054984920de7188ea688dd3d075aadafaff3d62a
[wimlib] / src / extract.c
1 /*
2  * extract.c
3  *
4  * Support for extracting WIM files.
5  *
6  * This code does NOT contain any filesystem-specific features.  In particular,
7  * security information (i.e. file permissions) and alternate data streams are
8  * ignored, except possibly to read an alternate data stream that contains
9  * symbolic link data.
10  */
11
12 /*
13  * Copyright (C) 2010 Carl Thijssen
14  * Copyright (C) 2012 Eric Biggers
15  *
16  * This file is part of wimlib, a library for working with WIM files.
17  *
18  * wimlib is free software; you can redistribute it and/or modify it under the
19  * terms of the GNU General Public License as published by the Free
20  * Software Foundation; either version 3 of the License, or (at your option)
21  * any later version.
22  *
23  * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
24  * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
25  * A PARTICULAR PURPOSE. See the GNU General Public License for more
26  * details.
27  *
28  * You should have received a copy of the GNU General Public License
29  * along with wimlib; if not, see http://www.gnu.org/licenses/.
30  */
31
32
33 #include <dirent.h>
34 #include <errno.h>
35 #include <fcntl.h>
36 #include <string.h>
37 #include <sys/stat.h>
38 #include <sys/time.h>
39 #include <unistd.h>
40
41 #include "config.h"
42 #include "dentry.h"
43 #include "lookup_table.h"
44 #include "timestamp.h"
45 #include "wimlib_internal.h"
46 #include "xml.h"
47
48 /* Internal */
49 #define WIMLIB_EXTRACT_FLAG_MULTI_IMAGE 0x80000000
50
51 static int extract_regular_file_linked(const struct dentry *dentry,
52                                        const char *output_dir,
53                                        const char *output_path,
54                                        int extract_flags,
55                                        struct lookup_table_entry *lte)
56 {
57         /* This mode overrides the normal hard-link extraction and
58          * instead either symlinks or hardlinks *all* identical files in
59          * the WIM, even if they are in a different image (in the case
60          * of a multi-image extraction) */
61         wimlib_assert(lte->extracted_file);
62
63         if (extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) {
64                 if (link(lte->extracted_file, output_path) != 0) {
65                         ERROR_WITH_ERRNO("Failed to hard link "
66                                          "`%s' to `%s'",
67                                          output_path, lte->extracted_file);
68                         return WIMLIB_ERR_LINK;
69                 }
70         } else {
71                 int num_path_components;
72                 int num_output_dir_path_components;
73                 size_t extracted_file_len;
74                 char *p;
75                 const char *p2;
76                 size_t i;
77
78                 wimlib_assert(extract_flags & WIMLIB_EXTRACT_FLAG_SYMLINK);
79
80                 num_path_components =
81                         get_num_path_components(dentry->full_path_utf8) - 1;
82                 num_output_dir_path_components =
83                         get_num_path_components(output_dir);
84
85                 if (extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE) {
86                         num_path_components++;
87                         num_output_dir_path_components--;
88                 }
89                 extracted_file_len = strlen(lte->extracted_file);
90
91                 char buf[extracted_file_len + 3 * num_path_components + 1];
92                 p = &buf[0];
93
94                 for (i = 0; i < num_path_components; i++) {
95                         *p++ = '.';
96                         *p++ = '.';
97                         *p++ = '/';
98                 }
99                 p2 = lte->extracted_file;
100                 while (*p2 == '/')
101                         p2++;
102                 while (num_output_dir_path_components--)
103                         p2 = path_next_part(p2, NULL);
104                 strcpy(p, p2);
105                 if (symlink(buf, output_path) != 0) {
106                         ERROR_WITH_ERRNO("Failed to symlink `%s' to "
107                                          "`%s'",
108                                          buf, lte->extracted_file);
109                         return WIMLIB_ERR_LINK;
110                 }
111
112         }
113         return 0;
114 }
115
116 static int extract_regular_file_unlinked(WIMStruct *w,
117                                          struct dentry *dentry,
118                                          const char *output_path,
119                                          int extract_flags,
120                                          struct lookup_table_entry *lte)
121 {
122         /* Normal mode of extraction.  Regular files and hard links are
123          * extracted in the way that they appear in the WIM. */
124
125         int out_fd;
126         int ret;
127         struct inode *inode = dentry->d_inode;
128
129         if (!((extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE)
130                 && (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
131                                      WIMLIB_EXTRACT_FLAG_HARDLINK))))
132         {
133                 /* If the dentry is one of a hard link set of at least 2
134                  * dentries and one of the other dentries has already been
135                  * extracted, make a hard link to the file corresponding to this
136                  * already-extracted directory.  Otherwise, extract the file,
137                  * and set the inode->extracted_file field so that other
138                  * dentries in the hard link group can link to it. */
139                 if (inode->link_count > 1) {
140                         if (inode->extracted_file) {
141                                 DEBUG("Extracting hard link `%s' => `%s'",
142                                       output_path, inode->extracted_file);
143                                 if (link(inode->extracted_file, output_path) != 0) {
144                                         ERROR_WITH_ERRNO("Failed to hard link "
145                                                          "`%s' to `%s'",
146                                                          output_path,
147                                                          inode->extracted_file);
148                                         return WIMLIB_ERR_LINK;
149                                 }
150                                 return 0;
151                         }
152                         FREE(inode->extracted_file);
153                         inode->extracted_file = STRDUP(output_path);
154                         if (!inode->extracted_file) {
155                                 ERROR("Failed to allocate memory for filename");
156                                 return WIMLIB_ERR_NOMEM;
157                         }
158                 }
159         }
160
161         /* Extract the contents of the file to @output_path. */
162
163         out_fd = open(output_path, O_WRONLY | O_CREAT | O_TRUNC, 0644);
164         if (out_fd == -1) {
165                 ERROR_WITH_ERRNO("Failed to open the file `%s' for writing",
166                                  output_path);
167                 return WIMLIB_ERR_OPEN;
168         }
169
170         if (!lte) {
171                 /* Empty file with no lookup table entry */
172                 DEBUG("Empty file `%s'.", output_path);
173                 ret = 0;
174                 goto out;
175         }
176
177         ret = extract_full_wim_resource_to_fd(lte, out_fd);
178         if (ret != 0) {
179                 ERROR("Failed to extract resource to `%s'", output_path);
180                 goto out;
181         }
182
183 out:
184         if (close(out_fd) != 0) {
185                 ERROR_WITH_ERRNO("Failed to close file `%s'", output_path);
186                 ret = WIMLIB_ERR_WRITE;
187         }
188         return ret;
189 }
190
191 /*
192  * Extracts a regular file from the WIM archive.
193  */
194 static int extract_regular_file(WIMStruct *w,
195                                 struct dentry *dentry,
196                                 const char *output_dir,
197                                 const char *output_path,
198                                 int extract_flags)
199 {
200         struct lookup_table_entry *lte;
201         const struct inode *inode = dentry->d_inode;
202
203         lte = inode_unnamed_lte(inode, w->lookup_table);
204
205         if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
206                               WIMLIB_EXTRACT_FLAG_HARDLINK)) && lte) {
207                 if (lte->extracted_file) {
208                         return extract_regular_file_linked(dentry, output_dir,
209                                                            output_path,
210                                                            extract_flags, lte);
211                 } else {
212                         lte->extracted_file = STRDUP(output_path);
213                         if (!lte->extracted_file)
214                                 return WIMLIB_ERR_NOMEM;
215                 }
216         }
217
218         return extract_regular_file_unlinked(w, dentry, output_path,
219                                              extract_flags, lte);
220
221 }
222
223 static int extract_symlink(const struct dentry *dentry, const char *output_path,
224                            const WIMStruct *w)
225 {
226         char target[4096];
227         ssize_t ret = inode_readlink(dentry->d_inode, target, sizeof(target), w);
228         if (ret <= 0) {
229                 ERROR("Could not read the symbolic link from dentry `%s'",
230                       dentry->full_path_utf8);
231                 return WIMLIB_ERR_INVALID_DENTRY;
232         }
233         ret = symlink(target, output_path);
234         if (ret != 0) {
235                 ERROR_WITH_ERRNO("Failed to symlink `%s' to `%s'",
236                                  output_path, target);
237                 return WIMLIB_ERR_LINK;
238         }
239         return 0;
240 }
241
242 /*
243  * Extracts a directory from the WIM archive.
244  *
245  * @dentry:             The directory entry for the directory.
246  * @output_path:        The path to which the directory is to be extracted to.
247  * @return:             True on success, false on failure.
248  */
249 static int extract_directory(const char *output_path, bool is_root)
250 {
251         int ret;
252         struct stat stbuf;
253         ret = stat(output_path, &stbuf);
254         if (ret == 0) {
255                 if (S_ISDIR(stbuf.st_mode)) {
256                         if (!is_root)
257                                 WARNING("`%s' already exists", output_path);
258                         return 0;
259                 } else {
260                         ERROR("`%s' is not a directory", output_path);
261                         return WIMLIB_ERR_MKDIR;
262                 }
263         } else {
264                 if (errno != ENOENT) {
265                         ERROR_WITH_ERRNO("Failed to stat `%s'", output_path);
266                         return WIMLIB_ERR_STAT;
267                 }
268         }
269         /* Compute the output path directory to the directory. */
270         if (mkdir(output_path, S_IRWXU | S_IRGRP | S_IXGRP |
271                                S_IROTH | S_IXOTH) != 0) {
272                 ERROR_WITH_ERRNO("Cannot create directory `%s'",
273                                  output_path);
274                 return WIMLIB_ERR_MKDIR;
275         }
276         return 0;
277 }
278
279 struct extract_args {
280         WIMStruct *w;
281         int extract_flags;
282         const char *output_dir;
283         unsigned num_lutimes_warnings;
284 };
285
286 /*
287  * Extracts a file, directory, or symbolic link from the WIM archive.  For use
288  * in for_dentry_in_tree().
289  */
290 static int extract_dentry(struct dentry *dentry, void *arg)
291 {
292         struct extract_args *args = arg;
293         WIMStruct *w = args->w;
294         int extract_flags = args->extract_flags;
295         size_t len = strlen(args->output_dir);
296         char output_path[len + dentry->full_path_utf8_len + 1];
297
298         if (extract_flags & WIMLIB_EXTRACT_FLAG_VERBOSE) {
299                 wimlib_assert(dentry->full_path_utf8);
300                 puts(dentry->full_path_utf8);
301         }
302
303         memcpy(output_path, args->output_dir, len);
304         memcpy(output_path + len, dentry->full_path_utf8, dentry->full_path_utf8_len);
305         output_path[len + dentry->full_path_utf8_len] = '\0';
306
307         if (dentry_is_symlink(dentry))
308                 return extract_symlink(dentry, output_path, w);
309         else if (dentry_is_directory(dentry))
310                 return extract_directory(output_path, dentry_is_root(dentry));
311         else
312                 return extract_regular_file(w, dentry, args->output_dir,
313                                             output_path, extract_flags);
314 }
315
316 /* Apply timestamp to extracted file */
317 static int apply_dentry_timestamps(struct dentry *dentry, void *arg)
318 {
319         struct extract_args *args = arg;
320         size_t len = strlen(args->output_dir);
321         char output_path[len + dentry->full_path_utf8_len + 1];
322
323         memcpy(output_path, args->output_dir, len);
324         memcpy(output_path + len, dentry->full_path_utf8, dentry->full_path_utf8_len);
325         output_path[len + dentry->full_path_utf8_len] = '\0';
326
327         struct timeval tv[2];
328         wim_timestamp_to_timeval(dentry->d_inode->last_access_time, &tv[0]);
329         wim_timestamp_to_timeval(dentry->d_inode->last_write_time, &tv[1]);
330         if (lutimes(output_path, tv) != 0) {
331                 if (errno != ENOSYS || args->num_lutimes_warnings < 10) {
332                         WARNING("Failed to set timestamp on file `%s': %s",
333                                 output_path, strerror(errno));
334                         args->num_lutimes_warnings++;
335                 }
336         }
337         return 0;
338 }
339
340
341 static int extract_single_image(WIMStruct *w, int image,
342                                 const char *output_dir, int extract_flags)
343 {
344         DEBUG("Extracting image %d", image);
345
346         int ret;
347         ret = select_wim_image(w, image);
348         if (ret != 0)
349                 return ret;
350
351         struct extract_args args = {
352                 .w                    = w,
353                 .extract_flags        = extract_flags,
354                 .output_dir           = output_dir,
355                 .num_lutimes_warnings = 0,
356         };
357
358         ret = for_dentry_in_tree(wim_root_dentry(w), extract_dentry, &args);
359         if (ret != 0)
360                 return ret;
361         return for_dentry_in_tree_depth(wim_root_dentry(w),
362                                         apply_dentry_timestamps, &args);
363
364 }
365
366
367 /* Extracts all images from the WIM to @output_dir, with the images placed in
368  * subdirectories named by their image names. */
369 static int extract_all_images(WIMStruct *w, const char *output_dir,
370                               int extract_flags)
371 {
372         size_t image_name_max_len = max(xml_get_max_image_name_len(w), 20);
373         size_t output_path_len = strlen(output_dir);
374         char buf[output_path_len + 1 + image_name_max_len + 1];
375         int ret;
376         int image;
377         const char *image_name;
378
379         DEBUG("Attempting to extract all images from `%s' to `%s'",
380               w->filename, output_dir);
381
382         ret = extract_directory(output_dir, true);
383         if (ret != 0)
384                 return ret;
385
386         memcpy(buf, output_dir, output_path_len);
387         buf[output_path_len] = '/';
388         for (image = 1; image <= w->hdr.image_count; image++) {
389
390                 image_name = wimlib_get_image_name(w, image);
391                 if (*image_name) {
392                         strcpy(buf + output_path_len + 1, image_name);
393                 } else {
394                         /* Image name is empty. Use image number instead */
395                         sprintf(buf + output_path_len + 1, "%d", image);
396                 }
397                 ret = extract_single_image(w, image, buf, extract_flags);
398                 if (ret != 0)
399                         return ret;
400         }
401         return 0;
402 }
403
404
405 /* Extracts a single image or all images from a WIM file. */
406 WIMLIBAPI int wimlib_extract_image(WIMStruct *w, int image,
407                                    const char *output_dir, int flags,
408                                    WIMStruct **additional_swms,
409                                    unsigned num_additional_swms)
410 {
411         struct lookup_table *joined_tab, *w_tab_save;
412         int ret;
413
414         DEBUG("w->filename = %s, image = %d, output_dir = %s, flags = 0x%x, "
415               "num_additional_swms = %u",
416               w->filename, image, output_dir, flags, num_additional_swms);
417
418         if (!w || !output_dir)
419                 return WIMLIB_ERR_INVALID_PARAM;
420
421         if ((flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK))
422                         == (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK))
423                 return WIMLIB_ERR_INVALID_PARAM;
424
425         ret = verify_swm_set(w, additional_swms, num_additional_swms);
426         if (ret != 0)
427                 return ret;
428
429         if (num_additional_swms) {
430                 ret = new_joined_lookup_table(w, additional_swms,
431                                               num_additional_swms, &joined_tab);
432                 if (ret != 0)
433                         return ret;
434                 w_tab_save = w->lookup_table;
435                 w->lookup_table = joined_tab;
436         }
437
438         for_lookup_table_entry(w->lookup_table, lte_free_extracted_file, NULL);
439
440         if (image == WIM_ALL_IMAGES) {
441                 flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
442                 ret = extract_all_images(w, output_dir, flags);
443         } else {
444                 flags &= ~WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
445                 ret = extract_single_image(w, image, output_dir, flags);
446         }
447         if (num_additional_swms) {
448                 free_lookup_table(w->lookup_table);
449                 w->lookup_table = w_tab_save;
450         }
451         return ret;
452
453 }