */
881         args.extract_flags |= WIMLIB_EXTRACT_FLAG_NO_STREAMS;
882         args.apply_dentry = ops->apply_dentry;
883         ret = for_dentry_in_tree(wim_root_dentry(w), maybe_apply_dentry, &args);
884         args.extract_flags &= ~WIMLIB_EXTRACT_FLAG_NO_STREAMS;
885         if (ret)
886                 goto out;
887
888         if (progress_func) {
889                 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
890                               &args.progress);
891         }
892
893         if (extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) {
894                 args.target_realpath = realpath(target, NULL);
895                 if (!args.target_realpath)
896                         return WIMLIB_ERR_NOMEM;
897                 args.target_realpath_len = tstrlen(args.target_realpath);
898         }
899
900         /* Extract non-empty files */
901         ret = apply_stream_list(&stream_list, &args, ops, progress_func);
902         if (ret)
903                 goto out_free_target_realpath;
904
905         if (progress_func) {
906                 progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS,
907                               &args.progress);
908         }
909
910         /* Apply timestamps */
911         ret = for_dentry_in_tree_depth(wim_root_dentry(w),
912                                        ops->apply_dentry_timestamps, &args);
913         if (ret)
914                 goto out_free_target_realpath;
915
916         if (progress_func) {
917                 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END,
918                               &args.progress);
919         }
920 out_free_target_realpath:
921         FREE(args.target_realpath);
922 out:
923 #ifdef WITH_NTFS_3G
924         /* Unmount the NTFS volume */
925         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
926                 if (ntfs_umount(args.vol, FALSE) != 0) {
927                         ERROR_WITH_ERRNO("Failed to unmount NTFS volume `%"TS"'",
928                                          args.target);
929                         if (ret == 0)
930                                 ret = WIMLIB_ERR_NTFS_3G;
931                 }
932         }
933 #endif
934         return ret;
935 }
936
937 static const tchar * const filename_forbidden_chars =
938 T(
939 #ifdef __WIN32__
940 "<>:\"/\\|?*"
941 #else
942 "/"
943 #endif
944 );
945
946 /* This function checks if it is okay to use a WIM image's name as a directory
947  * name.  */
948 static bool
949 image_name_ok_as_dir(const tchar *image_name)
950 {
951         return image_name && *image_name &&
952                 !tstrpbrk(image_name, filename_forbidden_chars);
953 }
954
955 /* Extracts all images from the WIM to the directory @target, with the images
956  * placed in subdirectories named by their image names. */
957 static int
958 extract_all_images(WIMStruct *w,
959                    const tchar *target,
960                    int extract_flags,
961                    wimlib_progress_func_t progress_func)
962 {
963         size_t image_name_max_len = max(xml_get_max_image_name_len(w), 20);
964         size_t output_path_len = tstrlen(target);
965         tchar buf[output_path_len + 1 + image_name_max_len + 1];
966         int ret;
967         int image;
968         const tchar *image_name;
969
970         ret = extract_directory(NULL, target, true);
971         if (ret)
972                 return ret;
973
974         tmemcpy(buf, target, output_path_len);
975         buf[output_path_len] = T('/');
976         for (image = 1; image <= w->hdr.image_count; image++) {
977                 image_name = wimlib_get_image_name(w, image);
978                 if (image_name_ok_as_dir(image_name)) {
979                         tstrcpy(buf + output_path_len + 1, image_name);
980                 } else {
981                         /* Image name is empty, or contains forbidden
982                          * characters. */
983                         tsprintf(buf + output_path_len + 1, T("%d"), image);
984                 }
985                 ret = extract_single_image(w, image, buf, extract_flags,
986                                            progress_func);
987                 if (ret != 0)
988                         return ret;
989         }
990         return 0;
991 }
992
993 /* Extracts a single image or all images from a WIM file to a directory or NTFS
994  * volume. */
995 WIMLIBAPI int
996 wimlib_extract_image(WIMStruct *w,
997                      int image,
998                      const tchar *target,
999                      int extract_flags,
1000                      WIMStruct **additional_swms,
1001                      unsigned num_additional_swms,
1002                      wimlib_progress_func_t progress_func)
1003 {
1004         struct wim_lookup_table *joined_tab, *w_tab_save;
1005         int ret;
1006
1007         if (!target)
1008                 return WIMLIB_ERR_INVALID_PARAM;
1009
1010         extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
1011
1012         if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK))
1013                         == (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK))
1014                 return WIMLIB_ERR_INVALID_PARAM;
1015
1016 #ifdef __WIN32__
1017         if (extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
1018                 ERROR("Extracting UNIX data is not supported on Windows");
1019                 return WIMLIB_ERR_INVALID_PARAM;
1020         }
1021         if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK)) {
1022                 ERROR("Linked extraction modes are not supported on Windows");
1023                 return WIMLIB_ERR_INVALID_PARAM;
1024         }
1025 #endif
1026
1027         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1028 #ifdef WITH_NTFS_3G
1029                 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK))) {
1030                         ERROR("Cannot specify symlink or hardlink flags when applying\n"
1031                               "        directly to a NTFS volume");
1032                         return WIMLIB_ERR_INVALID_PARAM;
1033                 }
1034                 if (image == WIMLIB_ALL_IMAGES) {
1035                         ERROR("Can only apply a single image when applying "
1036                               "directly to a NTFS volume");
1037                         return WIMLIB_ERR_INVALID_PARAM;
1038                 }
1039                 if (extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
1040                         ERROR("Cannot restore UNIX-specific data in the NTFS extraction mode");
1041                         return WIMLIB_ERR_INVALID_PARAM;
1042                 }
1043 #else
1044                 ERROR("wimlib was compiled without support for NTFS-3g, so");
1045                 ERROR("we cannot apply a WIM image directly to a NTFS volume");
1046                 return WIMLIB_ERR_UNSUPPORTED;
1047 #endif
1048         }
1049
1050         if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1051                               WIMLIB_EXTRACT_FLAG_RPFIX)) ==
1052                 (WIMLIB_EXTRACT_FLAG_RPFIX | WIMLIB_EXTRACT_FLAG_NORPFIX))
1053         {
1054                 ERROR("Cannot specify RPFIX and NORPFIX flags at the same time!");
1055                 return WIMLIB_ERR_INVALID_PARAM;
1056         }
1057
1058         if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1059                               WIMLIB_EXTRACT_FLAG_NORPFIX)) == 0)
1060                 if (w->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1061                         extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1062
1063         ret = verify_swm_set(w, additional_swms, num_additional_swms);
1064         if (ret)
1065                 return ret;
1066
1067         ret = wim_checksum_unhashed_streams(w);
1068         if (ret)
1069                 return ret;
1070
1071         if (num_additional_swms) {
1072                 ret = new_joined_lookup_table(w, additional_swms,
1073                                               num_additional_swms, &joined_tab);
1074                 if (ret)
1075                         return ret;
1076                 w_tab_save = w->lookup_table;
1077                 w->lookup_table = joined_tab;
1078         }
1079
1080         if (image == WIMLIB_ALL_IMAGES) {
1081                 extract_flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
1082                 ret = extract_all_images(w, target, extract_flags,
1083                                          progress_func);
1084         } else {
1085                 extract_flags &= ~WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
1086                 ret = extract_single_image(w, image, target, extract_flags,
1087                                            progress_func);
1088         }
1089
1090         if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
1091                              WIMLIB_EXTRACT_FLAG_HARDLINK))
1092         {
1093                 for_lookup_table_entry(w->lookup_table,
1094                                        lte_free_extracted_file,
1095                                        NULL);
1096         }
1097
1098         if (num_additional_swms) {
1099                 free_lookup_table(w->lookup_table);
1100                 w->lookup_table = w_tab_save;
1101         }
1102         return ret;
1103 }