5f311766364f1ae17beee27fc079da8c625fc265
[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 =
115                         get_num_path_components(dentry->full_path) - 1;
116                 num_output_dir_path_components =
117                         get_num_path_components(args->target);
118
119                 if (args->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE) {
120                         num_path_components++;
121                         num_output_dir_path_components--;
122                 }
123                 extracted_file_len = strlen(lte->extracted_file);
124
125                 char buf[extracted_file_len + 3 * num_path_components + 1];
126                 p = &buf[0];
127
128                 for (i = 0; i < num_path_components; i++) {
129                         *p++ = '.';
130                         *p++ = '.';
131                         *p++ = '/';
132                 }
133                 p2 = lte->extracted_file;
134                 while (*p2 == '/')
135                         p2++;
136                 while (num_output_dir_path_components > 0) {
137                         p2 = path_next_part(p2);
138                         num_output_dir_path_components--;
139                 }
140                 strcpy(p, p2);
141                 if (symlink(buf, output_path) != 0) {
142                         ERROR_WITH_ERRNO("Failed to symlink `%s' to `%s'",
143                                          buf, lte->extracted_file);
144                         return WIMLIB_ERR_LINK;
145                 }
146         }
147         return 0;
148 }
149
150 static int
151 symlink_apply_unix_data(const char *link,
152                         const struct wimlib_unix_data *unix_data)
153 {
154         if (lchown(link, unix_data->uid, unix_data->gid)) {
155                 if (errno == EPERM) {
156                         /* Ignore */
157                         WARNING_WITH_ERRNO("failed to set symlink UNIX owner/group");
158                 } else {
159                         ERROR_WITH_ERRNO("failed to set symlink UNIX owner/group");
160                         return WIMLIB_ERR_INVALID_DENTRY;
161                 }
162         }
163         return 0;
164 }
165
166 static int
167 fd_apply_unix_data(int fd, const struct wimlib_unix_data *unix_data)
168 {
169         if (fchown(fd, unix_data->uid, unix_data->gid)) {
170                 if (errno == EPERM) {
171                         WARNING_WITH_ERRNO("failed to set file UNIX owner/group");
172                         /* Ignore? */
173                 } else {
174                         ERROR_WITH_ERRNO("failed to set file UNIX owner/group");
175                         return WIMLIB_ERR_INVALID_DENTRY;
176                 }
177         }
178
179         if (fchmod(fd, unix_data->mode)) {
180                 if (errno == EPERM) {
181                         WARNING_WITH_ERRNO("failed to set UNIX file mode");
182                         /* Ignore? */
183                 } else {
184                         ERROR_WITH_ERRNO("failed to set UNIX file mode");
185                         return WIMLIB_ERR_INVALID_DENTRY;
186                 }
187         }
188         return 0;
189 }
190
191 static int
192 dir_apply_unix_data(const char *dir, const struct wimlib_unix_data *unix_data)
193 {
194         int dfd = open(dir, O_RDONLY);
195         int ret;
196         if (dfd >= 0) {
197                 ret = fd_apply_unix_data(dfd, unix_data);
198                 if (close(dfd)) {
199                         ERROR_WITH_ERRNO("can't close directory `%s'", dir);
200                         ret = WIMLIB_ERR_MKDIR;
201                 }
202         } else {
203                 ERROR_WITH_ERRNO("can't open directory `%s'", dir);
204                 ret = WIMLIB_ERR_MKDIR;
205         }
206         return ret;
207 }
208
209 static int
210 extract_regular_file_unlinked(struct wim_dentry *dentry,
211                               struct apply_args *args,
212                               const char *output_path,
213                               struct wim_lookup_table_entry *lte)
214 {
215         /* Normal mode of extraction.  Regular files and hard links are
216          * extracted in the way that they appear in the WIM. */
217
218         int out_fd;
219         int ret;
220         struct wim_inode *inode = dentry->d_inode;
221
222         if (!((args->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE)
223                 && (args->extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
224                                      WIMLIB_EXTRACT_FLAG_HARDLINK))))
225         {
226                 /* If the dentry is part of a hard link set of at least 2
227                  * dentries and one of the other dentries has already been
228                  * extracted, make a hard link to the file corresponding to this
229                  * already-extracted directory.  Otherwise, extract the file and
230                  * set the inode->i_extracted_file field so that other dentries
231                  * in the hard link group can link to it. */
232                 if (inode->i_nlink > 1) {
233                         if (inode->i_extracted_file) {
234                                 DEBUG("Extracting hard link `%s' => `%s'",
235                                       output_path, inode->i_extracted_file);
236                                 if (link(inode->i_extracted_file, output_path) != 0) {
237                                         ERROR_WITH_ERRNO("Failed to hard link "
238                                                          "`%s' to `%s'",
239                                                          output_path,
240                                                          inode->i_extracted_file);
241                                         return WIMLIB_ERR_LINK;
242                                 }
243                                 return 0;
244                         }
245                         FREE(inode->i_extracted_file);
246                         inode->i_extracted_file = STRDUP(output_path);
247                         if (!inode->i_extracted_file) {
248                                 ERROR("Failed to allocate memory for filename");
249                                 return WIMLIB_ERR_NOMEM;
250                         }
251                 }
252         }
253
254         /* Extract the contents of the file to @output_path. */
255
256         out_fd = open(output_path, O_WRONLY | O_CREAT | O_TRUNC, 0644);
257         if (out_fd == -1) {
258                 ERROR_WITH_ERRNO("Failed to open the file `%s' for writing",
259                                  output_path);
260                 return WIMLIB_ERR_OPEN;
261         }
262
263         if (!lte) {
264                 /* Empty file with no lookup table entry */
265                 DEBUG("Empty file `%s'.", output_path);
266                 ret = 0;
267                 goto out_extract_unix_data;
268         }
269
270         ret = extract_wim_resource_to_fd(lte, out_fd, wim_resource_size(lte));
271         if (ret != 0) {
272                 ERROR("Failed to extract resource to `%s'", output_path);
273                 goto out;
274         }
275
276 out_extract_unix_data:
277         if (args->extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
278                 struct wimlib_unix_data unix_data;
279                 ret = inode_get_unix_data(inode, &unix_data, NULL);
280                 if (ret > 0)
281                         ;
282                 else if (ret < 0)
283                         ret = 0;
284                 else
285                         ret = fd_apply_unix_data(out_fd, &unix_data);
286                 if (ret != 0)
287                         goto out;
288         }
289         if (lte)
290                 args->progress.extract.completed_bytes += wim_resource_size(lte);
291 out:
292         if (close(out_fd) != 0) {
293                 ERROR_WITH_ERRNO("Failed to close file `%s'", output_path);
294                 if (ret == 0)
295                         ret = WIMLIB_ERR_WRITE;
296         }
297         return ret;
298 }
299
300 static int
301 extract_regular_file(struct wim_dentry *dentry,
302                      struct apply_args *args,
303                      const char *output_path)
304 {
305         struct wim_lookup_table_entry *lte;
306         const struct wim_inode *inode = dentry->d_inode;
307
308         lte = inode_unnamed_lte_resolved(inode);
309
310         if (lte && (args->extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
311                                            WIMLIB_EXTRACT_FLAG_HARDLINK)))
312         {
313                 if (lte->extracted_file) {
314                         return extract_regular_file_linked(dentry, output_path, args, lte);
315                 } else {
316                         lte->extracted_file = STRDUP(output_path);
317                         if (!lte->extracted_file)
318                                 return WIMLIB_ERR_NOMEM;
319                 }
320         }
321         return extract_regular_file_unlinked(dentry, args, output_path, lte);
322 }
323
324 static int
325 extract_symlink(struct wim_dentry *dentry,
326                 struct apply_args *args,
327                 const char *output_path)
328 {
329         char target[4096];
330         ssize_t ret = inode_readlink(dentry->d_inode, target,
331                                      sizeof(target), args->w, 0);
332         struct wim_lookup_table_entry *lte;
333
334         if (ret <= 0) {
335                 ERROR("Could not read the symbolic link from dentry `%s'",
336                       dentry->full_path);
337                 return WIMLIB_ERR_INVALID_DENTRY;
338         }
339         ret = symlink(target, output_path);
340         if (ret != 0) {
341                 ERROR_WITH_ERRNO("Failed to symlink `%s' to `%s'",
342                                  output_path, target);
343                 return WIMLIB_ERR_LINK;
344         }
345         lte = inode_unnamed_lte_resolved(dentry->d_inode);
346         wimlib_assert(lte != NULL);
347         if (args->extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
348                 struct wimlib_unix_data unix_data;
349                 ret = inode_get_unix_data(dentry->d_inode, &unix_data, NULL);
350                 if (ret > 0)
351                         ;
352                 else if (ret < 0)
353                         ret = 0;
354                 else
355                         ret = symlink_apply_unix_data(output_path, &unix_data);
356                 if (ret != 0)
357                         return ret;
358         }
359         args->progress.extract.completed_bytes += wim_resource_size(lte);
360         return 0;
361 }
362
363 #endif /* !__WIN32__ */
364
365 static int
366 extract_directory(struct wim_dentry *dentry,
367                   const tchar *output_path, bool is_root)
368 {
369         int ret;
370         struct stat stbuf;
371
372         ret = tstat(output_path, &stbuf);
373         if (ret == 0) {
374                 if (S_ISDIR(stbuf.st_mode)) {
375                         /*if (!is_root)*/
376                                 /*WARNING("`%s' already exists", output_path);*/
377                         goto dir_exists;
378                 } else {
379                         ERROR("`%"TS"' is not a directory", output_path);
380                         return WIMLIB_ERR_MKDIR;
381                 }
382         } else {
383                 if (errno != ENOENT) {
384                         ERROR_WITH_ERRNO("Failed to stat `%"TS"'", output_path);
385                         return WIMLIB_ERR_STAT;
386                 }
387         }
388
389         if (tmkdir(output_path, S_IRWXU | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH))
390         {
391                 ERROR_WITH_ERRNO("Cannot create directory `%"TS"'", output_path);
392                 return WIMLIB_ERR_MKDIR;
393         }
394 dir_exists:
395         ret = 0;
396 #ifndef __WIN32__
397         if (dentry) {
398                 struct wimlib_unix_data unix_data;
399                 ret = inode_get_unix_data(dentry->d_inode, &unix_data, NULL);
400                 if (ret > 0)
401                         ;
402                 else if (ret < 0)
403                         ret = 0;
404                 else
405                         ret = dir_apply_unix_data(output_path, &unix_data);
406         }
407 #endif
408         return ret;
409 }
410
411 #ifndef __WIN32__
412 static int unix_do_apply_dentry(const char *output_path,
413                                 size_t output_path_len,
414                                 struct wim_dentry *dentry,
415                                 struct apply_args *args)
416 {
417         const struct wim_inode *inode = dentry->d_inode;
418
419         if (inode_is_symlink(inode))
420                 return extract_symlink(dentry, args, output_path);
421         else if (inode_is_directory(inode))
422                 return extract_directory((args->extract_flags &
423                                            WIMLIB_EXTRACT_FLAG_UNIX_DATA) ? dentry : NULL,
424                                          output_path, false);
425         else
426                 return extract_regular_file(dentry, args, output_path);
427 }
428
429 static int
430 unix_do_apply_dentry_timestamps(const char *output_path,
431                                 size_t output_path_len,
432                                 const struct wim_dentry *dentry,
433                                 struct apply_args *args)
434 {
435         int ret;
436         const struct wim_inode *inode = dentry->d_inode;
437
438 #ifdef HAVE_UTIMENSAT
439         /* Convert the WIM timestamps, which are accurate to 100 nanoseconds,
440          * into `struct timespec's for passing to utimensat(), which is accurate
441          * to 1 nanosecond. */
442
443         struct timespec ts[2];
444         ts[0] = wim_timestamp_to_timespec(inode->i_last_access_time);
445         ts[1] = wim_timestamp_to_timespec(inode->i_last_write_time);
446         ret = utimensat(AT_FDCWD, output_path, ts, AT_SYMLINK_NOFOLLOW);
447         if (ret)
448                 ret = errno;
449 #else
450         ret = ENOSYS;
451 #endif
452
453         if (ret == ENOSYS) {
454                 /* utimensat() not implemented or not available */
455         #ifdef HAVE_LUTIMES
456                 /* Convert the WIM timestamps, which are accurate to 100
457                  * nanoseconds, into `struct timeval's for passing to lutimes(),
458                  * which is accurate to 1 microsecond. */
459                 struct timeval tv[2];
460                 tv[0] = wim_timestamp_to_timeval(inode->i_last_access_time);
461                 tv[1] = wim_timestamp_to_timeval(inode->i_last_write_time);
462                 ret = lutimes(output_path, tv);
463                 if (ret)
464                         ret = errno;
465         #endif
466         }
467
468         if (ret == ENOSYS) {
469                 /* utimensat() and lutimes() both not implemented or not
470                  * available */
471         #ifdef HAVE_UTIME
472                 /* Convert the WIM timestamps, which are accurate to 100
473                  * nanoseconds, into a `struct utimbuf's for passing to
474                  * utime(), which is accurate to 1 second. */
475                 struct utimbuf buf;
476                 buf.actime = wim_timestamp_to_unix(inode->i_last_access_time);
477                 buf.modtime = wim_timestamp_to_unix(inode->i_last_write_time);
478                 ret = utime(output_path, &buf);
479         #endif
480         }
481         if (ret && args->num_utime_warnings < 10) {
482                 WARNING_WITH_ERRNO("Failed to set timestamp on file `%s'",
483                                     output_path);
484                 args->num_utime_warnings++;
485         }
486         return 0;
487 }
488 #endif /* !__WIN32__ */
489
490 /* Extracts a file, directory, or symbolic link from the WIM archive. */
491 static int
492 apply_dentry_normal(struct wim_dentry *dentry, void *arg)
493 {
494         struct apply_args *args = arg;
495         tchar *output_path;
496         size_t len;
497
498         len = tstrlen(args->target);
499         if (dentry_is_root(dentry)) {
500                 output_path = (tchar*)args->target;
501         } else {
502                 output_path = alloca(len * sizeof(tchar) + dentry->full_path_nbytes +
503                                      sizeof(tchar));
504                 memcpy(output_path, args->target, len * sizeof(tchar));
505                 memcpy(output_path + len, dentry->full_path, dentry->full_path_nbytes);
506                 len += dentry->full_path_nbytes / sizeof(tchar);
507                 output_path[len] = T('\0');
508         }
509 #ifdef __WIN32__
510         return win32_do_apply_dentry(output_path, len, dentry, args);
511 #else
512         return unix_do_apply_dentry(output_path, len, dentry, args);
513 #endif
514 }
515
516
517 /* Apply timestamps to an extracted file or directory */
518 static int
519 apply_dentry_timestamps_normal(struct wim_dentry *dentry, void *arg)
520 {
521         struct apply_args *args = arg;
522         size_t len;
523         tchar *output_path;
524
525         len = tstrlen(args->target);
526         if (dentry_is_root(dentry)) {
527                 output_path = (tchar*)args->target;
528         } else {
529                 output_path = alloca(len * sizeof(tchar) + dentry->full_path_nbytes +
530                                      sizeof(tchar));
531                 memcpy(output_path, args->target, len * sizeof(tchar));
532                 memcpy(output_path + len, dentry->full_path, dentry->full_path_nbytes);
533                 len += dentry->full_path_nbytes / sizeof(tchar);
534                 output_path[len] = T('\0');
535         }
536
537
538 #ifdef __WIN32__
539         return win32_do_apply_dentry_timestamps(output_path, len, dentry, args);
540 #else
541         return unix_do_apply_dentry_timestamps(output_path, len, dentry, args);
542 #endif
543 }
544
545 /* Extract a dentry if it hasn't already been extracted, and either the dentry
546  * has no streams or WIMLIB_EXTRACT_FLAG_NO_STREAMS is not specified. */
547 static int
548 maybe_apply_dentry(struct wim_dentry *dentry, void *arg)
549 {
550         struct apply_args *args = arg;
551         int ret;
552
553         if (dentry->is_extracted)
554                 return 0;
555
556         if (args->extract_flags & WIMLIB_EXTRACT_FLAG_NO_STREAMS)
557                 if (inode_unnamed_lte_resolved(dentry->d_inode))
558                         return 0;
559
560         if ((args->extract_flags & WIMLIB_EXTRACT_FLAG_VERBOSE) &&
561              args->progress_func) {
562                 args->progress.extract.cur_path = dentry->full_path;
563                 args->progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY,
564                                     &args->progress);
565         }
566         ret = args->apply_dentry(dentry, args);
567         if (ret == 0)
568                 dentry->is_extracted = 1;
569         return ret;
570 }
571
572 static int
573 cmp_streams_by_wim_position(const void *p1, const void *p2)
574 {
575         const struct wim_lookup_table_entry *lte1, *lte2;
576         lte1 = *(const struct wim_lookup_table_entry**)p1;
577         lte2 = *(const struct wim_lookup_table_entry**)p2;
578         if (lte1->resource_entry.offset < lte2->resource_entry.offset)
579                 return -1;
580         else if (lte1->resource_entry.offset > lte2->resource_entry.offset)
581                 return 1;
582         else
583                 return 0;
584 }
585
586 static int
587 sort_stream_list_by_wim_position(struct list_head *stream_list)
588 {
589         struct list_head *cur;
590         size_t num_streams;
591         struct wim_lookup_table_entry **array;
592         size_t i;
593         size_t array_size;
594
595         num_streams = 0;
596         list_for_each(cur, stream_list)
597                 num_streams++;
598         array_size = num_streams * sizeof(array[0]);
599         array = MALLOC(array_size);
600         if (!array) {
601                 ERROR("Failed to allocate %zu bytes to sort stream entries",
602                       array_size);
603                 return WIMLIB_ERR_NOMEM;
604         }
605         cur = stream_list->next;
606         for (i = 0; i < num_streams; i++) {
607                 array[i] = container_of(cur, struct wim_lookup_table_entry, staging_list);
608                 cur = cur->next;
609         }
610
611         qsort(array, num_streams, sizeof(array[0]), cmp_streams_by_wim_position);
612
613         INIT_LIST_HEAD(stream_list);
614         for (i = 0; i < num_streams; i++)
615                 list_add_tail(&array[i]->staging_list, stream_list);
616         FREE(array);
617         return 0;
618 }
619
620 static void
621 calculate_bytes_to_extract(struct list_head *stream_list,
622                            int extract_flags,
623                            union wimlib_progress_info *progress)
624 {
625         struct wim_lookup_table_entry *lte;
626         u64 total_bytes = 0;
627         u64 num_streams = 0;
628
629         /* For each stream to be extracted... */
630         list_for_each_entry(lte, stream_list, staging_list) {
631                 if (extract_flags &
632                     (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK))
633                 {
634                         /* In the symlink or hard link extraction mode, each
635                          * stream will be extracted one time regardless of how
636                          * many dentries share the stream. */
637                         wimlib_assert(!(extract_flags & WIMLIB_EXTRACT_FLAG_NTFS));
638                         if (!lte->extracted_file) {
639                                 num_streams++;
640                                 total_bytes += wim_resource_size(lte);
641                         }
642                 } else {
643                         num_streams += lte->out_refcnt;
644                         total_bytes += lte->out_refcnt * wim_resource_size(lte);
645                 }
646         }
647         progress->extract.num_streams = num_streams;
648         progress->extract.total_bytes = total_bytes;
649         progress->extract.completed_bytes = 0;
650 }
651
652 static void
653 maybe_add_stream_for_extraction(struct wim_lookup_table_entry *lte,
654                                 struct list_head *stream_list)
655 {
656         if (++lte->out_refcnt == 1) {
657                 INIT_LIST_HEAD(&lte->inode_list);
658                 list_add_tail(&lte->staging_list, stream_list);
659         }
660 }
661
662 static void
663 inode_find_streams_for_extraction(struct wim_inode *inode,
664                                   struct list_head *stream_list,
665                                   int extract_flags)
666 {
667         struct wim_lookup_table_entry *lte;
668         bool inode_added = false;
669
670         lte = inode_unnamed_lte_resolved(inode);
671         if (lte) {
672                 maybe_add_stream_for_extraction(lte, stream_list);
673                 list_add_tail(&inode->i_lte_inode_list, &lte->inode_list);
674                 inode_added = true;
675         }
676 #ifdef WITH_NTFS_3G
677         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
678                 for (unsigned i = 0; i < inode->i_num_ads; i++) {
679                         if (inode->i_ads_entries[i].stream_name_nbytes != 0) {
680                                 lte = inode->i_ads_entries[i].lte;
681                                 if (lte) {
682                                         maybe_add_stream_for_extraction(lte,
683                                                                         stream_list);
684                                         if (!inode_added) {
685                                                 list_add_tail(&inode->i_lte_inode_list,
686                                                               &lte->inode_list);
687                                                 inode_added = true;
688                                         }
689                                 }
690                         }
691                 }
692         }
693 #endif
694 }
695
696 static void
697 find_streams_for_extraction(struct hlist_head *inode_list,
698                             struct list_head *stream_list,
699                             struct wim_lookup_table *lookup_table,
700                             int extract_flags)
701 {
702         struct wim_inode *inode;
703         struct hlist_node *cur;
704         struct wim_dentry *dentry;
705
706         for_lookup_table_entry(lookup_table, lte_zero_out_refcnt, NULL);
707         INIT_LIST_HEAD(stream_list);
708         hlist_for_each_entry(inode, cur, inode_list, i_hlist) {
709                 if (!inode->i_resolved)
710                         inode_resolve_ltes(inode, lookup_table);
711                 inode_for_each_dentry(dentry, inode)
712                         dentry->is_extracted = 0;
713                 inode_find_streams_for_extraction(inode, stream_list,
714                                                   extract_flags);
715         }
716 }
717
718 struct apply_operations {
719         int (*apply_dentry)(struct wim_dentry *dentry, void *arg);
720         int (*apply_dentry_timestamps)(struct wim_dentry *dentry, void *arg);
721 };
722
723 static const struct apply_operations normal_apply_operations = {
724         .apply_dentry = apply_dentry_normal,
725         .apply_dentry_timestamps = apply_dentry_timestamps_normal,
726 };
727
728 #ifdef WITH_NTFS_3G
729 static const struct apply_operations ntfs_apply_operations = {
730         .apply_dentry = apply_dentry_ntfs,
731         .apply_dentry_timestamps = apply_dentry_timestamps_ntfs,
732 };
733 #endif
734
735 static int
736 apply_stream_list(struct list_head *stream_list,
737                   struct apply_args *args,
738                   const struct apply_operations *ops,
739                   wimlib_progress_func_t progress_func)
740 {
741         uint64_t bytes_per_progress = args->progress.extract.total_bytes / 100;
742         uint64_t next_progress = bytes_per_progress;
743         struct wim_lookup_table_entry *lte;
744         struct wim_inode *inode;
745         struct wim_dentry *dentry;
746         int ret;
747
748         /* This complicated loop is essentially looping through the dentries,
749          * although dentries may be visited more than once (if a dentry contains
750          * two different nonempty streams) or not at all (if a dentry contains
751          * no non-empty streams).
752          *
753          * The outer loop is over the distinct streams to be extracted so that
754          * sequential reading of the WIM can be implemented. */
755
756         /* For each distinct stream to be extracted */
757         list_for_each_entry(lte, stream_list, staging_list) {
758                 /* For each inode that contains the stream */
759                 list_for_each_entry(inode, &lte->inode_list, i_lte_inode_list) {
760                         /* For each dentry that points to the inode */
761                         inode_for_each_dentry(dentry, inode) {
762                                 /* Extract the dentry if it was not already
763                                  * extracted */
764                                 ret = maybe_apply_dentry(dentry, args);
765                                 if (ret != 0)
766                                         return ret;
767                                 if (progress_func &&
768                                     args->progress.extract.completed_bytes >= next_progress)
769                                 {
770                                         progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS,
771                                                       &args->progress);
772                                         if (args->progress.extract.completed_bytes >=
773                                             args->progress.extract.total_bytes)
774                                         {
775                                                 next_progress = ~0ULL;
776                                         } else {
777                                                 next_progress =
778                                                         min (args->progress.extract.completed_bytes +
779                                                              bytes_per_progress,
780                                                              args->progress.extract.total_bytes);
781                                         }
782                                 }
783                         }
784                 }
785         }
786         return 0;
787 }
788
789 /* Extracts the image @image from the WIM @w to the directory or NTFS volume
790  * @target. */
791 static int
792 extract_single_image(WIMStruct *w, int image,
793                      const tchar *target, int extract_flags,
794                      wimlib_progress_func_t progress_func)
795 {
796         int ret;
797         struct list_head stream_list;
798         struct hlist_head *inode_list;
799
800         struct apply_args args;
801         const struct apply_operations *ops;
802
803         memset(&args, 0, sizeof(args));
804
805         args.w                  = w;
806         args.target             = target;
807         args.extract_flags      = extract_flags;
808         args.progress_func      = progress_func;
809
810         if (progress_func) {
811                 args.progress.extract.wimfile_name = w->filename;
812                 args.progress.extract.image = image;
813                 args.progress.extract.extract_flags = (extract_flags &
814                                                        WIMLIB_EXTRACT_MASK_PUBLIC);
815                 args.progress.extract.image_name = wimlib_get_image_name(w, image);
816                 args.progress.extract.target = target;
817         }
818
819 #ifdef WITH_NTFS_3G
820         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
821                 args.vol = ntfs_mount(target, 0);
822                 if (!args.vol) {
823                         ERROR_WITH_ERRNO("Failed to mount NTFS volume `%"TS"'",
824                                          target);
825                         return WIMLIB_ERR_NTFS_3G;
826                 }
827                 ops = &ntfs_apply_operations;
828         } else
829 #endif
830                 ops = &normal_apply_operations;
831
832         ret = select_wim_image(w, image);
833         if (ret != 0)
834                 goto out;
835
836         inode_list = &w->image_metadata[image - 1].inode_list;
837
838         /* Build a list of the streams that need to be extracted */
839         find_streams_for_extraction(inode_list, &stream_list,
840                                     w->lookup_table, extract_flags);
841
842         /* Calculate the number of bytes of data that will be extracted */
843         calculate_bytes_to_extract(&stream_list, extract_flags,
844                                    &args.progress);
845
846         if (progress_func) {
847                 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN,
848                               &args.progress);
849         }
850
851         /* If a sequential extraction was specified, sort the streams to be
852          * extracted by their position in the WIM file, so that the WIM file can
853          * be read sequentially. */
854         if (extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) {
855                 ret = sort_stream_list_by_wim_position(&stream_list);
856                 if (ret != 0) {
857                         WARNING("Falling back to non-sequential extraction");
858                         extract_flags &= ~WIMLIB_EXTRACT_FLAG_SEQUENTIAL;
859                 }
860         }
861
862         if (progress_func) {
863                 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
864                               &args.progress);
865         }
866
867         /* Make the directory structure and extract empty files */
868         args.extract_flags |= WIMLIB_EXTRACT_FLAG_NO_STREAMS;
869         args.apply_dentry = ops->apply_dentry;
870         ret = for_dentry_in_tree(wim_root_dentry(w), maybe_apply_dentry, &args);
871         args.extract_flags &= ~WIMLIB_EXTRACT_FLAG_NO_STREAMS;
872         if (ret != 0)
873                 goto out;
874
875         if (progress_func) {
876                 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
877                               &args.progress);
878         }
879
880         /* Extract non-empty files */
881         ret = apply_stream_list(&stream_list, &args, ops, progress_func);
882         if (ret != 0)
883                 goto out;
884
885         if (progress_func) {
886                 progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS,
887                               &args.progress);
888         }
889
890         /* Apply timestamps */
891         ret = for_dentry_in_tree_depth(wim_root_dentry(w),
892                                        ops->apply_dentry_timestamps, &args);
893         if (ret != 0)
894                 goto out;
895
896         if (progress_func) {
897                 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END,
898                               &args.progress);
899         }
900 out:
901 #ifdef WITH_NTFS_3G
902         /* Unmount the NTFS volume */
903         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
904                 if (ntfs_umount(args.vol, FALSE) != 0) {
905                         ERROR_WITH_ERRNO("Failed to unmount NTFS volume `%"TS"'",
906                                          args.target);
907                         if (ret == 0)
908                                 ret = WIMLIB_ERR_NTFS_3G;
909                 }
910         }
911 #endif
912         return ret;
913 }
914
915 static const tchar *filename_forbidden_chars = 
916 T(
917 #ifdef __WIN32__
918 "<>:\"/\\|?*"
919 #else
920 "/"
921 #endif
922 );
923
924 /* This function checks if it is okay to use a WIM image's name as a directory
925  * name.  */
926 static bool
927 image_name_ok_as_dir(const tchar *image_name)
928 {
929         return image_name && *image_name &&
930                 !tstrpbrk(image_name, filename_forbidden_chars);
931 }
932
933 /* Extracts all images from the WIM to the directory @target, with the images
934  * placed in subdirectories named by their image names. */
935 static int
936 extract_all_images(WIMStruct *w,
937                    const tchar *target,
938                    int extract_flags,
939                    wimlib_progress_func_t progress_func)
940 {
941         size_t image_name_max_len = max(xml_get_max_image_name_len(w), 20);
942         size_t output_path_len = tstrlen(target);
943         tchar buf[output_path_len + 1 + image_name_max_len + 1];
944         int ret;
945         int image;
946         const tchar *image_name;
947
948         ret = extract_directory(NULL, target, true);
949         if (ret)
950                 return ret;
951
952         tmemcpy(buf, target, output_path_len);
953         buf[output_path_len] = T('/');
954         for (image = 1; image <= w->hdr.image_count; image++) {
955                 image_name = wimlib_get_image_name(w, image);
956                 if (image_name_ok_as_dir(image_name)) {
957                         tstrcpy(buf + output_path_len + 1, image_name);
958                 } else {
959                         /* Image name is empty, or contains forbidden
960                          * characters. */
961                         tsprintf(buf + output_path_len + 1, T("%d"), image);
962                 }
963                 ret = extract_single_image(w, image, buf, extract_flags,
964                                            progress_func);
965                 if (ret != 0)
966                         return ret;
967         }
968         return 0;
969 }
970
971 /* Extracts a single image or all images from a WIM file to a directory or NTFS
972  * volume. */
973 WIMLIBAPI int
974 wimlib_extract_image(WIMStruct *w,
975                      int image,
976                      const tchar *target,
977                      int extract_flags,
978                      WIMStruct **additional_swms,
979                      unsigned num_additional_swms,
980                      wimlib_progress_func_t progress_func)
981 {
982         struct wim_lookup_table *joined_tab, *w_tab_save;
983         int ret;
984
985         if (!target)
986                 return WIMLIB_ERR_INVALID_PARAM;
987
988         extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC;
989
990         if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK))
991                         == (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK))
992                 return WIMLIB_ERR_INVALID_PARAM;
993
994 #ifdef __WIN32__
995         if (extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
996                 ERROR("Extracting UNIX data is not supported on Windows");
997                 return WIMLIB_ERR_INVALID_PARAM;
998         }
999         if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK)) {
1000                 ERROR("Linked extraction modes are not supported on Windows");
1001                 return WIMLIB_ERR_INVALID_PARAM;
1002         }
1003 #endif
1004
1005         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1006 #ifdef WITH_NTFS_3G
1007                 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK))) {
1008                         ERROR("Cannot specify symlink or hardlink flags when applying\n"
1009                               "        directly to a NTFS volume");
1010                         return WIMLIB_ERR_INVALID_PARAM;
1011                 }
1012                 if (image == WIMLIB_ALL_IMAGES) {
1013                         ERROR("Can only apply a single image when applying "
1014                               "directly to a NTFS volume");
1015                         return WIMLIB_ERR_INVALID_PARAM;
1016                 }
1017                 if (extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
1018                         ERROR("Cannot restore UNIX-specific data in the NTFS extraction mode");
1019                         return WIMLIB_ERR_INVALID_PARAM;
1020                 }
1021 #else
1022                 ERROR("wimlib was compiled without support for NTFS-3g, so");
1023                 ERROR("we cannot apply a WIM image directly to a NTFS volume");
1024                 return WIMLIB_ERR_UNSUPPORTED;
1025 #endif
1026         }
1027
1028         ret = verify_swm_set(w, additional_swms, num_additional_swms);
1029         if (ret != 0)
1030                 return ret;
1031
1032         if (num_additional_swms) {
1033                 ret = new_joined_lookup_table(w, additional_swms,
1034                                               num_additional_swms, &joined_tab);
1035                 if (ret != 0)
1036                         return ret;
1037                 w_tab_save = w->lookup_table;
1038                 w->lookup_table = joined_tab;
1039         }
1040
1041         if (image == WIMLIB_ALL_IMAGES) {
1042                 extract_flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
1043                 ret = extract_all_images(w, target, extract_flags,
1044                                          progress_func);
1045         } else {
1046                 extract_flags &= ~WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
1047                 ret = extract_single_image(w, image, target, extract_flags,
1048                                            progress_func);
1049         }
1050
1051         if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
1052                              WIMLIB_EXTRACT_FLAG_HARDLINK))
1053         {
1054                 for_lookup_table_entry(w->lookup_table,
1055                                        lte_free_extracted_file,
1056                                        NULL);
1057         }
1058
1059         if (num_additional_swms) {
1060                 free_lookup_table(w->lookup_table);
1061                 w->lookup_table = w_tab_save;
1062         }
1063         return ret;
1064 }