960a9d50aafc5b85bfd37b1fa9b2302d55d20d55
[wimlib] / src / wim.c
1 /*
2  * wim.c - Stuff that doesn't fit into any other file
3  */
4
5 /*
6  * Copyright (C) 2012, 2013 Eric Biggers
7  *
8  * This file is part of wimlib, a library for working with WIM files.
9  *
10  * wimlib is free software; you can redistribute it and/or modify it under the
11  * terms of the GNU General Public License as published by the Free
12  * Software Foundation; either version 3 of the License, or (at your option)
13  * any later version.
14  *
15  * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
16  * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
17  * A PARTICULAR PURPOSE. See the GNU General Public License for more
18  * details.
19  *
20  * You should have received a copy of the GNU General Public License
21  * along with wimlib; if not, see http://www.gnu.org/licenses/.
22  */
23
24 #ifdef HAVE_CONFIG_H
25 #  include "config.h"
26 #endif
27
28 #include "wimlib/error.h"
29 #include "wimlib/dentry.h"
30 #include "wimlib/encoding.h"
31 #include "wimlib/file_io.h"
32 #include "wimlib/integrity.h"
33 #include "wimlib/lookup_table.h"
34 #include "wimlib/metadata.h"
35 #ifdef WITH_NTFS_3G
36 #  include "wimlib/ntfs_3g.h" /* for do_ntfs_umount() */
37 #endif
38 #include "wimlib/security.h"
39 #include "wimlib/wim.h"
40 #include "wimlib/xml.h"
41 #include "wimlib/compressor_ops.h"
42 #include "wimlib/decompressor_ops.h"
43 #include "wimlib/version.h"
44
45 #ifdef __WIN32__
46 #  include "wimlib/win32.h" /* for realpath() replacement */
47 #endif
48
49 #include <errno.h>
50 #include <fcntl.h>
51 #ifndef __WIN32__
52 #  include <langinfo.h>
53 #endif
54 #include <limits.h>
55 #include <stdarg.h>
56 #include <stdlib.h>
57 #include <unistd.h>
58
59 static int
60 wim_default_pack_compression_type(void)
61 {
62         return WIMLIB_COMPRESSION_TYPE_LZMS;
63 }
64
65 static u32
66 wim_default_pack_chunk_size(int ctype) {
67         switch (ctype) {
68         case WIMLIB_COMPRESSION_TYPE_LZMS:
69                 /* Note: WIMGAPI uses 1 << 26, but lower sizes are compatible.
70                  * */
71                 return 1U << 25; /* 33554432  */
72         default:
73                 return 1U << 15; /* 32768     */
74         }
75 }
76
77 static WIMStruct *
78 new_wim_struct(void)
79 {
80         WIMStruct *wim = CALLOC(1, sizeof(WIMStruct));
81         if (wim) {
82                 filedes_invalidate(&wim->in_fd);
83                 filedes_invalidate(&wim->out_fd);
84                 wim->out_pack_compression_type = wim_default_pack_compression_type();
85                 wim->out_pack_chunk_size = wim_default_pack_chunk_size(
86                                                 wim->out_pack_compression_type);
87                 INIT_LIST_HEAD(&wim->subwims);
88         }
89         return wim;
90 }
91
92 /* Determine if the chunk size is valid for the specified compression type.  */
93 static bool
94 wim_chunk_size_valid(u32 chunk_size, int ctype)
95 {
96         u32 order;
97
98         /* Chunk size is meaningless for uncompressed WIMs --- any value is
99          * okay.  */
100         if (ctype == WIMLIB_COMPRESSION_TYPE_NONE)
101                 return true;
102
103         /* Chunk size must be power of 2.  */
104         if (chunk_size == 0)
105                 return false;
106         order = bsr32(chunk_size);
107         if (chunk_size != 1U << order)
108                 return false;
109
110         /* Order        Size
111          * =====        ====
112          * 15           32768
113          * 16           65536
114          * 17           131072
115          * 18           262144
116          * 19           524288
117          * 20           1048576
118          * 21           2097152
119          * 22           4194304
120          * 23           8388608
121          * 24           16777216
122          * 25           33554432
123          * 26           67108864
124          */
125
126         /* See the documentation for the --chunk-size option of `wimlib-imagex
127          * capture' for information about allowed chunk sizes.  */
128         switch (ctype) {
129         case WIMLIB_COMPRESSION_TYPE_LZX:
130                 return order >= 15 && order <= 21;
131
132         case WIMLIB_COMPRESSION_TYPE_XPRESS:
133                 return order >= 12 && order <= 26;
134         case WIMLIB_COMPRESSION_TYPE_LZMS:
135                 return order >= 15 && order <= 30;
136         }
137         return false;
138 }
139
140 /* Return the default chunk size to use for the specified compression type.
141  *
142  * See notes above in wim_chunk_size_valid().  */
143 static u32
144 wim_default_chunk_size(int ctype)
145 {
146         switch (ctype) {
147         case WIMLIB_COMPRESSION_TYPE_LZMS:
148                 return 1U << 17; /* 131072  */
149         default:
150                 return 1U << 15; /* 32768   */
151         }
152 }
153
154 /*
155  * Calls a function on images in the WIM.  If @image is WIMLIB_ALL_IMAGES,
156  * @visitor is called on the WIM once for each image, with each image selected
157  * as the current image in turn.  If @image is a certain image, @visitor is
158  * called on the WIM only once, with that image selected.
159  */
160 int
161 for_image(WIMStruct *wim, int image, int (*visitor)(WIMStruct *))
162 {
163         int ret;
164         int start;
165         int end;
166         int i;
167
168         if (image == WIMLIB_ALL_IMAGES) {
169                 start = 1;
170                 end = wim->hdr.image_count;
171         } else if (image >= 1 && image <= wim->hdr.image_count) {
172                 start = image;
173                 end = image;
174         } else {
175                 return WIMLIB_ERR_INVALID_IMAGE;
176         }
177         for (i = start; i <= end; i++) {
178                 ret = select_wim_image(wim, i);
179                 if (ret != 0)
180                         return ret;
181                 ret = visitor(wim);
182                 if (ret != 0)
183                         return ret;
184         }
185         return 0;
186 }
187
188 /* API function documented in wimlib.h  */
189 WIMLIBAPI int
190 wimlib_create_new_wim(int ctype, WIMStruct **wim_ret)
191 {
192         WIMStruct *wim;
193         struct wim_lookup_table *table;
194         int ret;
195
196         wimlib_global_init(WIMLIB_INIT_FLAG_ASSUME_UTF8);
197
198         DEBUG("Creating new WIM with %"TS" compression.",
199               wimlib_get_compression_type_string(ctype));
200
201         /* Allocate the WIMStruct. */
202         wim = new_wim_struct();
203         if (wim == NULL)
204                 return WIMLIB_ERR_NOMEM;
205
206         ret = init_wim_header(&wim->hdr, ctype, wim_default_chunk_size(ctype));
207         if (ret)
208                 goto out_free;
209
210         table = new_lookup_table(9001);
211         if (table == NULL) {
212                 ret = WIMLIB_ERR_NOMEM;
213                 goto out_free;
214         }
215         wim->lookup_table = table;
216         wim->refcnts_ok = 1;
217         wim->compression_type = ctype;
218         wim->out_compression_type = ctype;
219         wim->chunk_size = wim->hdr.chunk_size;
220         wim->out_chunk_size = wim->hdr.chunk_size;
221         *wim_ret = wim;
222         return 0;
223 out_free:
224         FREE(wim);
225         return ret;
226 }
227
228 static void
229 destroy_image_metadata(struct wim_image_metadata *imd,
230                        struct wim_lookup_table *table,
231                        bool free_metadata_lte)
232 {
233         free_dentry_tree(imd->root_dentry, table);
234         imd->root_dentry = NULL;
235         free_wim_security_data(imd->security_data);
236         imd->security_data = NULL;
237
238         if (free_metadata_lte) {
239                 free_lookup_table_entry(imd->metadata_lte);
240                 imd->metadata_lte = NULL;
241         }
242         if (table == NULL) {
243                 struct wim_lookup_table_entry *lte, *tmp;
244                 list_for_each_entry_safe(lte, tmp, &imd->unhashed_streams, unhashed_list)
245                         free_lookup_table_entry(lte);
246         }
247         INIT_LIST_HEAD(&imd->unhashed_streams);
248         INIT_LIST_HEAD(&imd->inode_list);
249 #ifdef WITH_NTFS_3G
250         if (imd->ntfs_vol) {
251                 do_ntfs_umount(imd->ntfs_vol);
252                 imd->ntfs_vol = NULL;
253         }
254 #endif
255 }
256
257 void
258 put_image_metadata(struct wim_image_metadata *imd,
259                    struct wim_lookup_table *table)
260 {
261         if (imd && --imd->refcnt == 0) {
262                 destroy_image_metadata(imd, table, true);
263                 FREE(imd);
264         }
265 }
266
267 /* Appends the specified image metadata structure to the array of image metadata
268  * for a WIM, and increments the image count. */
269 int
270 append_image_metadata(WIMStruct *wim, struct wim_image_metadata *imd)
271 {
272         struct wim_image_metadata **imd_array;
273
274         DEBUG("Reallocating image metadata array for image_count = %u",
275               wim->hdr.image_count + 1);
276         imd_array = REALLOC(wim->image_metadata,
277                             sizeof(wim->image_metadata[0]) * (wim->hdr.image_count + 1));
278
279         if (imd_array == NULL)
280                 return WIMLIB_ERR_NOMEM;
281         wim->image_metadata = imd_array;
282         imd_array[wim->hdr.image_count++] = imd;
283         return 0;
284 }
285
286 struct wim_image_metadata *
287 new_image_metadata(void)
288 {
289         struct wim_image_metadata *imd;
290
291         imd = CALLOC(1, sizeof(*imd));
292         if (imd) {
293                 imd->refcnt = 1;
294                 INIT_LIST_HEAD(&imd->inode_list);
295                 INIT_LIST_HEAD(&imd->unhashed_streams);
296                 DEBUG("Created new image metadata (refcnt=1)");
297         } else {
298                 ERROR_WITH_ERRNO("Failed to allocate new image metadata structure");
299         }
300         return imd;
301 }
302
303 static struct wim_image_metadata **
304 new_image_metadata_array(unsigned num_images)
305 {
306         struct wim_image_metadata **imd_array;
307
308         DEBUG("Creating new image metadata array for %u images",
309               num_images);
310
311         imd_array = CALLOC(num_images, sizeof(imd_array[0]));
312
313         if (imd_array == NULL) {
314                 ERROR("Failed to allocate memory for %u image metadata structures",
315                       num_images);
316                 return NULL;
317         }
318         for (unsigned i = 0; i < num_images; i++) {
319                 imd_array[i] = new_image_metadata();
320                 if (imd_array[i] == NULL) {
321                         for (unsigned j = 0; j < i; j++)
322                                 put_image_metadata(imd_array[j], NULL);
323                         FREE(imd_array);
324                         return NULL;
325                 }
326         }
327         return imd_array;
328 }
329
330
331 /* Load the metadata for the specified WIM image into memory and set it as the
332  * WIMStruct's currently selected WIM image.  */
333 int
334 select_wim_image(WIMStruct *wim, int image)
335 {
336         struct wim_image_metadata *imd;
337         int ret;
338
339         DEBUG("Selecting image %d", image);
340
341         if (image == WIMLIB_NO_IMAGE) {
342                 ERROR("Invalid image: %d", WIMLIB_NO_IMAGE);
343                 return WIMLIB_ERR_INVALID_IMAGE;
344         }
345
346         if (image == wim->current_image)
347                 return 0;
348
349         if (image < 1 || image > wim->hdr.image_count) {
350                 ERROR("Cannot select image %d: There are only %u images",
351                       image, wim->hdr.image_count);
352                 return WIMLIB_ERR_INVALID_IMAGE;
353         }
354
355         if (!wim_has_metadata(wim)) {
356                 ERROR("\"%"TS"\" does not contain metadata resources!", wim->filename);
357                 if (wim->hdr.part_number != 1)
358                         ERROR("Specify the first part of the split WIM instead.");
359                 return WIMLIB_ERR_METADATA_NOT_FOUND;
360         }
361
362         /* If a valid image is currently selected, its metadata can be freed if
363          * it has not been modified.  */
364         if (wim->current_image != WIMLIB_NO_IMAGE) {
365                 imd = wim_get_current_image_metadata(wim);
366                 if (!imd->modified) {
367                         wimlib_assert(list_empty(&imd->unhashed_streams));
368                         DEBUG("Freeing image %u", wim->current_image);
369                         destroy_image_metadata(imd, NULL, false);
370                 }
371         }
372         wim->current_image = image;
373         imd = wim_get_current_image_metadata(wim);
374         if (imd->root_dentry || imd->modified) {
375                 ret = 0;
376         } else {
377                 ret = read_metadata_resource(wim, imd);
378                 if (ret)
379                         wim->current_image = WIMLIB_NO_IMAGE;
380         }
381         return ret;
382 }
383
384
385 /* API function documented in wimlib.h  */
386 WIMLIBAPI const tchar *
387 wimlib_get_compression_type_string(int ctype)
388 {
389         switch (ctype) {
390                 case WIMLIB_COMPRESSION_TYPE_NONE:
391                         return T("None");
392                 case WIMLIB_COMPRESSION_TYPE_LZX:
393                         return T("LZX");
394                 case WIMLIB_COMPRESSION_TYPE_XPRESS:
395                         return T("XPRESS");
396                 case WIMLIB_COMPRESSION_TYPE_LZMS:
397                         return T("LZMS");
398                 default:
399                         return T("Invalid");
400         }
401 }
402
403 /* API function documented in wimlib.h  */
404 WIMLIBAPI int
405 wimlib_resolve_image(WIMStruct *wim, const tchar *image_name_or_num)
406 {
407         tchar *p;
408         long image;
409         int i;
410
411         if (!image_name_or_num || !*image_name_or_num)
412                 return WIMLIB_NO_IMAGE;
413
414         if (!tstrcasecmp(image_name_or_num, T("all"))
415             || !tstrcasecmp(image_name_or_num, T("*")))
416                 return WIMLIB_ALL_IMAGES;
417         image = tstrtol(image_name_or_num, &p, 10);
418         if (p != image_name_or_num && *p == T('\0') && image > 0) {
419                 if (image > wim->hdr.image_count)
420                         return WIMLIB_NO_IMAGE;
421                 return image;
422         } else {
423                 for (i = 1; i <= wim->hdr.image_count; i++) {
424                         if (!tstrcmp(image_name_or_num,
425                                      wimlib_get_image_name(wim, i)))
426                                 return i;
427                 }
428                 return WIMLIB_NO_IMAGE;
429         }
430 }
431
432 /* API function documented in wimlib.h  */
433 WIMLIBAPI void
434 wimlib_print_available_images(const WIMStruct *wim, int image)
435 {
436         int first;
437         int last;
438         int i;
439         int n;
440         if (image == WIMLIB_ALL_IMAGES) {
441                 n = tprintf(T("Available Images:\n"));
442                 first = 1;
443                 last = wim->hdr.image_count;
444         } else if (image >= 1 && image <= wim->hdr.image_count) {
445                 n = tprintf(T("Information for Image %d\n"), image);
446                 first = image;
447                 last = image;
448         } else {
449                 tprintf(T("wimlib_print_available_images(): Invalid image %d"),
450                         image);
451                 return;
452         }
453         for (i = 0; i < n - 1; i++)
454                 tputchar(T('-'));
455         tputchar(T('\n'));
456         for (i = first; i <= last; i++)
457                 print_image_info(wim->wim_info, i);
458 }
459
460 /* API function documented in wimlib.h  */
461 WIMLIBAPI int
462 wimlib_get_wim_info(WIMStruct *wim, struct wimlib_wim_info *info)
463 {
464         memset(info, 0, sizeof(struct wimlib_wim_info));
465         memcpy(info->guid, wim->hdr.guid, WIMLIB_GUID_LEN);
466         info->image_count = wim->hdr.image_count;
467         info->boot_index = wim->hdr.boot_idx;
468         info->wim_version = wim->hdr.wim_version;
469         info->chunk_size = wim->chunk_size;
470         info->part_number = wim->hdr.part_number;
471         info->total_parts = wim->hdr.total_parts;
472         info->compression_type = wim->compression_type;
473         info->total_bytes = wim_info_get_total_bytes(wim->wim_info);
474         info->has_integrity_table = wim_has_integrity_table(wim);
475         info->opened_from_file = (wim->filename != NULL);
476         info->is_readonly = (wim->hdr.flags & WIM_HDR_FLAG_READONLY) ||
477                              (wim->hdr.total_parts != 1) ||
478                              (wim->filename && taccess(wim->filename, W_OK));
479         info->has_rpfix = (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX) != 0;
480         info->is_marked_readonly = (wim->hdr.flags & WIM_HDR_FLAG_READONLY) != 0;
481         info->write_in_progress = (wim->hdr.flags & WIM_HDR_FLAG_WRITE_IN_PROGRESS) != 0;
482         info->metadata_only = (wim->hdr.flags & WIM_HDR_FLAG_METADATA_ONLY) != 0;
483         info->resource_only = (wim->hdr.flags & WIM_HDR_FLAG_RESOURCE_ONLY) != 0;
484         info->spanned = (wim->hdr.flags & WIM_HDR_FLAG_SPANNED) != 0;
485         info->pipable = wim_is_pipable(wim);
486         return 0;
487 }
488
489 /* API function documented in wimlib.h  */
490 WIMLIBAPI int
491 wimlib_set_wim_info(WIMStruct *wim, const struct wimlib_wim_info *info, int which)
492 {
493         int ret;
494
495         if (which & ~(WIMLIB_CHANGE_READONLY_FLAG |
496                       WIMLIB_CHANGE_GUID |
497                       WIMLIB_CHANGE_BOOT_INDEX |
498                       WIMLIB_CHANGE_RPFIX_FLAG))
499                 return WIMLIB_ERR_INVALID_PARAM;
500
501         if (which & WIMLIB_CHANGE_READONLY_FLAG) {
502                 if (info->is_marked_readonly)
503                         wim->hdr.flags |= WIM_HDR_FLAG_READONLY;
504                 else
505                         wim->hdr.flags &= ~WIM_HDR_FLAG_READONLY;
506         }
507
508         if ((which & ~WIMLIB_CHANGE_READONLY_FLAG) == 0)
509                 return 0;
510
511         ret = can_modify_wim(wim);
512         if (ret)
513                 return ret;
514
515         if (which & WIMLIB_CHANGE_GUID)
516                 memcpy(wim->hdr.guid, info->guid, WIM_GUID_LEN);
517
518         if (which & WIMLIB_CHANGE_BOOT_INDEX) {
519                 if (info->boot_index > wim->hdr.image_count) {
520                         ERROR("%u is not 0 or a valid image in the WIM to mark as bootable",
521                               info->boot_index);
522                         return WIMLIB_ERR_INVALID_IMAGE;
523                 }
524                 wim->hdr.boot_idx = info->boot_index;
525         }
526
527         if (which & WIMLIB_CHANGE_RPFIX_FLAG) {
528                 if (info->has_rpfix)
529                         wim->hdr.flags |= WIM_HDR_FLAG_RP_FIX;
530                 else
531                         wim->hdr.flags &= ~WIM_HDR_FLAG_RP_FIX;
532         }
533         return 0;
534 }
535
536 static int
537 set_out_ctype(int ctype, u8 *out_ctype_p)
538 {
539         switch (ctype) {
540         case WIMLIB_COMPRESSION_TYPE_INVALID:
541                 break;
542         case WIMLIB_COMPRESSION_TYPE_NONE:
543         case WIMLIB_COMPRESSION_TYPE_LZX:
544         case WIMLIB_COMPRESSION_TYPE_XPRESS:
545         case WIMLIB_COMPRESSION_TYPE_LZMS:
546                 *out_ctype_p = ctype;
547                 return 0;
548         }
549         return WIMLIB_ERR_INVALID_COMPRESSION_TYPE;
550 }
551
552 /* API function documented in wimlib.h  */
553 WIMLIBAPI int
554 wimlib_set_output_compression_type(WIMStruct *wim, int ctype)
555 {
556         int ret = set_out_ctype(ctype, &wim->out_compression_type);
557         if (ret)
558                 return ret;
559
560         /* Reset the chunk size if it's no longer valid.  */
561         if (!wim_chunk_size_valid(wim->out_chunk_size, ctype))
562                 wim->out_chunk_size = wim_default_chunk_size(ctype);
563         return 0;
564 }
565
566 /* API function documented in wimlib.h  */
567 WIMLIBAPI int
568 wimlib_set_output_pack_compression_type(WIMStruct *wim, int ctype)
569 {
570         int ret = set_out_ctype(ctype, &wim->out_pack_compression_type);
571         if (ret)
572                 return ret;
573
574         /* Reset the chunk size if it's no longer valid.  */
575         if (!wim_chunk_size_valid(wim->out_pack_chunk_size, ctype))
576                 wim->out_pack_chunk_size = wim_default_pack_chunk_size(ctype);
577         return 0;
578 }
579
580 static int
581 set_out_chunk_size(u32 chunk_size, int ctype, u32 *out_chunk_size_p)
582 {
583         if (!wim_chunk_size_valid(chunk_size, ctype)) {
584                 ERROR("Invalid chunk size (%"PRIu32" bytes) "
585                       "for compression type %"TS"!",
586                       chunk_size,
587                       wimlib_get_compression_type_string(ctype));
588                 return WIMLIB_ERR_INVALID_CHUNK_SIZE;
589         }
590
591         *out_chunk_size_p = chunk_size;
592         return 0;
593 }
594
595 /* API function documented in wimlib.h  */
596 WIMLIBAPI int
597 wimlib_set_output_chunk_size(WIMStruct *wim, uint32_t chunk_size)
598 {
599         if (chunk_size == 0) {
600                 wim->out_chunk_size =
601                         wim_default_chunk_size(wim->out_compression_type);
602                 return 0;
603         }
604
605         return set_out_chunk_size(chunk_size,
606                                   wim->out_compression_type,
607                                   &wim->out_chunk_size);
608 }
609
610 /* API function documented in wimlib.h  */
611 WIMLIBAPI int
612 wimlib_set_output_pack_chunk_size(WIMStruct *wim, uint32_t chunk_size)
613 {
614         if (chunk_size == 0) {
615                 wim->out_pack_chunk_size =
616                         wim_default_pack_chunk_size(wim->out_pack_compression_type);
617                 return 0;
618         }
619
620         return set_out_chunk_size(chunk_size,
621                                   wim->out_pack_compression_type,
622                                   &wim->out_pack_chunk_size);
623 }
624
625 static int
626 open_wim_file(const tchar *filename, struct filedes *fd_ret)
627 {
628         int raw_fd;
629
630         raw_fd = topen(filename, O_RDONLY | O_BINARY);
631         if (raw_fd < 0) {
632                 ERROR_WITH_ERRNO("Can't open \"%"TS"\" read-only", filename);
633                 return WIMLIB_ERR_OPEN;
634         }
635         filedes_init(fd_ret, raw_fd);
636         return 0;
637 }
638
639 /*
640  * Begins the reading of a WIM file; opens the file and reads its header and
641  * lookup table, and optionally checks the integrity.
642  */
643 static int
644 begin_read(WIMStruct *wim, const void *wim_filename_or_fd,
645            int open_flags, wimlib_progress_func_t progress_func)
646 {
647         int ret;
648         int xml_num_images;
649         const tchar *wimfile;
650
651         if (open_flags & WIMLIB_OPEN_FLAG_FROM_PIPE) {
652                 wimfile = NULL;
653                 filedes_init(&wim->in_fd, *(const int*)wim_filename_or_fd);
654                 wim->in_fd.is_pipe = 1;
655         } else {
656                 wimfile = wim_filename_or_fd;
657                 ret = open_wim_file(wimfile, &wim->in_fd);
658                 if (ret)
659                         return ret;
660
661                 /* The absolute path to the WIM is requested so that
662                  * wimlib_overwrite() still works even if the process changes
663                  * its working directory.  This actually happens if a WIM is
664                  * mounted read-write, since the FUSE thread changes directory
665                  * to "/", and it needs to be able to find the WIM file again.
666                  *
667                  * This will break if the full path to the WIM changes in the
668                  * intervening time...
669                  *
670                  * Warning: in Windows native builds, realpath() calls the
671                  * replacement function in win32_replacements.c.
672                  */
673                 wim->filename = realpath(wimfile, NULL);
674                 if (wim->filename == NULL) {
675                         ERROR_WITH_ERRNO("Failed to resolve WIM filename");
676                         if (errno == ENOMEM)
677                                 return WIMLIB_ERR_NOMEM;
678                         else
679                                 return WIMLIB_ERR_OPEN;
680                 }
681         }
682
683         ret = read_wim_header(wim, &wim->hdr);
684         if (ret)
685                 return ret;
686
687         if (wim->hdr.flags & WIM_HDR_FLAG_WRITE_IN_PROGRESS) {
688                 WARNING("The WIM_HDR_FLAG_WRITE_IN_PROGRESS flag is set in the header of\n"
689                         "          \"%"TS"\".  It may be being changed by another process,\n"
690                         "          or a process may have crashed while writing the WIM.",
691                         wimfile);
692         }
693
694         if (open_flags & WIMLIB_OPEN_FLAG_WRITE_ACCESS) {
695                 ret = can_modify_wim(wim);
696                 if (ret)
697                         return ret;
698         }
699
700         if ((open_flags & WIMLIB_OPEN_FLAG_ERROR_IF_SPLIT) &&
701             (wim->hdr.total_parts != 1))
702                 return WIMLIB_ERR_IS_SPLIT_WIM;
703
704         DEBUG("According to header, WIM contains %u images", wim->hdr.image_count);
705
706         /* If the boot index is invalid, print a warning and set it to 0 */
707         if (wim->hdr.boot_idx > wim->hdr.image_count) {
708                 WARNING("In `%"TS"', image %u is marked as bootable, "
709                         "but there are only %u images in the WIM",
710                         wimfile, wim->hdr.boot_idx, wim->hdr.image_count);
711                 wim->hdr.boot_idx = 0;
712         }
713
714         /* Check and cache the compression type */
715         if (wim->hdr.flags & WIM_HDR_FLAG_COMPRESSION) {
716                 if (wim->hdr.flags & WIM_HDR_FLAG_COMPRESS_LZX) {
717                         wim->compression_type = WIMLIB_COMPRESSION_TYPE_LZX;
718                 } else if (wim->hdr.flags & (WIM_HDR_FLAG_COMPRESS_XPRESS |
719                                              WIM_HDR_FLAG_COMPRESS_XPRESS_2)) {
720                         wim->compression_type = WIMLIB_COMPRESSION_TYPE_XPRESS;
721                 } else if (wim->hdr.flags & WIM_HDR_FLAG_COMPRESS_LZMS) {
722                         wim->compression_type = WIMLIB_COMPRESSION_TYPE_LZMS;
723                 } else {
724                         ERROR("The compression flag is set on \"%"TS"\", but "
725                               "a flag for a recognized format is not",
726                               wimfile);
727                         return WIMLIB_ERR_INVALID_COMPRESSION_TYPE;
728                 }
729         } else {
730                 wim->compression_type = WIMLIB_COMPRESSION_TYPE_NONE;
731         }
732         wim->out_compression_type = wim->compression_type;
733
734         /* Check and cache the chunk size.  */
735         wim->chunk_size = wim->hdr.chunk_size;
736         wim->out_chunk_size = wim->chunk_size;
737         if (!wim_chunk_size_valid(wim->chunk_size, wim->compression_type)) {
738                 ERROR("Invalid chunk size (%"PRIu32" bytes) "
739                       "for compression type %"TS"!",
740                       wim->chunk_size,
741                       wimlib_get_compression_type_string(wim->compression_type));
742                 return WIMLIB_ERR_INVALID_CHUNK_SIZE;
743         }
744
745         if (open_flags & WIMLIB_OPEN_FLAG_CHECK_INTEGRITY) {
746                 ret = check_wim_integrity(wim, progress_func);
747                 if (ret == WIM_INTEGRITY_NONEXISTENT) {
748                         WARNING("No integrity information for `%"TS"'; skipping "
749                                 "integrity check.", wimfile);
750                 } else if (ret == WIM_INTEGRITY_NOT_OK) {
751                         ERROR("WIM is not intact! (Failed integrity check)");
752                         return WIMLIB_ERR_INTEGRITY;
753                 } else if (ret != WIM_INTEGRITY_OK) {
754                         return ret;
755                 }
756         }
757
758         if (wim->hdr.image_count != 0 && wim->hdr.part_number == 1) {
759                 wim->image_metadata = new_image_metadata_array(wim->hdr.image_count);
760                 if (wim->image_metadata == NULL)
761                         return WIMLIB_ERR_NOMEM;
762         }
763
764         if (open_flags & WIMLIB_OPEN_FLAG_FROM_PIPE) {
765                 wim->lookup_table = new_lookup_table(9001);
766                 if (wim->lookup_table == NULL)
767                         return WIMLIB_ERR_NOMEM;
768         } else {
769
770                 ret = read_wim_xml_data(wim);
771                 if (ret)
772                         return ret;
773
774                 xml_num_images = wim_info_get_num_images(wim->wim_info);
775                 if (xml_num_images != wim->hdr.image_count) {
776                         ERROR("In the file `%"TS"', there are %u <IMAGE> elements "
777                               "in the XML data,", wimfile, xml_num_images);
778                         ERROR("but %u images in the WIM!  There must be exactly one "
779                               "<IMAGE> element per image.", wim->hdr.image_count);
780                         return WIMLIB_ERR_IMAGE_COUNT;
781                 }
782
783                 ret = read_wim_lookup_table(wim);
784                 if (ret)
785                         return ret;
786
787                 DEBUG("Done beginning read of WIM file `%"TS"'.", wimfile);
788         }
789         return 0;
790 }
791
792 int
793 open_wim_as_WIMStruct(const void *wim_filename_or_fd, int open_flags,
794                       WIMStruct **wim_ret, wimlib_progress_func_t progress_func)
795 {
796         WIMStruct *wim;
797         int ret;
798
799         if (open_flags & WIMLIB_OPEN_FLAG_FROM_PIPE)
800                 DEBUG("Opening pipable WIM from file descriptor %d.", *(const int*)wim_filename_or_fd);
801         else
802                 DEBUG("Opening WIM file \"%"TS"\"", (const tchar*)wim_filename_or_fd);
803
804         wimlib_global_init(WIMLIB_INIT_FLAG_ASSUME_UTF8);
805
806         if (wim_ret == NULL)
807                 return WIMLIB_ERR_INVALID_PARAM;
808
809         wim = new_wim_struct();
810         if (wim == NULL)
811                 return WIMLIB_ERR_NOMEM;
812
813         ret = begin_read(wim, wim_filename_or_fd, open_flags, progress_func);
814         if (ret) {
815                 wimlib_free(wim);
816                 return ret;
817         }
818
819         DEBUG("Successfully opened WIM and created WIMStruct.");
820         *wim_ret = wim;
821         return 0;
822 }
823
824 /* API function documented in wimlib.h  */
825 WIMLIBAPI int
826 wimlib_open_wim(const tchar *wimfile, int open_flags,
827                 WIMStruct **wim_ret, wimlib_progress_func_t progress_func)
828 {
829         if (open_flags & ~(WIMLIB_OPEN_FLAG_CHECK_INTEGRITY |
830                            WIMLIB_OPEN_FLAG_ERROR_IF_SPLIT |
831                            WIMLIB_OPEN_FLAG_WRITE_ACCESS))
832                 return WIMLIB_ERR_INVALID_PARAM;
833
834         return open_wim_as_WIMStruct(wimfile, open_flags, wim_ret,
835                                      progress_func);
836 }
837
838 /* Checksum all streams that are unhashed (other than the metadata streams),
839  * merging them into the lookup table as needed.  This is a no-op unless the
840  * library has previously used to add or mount an image using the same
841  * WIMStruct. */
842 int
843 wim_checksum_unhashed_streams(WIMStruct *wim)
844 {
845         int ret;
846
847         if (!wim_has_metadata(wim))
848                 return 0;
849         for (int i = 0; i < wim->hdr.image_count; i++) {
850                 struct wim_lookup_table_entry *lte, *tmp;
851                 struct wim_image_metadata *imd = wim->image_metadata[i];
852                 image_for_each_unhashed_stream_safe(lte, tmp, imd) {
853                         struct wim_lookup_table_entry *new_lte;
854                         ret = hash_unhashed_stream(lte, wim->lookup_table, &new_lte);
855                         if (ret)
856                                 return ret;
857                         if (new_lte != lte)
858                                 free_lookup_table_entry(lte);
859                 }
860         }
861         return 0;
862 }
863
864 /*
865  * can_modify_wim - Check if a given WIM is writeable.  This is only the case if
866  * it meets the following three conditions:
867  *
868  * 1. Write access is allowed to the underlying file (if any) at the filesystem level.
869  * 2. The WIM is not part of a spanned set.
870  * 3. The WIM_HDR_FLAG_READONLY flag is not set in the WIM header.
871  *
872  * Return value is 0 if writable; WIMLIB_ERR_WIM_IS_READONLY otherwise.
873  */
874 int
875 can_modify_wim(WIMStruct *wim)
876 {
877         if (wim->filename) {
878                 if (taccess(wim->filename, W_OK)) {
879                         ERROR_WITH_ERRNO("Can't modify \"%"TS"\"", wim->filename);
880                         return WIMLIB_ERR_WIM_IS_READONLY;
881                 }
882         }
883         if (wim->hdr.total_parts != 1) {
884                 ERROR("Cannot modify \"%"TS"\": is part of a split WIM",
885                       wim->filename);
886                 return WIMLIB_ERR_WIM_IS_READONLY;
887         }
888         if (wim->hdr.flags & WIM_HDR_FLAG_READONLY) {
889                 ERROR("Cannot modify \"%"TS"\": is marked read-only",
890                       wim->filename);
891                 return WIMLIB_ERR_WIM_IS_READONLY;
892         }
893         return 0;
894 }
895
896 /*
897  * can_delete_from_wim - Check if files or images can be deleted from a given
898  * WIM file.
899  *
900  * This theoretically should be exactly the same as can_modify_wim(), but
901  * unfortunately, due to bugs in Microsoft's software that generate incorrect
902  * reference counts for some WIM resources, we need to run expensive
903  * verifications to make sure the reference counts are correct on all WIM
904  * resources.  Otherwise we might delete a WIM resource whose reference count
905  * has fallen to 0, but is actually still referenced somewhere.
906  */
907 int
908 can_delete_from_wim(WIMStruct *wim)
909 {
910         int ret;
911
912         ret = can_modify_wim(wim);
913         if (ret)
914                 return ret;
915         if (!wim->refcnts_ok) {
916                 ret = wim_recalculate_refcnts(wim);
917                 if (ret)
918                         return ret;
919         }
920         return 0;
921 }
922
923 /* API function documented in wimlib.h  */
924 WIMLIBAPI void
925 wimlib_free(WIMStruct *wim)
926 {
927         if (wim == NULL)
928                 return;
929
930         DEBUG("Freeing WIMStruct (filename=\"%"TS"\", image_count=%u)",
931               wim->filename, wim->hdr.image_count);
932
933         while (!list_empty(&wim->subwims)) {
934                 WIMStruct *subwim;
935
936                 subwim = list_entry(wim->subwims.next, WIMStruct, subwim_node);
937                 list_del(&subwim->subwim_node);
938                 DEBUG("Freeing subwim.");
939                 wimlib_free(subwim);
940         }
941
942         if (filedes_valid(&wim->in_fd))
943                 filedes_close(&wim->in_fd);
944         if (filedes_valid(&wim->out_fd))
945                 filedes_close(&wim->out_fd);
946
947         free_lookup_table(wim->lookup_table);
948
949         wimlib_free_decompressor(wim->decompressor);
950
951         FREE(wim->filename);
952         free_wim_info(wim->wim_info);
953         if (wim->image_metadata) {
954                 for (unsigned i = 0; i < wim->hdr.image_count; i++)
955                         put_image_metadata(wim->image_metadata[i], NULL);
956                 FREE(wim->image_metadata);
957         }
958         FREE(wim);
959 }
960
961 static bool
962 test_locale_ctype_utf8(void)
963 {
964 #ifdef __WIN32__
965         return false;
966 #else
967         char *ctype = nl_langinfo(CODESET);
968
969         return (!strstr(ctype, "UTF-8") ||
970                 !strstr(ctype, "UTF8") ||
971                 !strstr(ctype, "utf8") ||
972                 !strstr(ctype, "utf-8"));
973 #endif
974 }
975
976 /* API function documented in wimlib.h  */
977 WIMLIBAPI u32
978 wimlib_get_version(void)
979 {
980         return WIMLIB_VERSION_CODE;
981 }
982
983 /* API function documented in wimlib.h  */
984 WIMLIBAPI int
985 wimlib_global_init(int init_flags)
986 {
987         static bool already_inited = false;
988
989         if (already_inited)
990                 return 0;
991
992         if (init_flags & ~(WIMLIB_INIT_FLAG_ASSUME_UTF8 |
993                            WIMLIB_INIT_FLAG_DONT_ACQUIRE_PRIVILEGES |
994                            WIMLIB_INIT_FLAG_STRICT_CAPTURE_PRIVILEGES |
995                            WIMLIB_INIT_FLAG_STRICT_APPLY_PRIVILEGES |
996                            WIMLIB_INIT_FLAG_DEFAULT_CASE_SENSITIVE |
997                            WIMLIB_INIT_FLAG_DEFAULT_CASE_INSENSITIVE))
998                 return WIMLIB_ERR_INVALID_PARAM;
999
1000         libxml_global_init();
1001         if (!(init_flags & WIMLIB_INIT_FLAG_ASSUME_UTF8)) {
1002                 wimlib_mbs_is_utf8 = test_locale_ctype_utf8();
1003         #ifdef WITH_NTFS_3G
1004                 if (!wimlib_mbs_is_utf8)
1005                         libntfs3g_global_init();
1006         #endif
1007         }
1008 #ifdef __WIN32__
1009         {
1010                 int ret = win32_global_init(init_flags);
1011                 if (ret)
1012                         return ret;
1013         }
1014 #endif
1015         init_upcase();
1016         if (init_flags & WIMLIB_INIT_FLAG_DEFAULT_CASE_SENSITIVE)
1017                 default_ignore_case = false;
1018         else if (init_flags & WIMLIB_INIT_FLAG_DEFAULT_CASE_INSENSITIVE)
1019                 default_ignore_case = true;
1020         already_inited = true;
1021         return 0;
1022 }
1023
1024 /* API function documented in wimlib.h  */
1025 WIMLIBAPI void
1026 wimlib_global_cleanup(void)
1027 {
1028         libxml_global_cleanup();
1029         iconv_global_cleanup();
1030 #ifdef __WIN32__
1031         win32_global_cleanup();
1032 #endif
1033         cleanup_decompressor_params();
1034         cleanup_compressor_params();
1035 }