a1725a918880b84e79e088815fb0518ecbc54583
[wimlib] / src / extract.c
1 /*
2  * extract.c
3  *
4  * Support for extracting WIM images, or files or directories contained in a WIM
5  * image.
6  */
7
8 /*
9  * Copyright (C) 2012, 2013 Eric Biggers
10  *
11  * This file is part of wimlib, a library for working with WIM files.
12  *
13  * wimlib is free software; you can redistribute it and/or modify it under the
14  * terms of the GNU General Public License as published by the Free
15  * Software Foundation; either version 3 of the License, or (at your option)
16  * any later version.
17  *
18  * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
19  * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
20  * A PARTICULAR PURPOSE. See the GNU General Public License for more
21  * details.
22  *
23  * You should have received a copy of the GNU General Public License
24  * along with wimlib; if not, see http://www.gnu.org/licenses/.
25  */
26
27 /*
28  * This file provides the API functions wimlib_extract_image(),
29  * wimlib_extract_image_from_pipe(), wimlib_extract_files(),
30  * wimlib_extract_paths(), and wimlib_extract_pathlist().  Internally, all end
31  * up calling do_wimlib_extract_paths() and extract_trees().
32  *
33  * Although wimlib supports multiple extraction modes/backends (NTFS-3g, UNIX,
34  * Win32), this file does not itself have code to extract files or directories
35  * to any specific target; instead, it handles generic functionality and relies
36  * on lower-level callback functions declared in `struct apply_operations' to do
37  * the actual extraction.
38  */
39
40 #ifdef HAVE_CONFIG_H
41 #  include "config.h"
42 #endif
43
44 #include "wimlib/apply.h"
45 #include "wimlib/dentry.h"
46 #include "wimlib/encoding.h"
47 #include "wimlib/endianness.h"
48 #include "wimlib/error.h"
49 #include "wimlib/lookup_table.h"
50 #include "wimlib/metadata.h"
51 #include "wimlib/pathlist.h"
52 #include "wimlib/paths.h"
53 #include "wimlib/reparse.h"
54 #include "wimlib/resource.h"
55 #include "wimlib/security.h"
56 #ifdef __WIN32__
57 #  include "wimlib/win32.h" /* for realpath() equivalent */
58 #endif
59 #include "wimlib/xml.h"
60 #include "wimlib/wildcard.h"
61 #include "wimlib/wim.h"
62
63 #include <errno.h>
64 #include <fcntl.h>
65 #include <stdlib.h>
66 #include <sys/stat.h>
67 #include <unistd.h>
68
69 #define WIMLIB_EXTRACT_FLAG_MULTI_IMAGE 0x80000000
70 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE   0x40000000
71 #define WIMLIB_EXTRACT_FLAG_FILEMODE    0x20000000
72 #define WIMLIB_EXTRACT_FLAG_IMAGEMODE   0x10000000
73
74 /* Keep in sync with wimlib.h  */
75 #define WIMLIB_EXTRACT_MASK_PUBLIC                              \
76         (WIMLIB_EXTRACT_FLAG_NTFS                       |       \
77          WIMLIB_EXTRACT_FLAG_HARDLINK                   |       \
78          WIMLIB_EXTRACT_FLAG_SYMLINK                    |       \
79          WIMLIB_EXTRACT_FLAG_VERBOSE                    |       \
80          WIMLIB_EXTRACT_FLAG_SEQUENTIAL                 |       \
81          WIMLIB_EXTRACT_FLAG_UNIX_DATA                  |       \
82          WIMLIB_EXTRACT_FLAG_NO_ACLS                    |       \
83          WIMLIB_EXTRACT_FLAG_STRICT_ACLS                |       \
84          WIMLIB_EXTRACT_FLAG_RPFIX                      |       \
85          WIMLIB_EXTRACT_FLAG_NORPFIX                    |       \
86          WIMLIB_EXTRACT_FLAG_TO_STDOUT                  |       \
87          WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES  |       \
88          WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS         |       \
89          WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS          |       \
90          WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES         |       \
91          WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS            |       \
92          WIMLIB_EXTRACT_FLAG_RESUME                     |       \
93          WIMLIB_EXTRACT_FLAG_FILE_ORDER                 |       \
94          WIMLIB_EXTRACT_FLAG_GLOB_PATHS                 |       \
95          WIMLIB_EXTRACT_FLAG_STRICT_GLOB                |       \
96          WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES              |       \
97          WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)
98
99 static bool
100 dentry_in_list(const struct wim_dentry *dentry)
101 {
102         return dentry->extraction_list.next != NULL;
103 }
104
105 static inline bool
106 is_linked_extraction(const struct apply_ctx *ctx)
107 {
108         return 0 != (ctx->extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK |
109                                            WIMLIB_EXTRACT_FLAG_SYMLINK));
110 }
111
112 static inline bool
113 can_extract_named_data_streams(const struct apply_ctx *ctx)
114 {
115         return ctx->supported_features.named_data_streams &&
116                 !is_linked_extraction(ctx);
117 }
118 /* Inform library user of progress of stream extraction following the successful
119  * extraction of a copy of the stream specified by @lte.  */
120 static void
121 update_extract_progress(struct apply_ctx *ctx,
122                         const struct wim_lookup_table_entry *lte)
123 {
124         wimlib_progress_func_t progress_func = ctx->progress_func;
125         union wimlib_progress_info *progress = &ctx->progress;
126
127         progress->extract.completed_bytes += lte->size;
128         if (progress_func &&
129             progress->extract.completed_bytes >= ctx->next_progress)
130         {
131                 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, progress);
132                 if (progress->extract.completed_bytes >=
133                     progress->extract.total_bytes)
134                 {
135                         ctx->next_progress = ~0ULL;
136                 } else {
137                         ctx->next_progress += progress->extract.total_bytes / 128;
138                         if (ctx->next_progress > progress->extract.total_bytes)
139                                 ctx->next_progress = progress->extract.total_bytes;
140                 }
141         }
142 }
143
144 #ifndef __WIN32__
145 /* Extract a symbolic link (not directly as reparse data), handling fixing up
146  * the target of absolute symbolic links and updating the extract progress.
147  *
148  * @inode must specify the WIM inode for a symbolic link or junction reparse
149  * point.
150  *
151  * @lte_override overrides the resource used as the reparse data for the
152  * symbolic link.  */
153 static int
154 extract_symlink(const tchar *path, struct apply_ctx *ctx,
155                 struct wim_inode *inode,
156                 struct wim_lookup_table_entry *lte_override)
157 {
158         ssize_t bufsize = ctx->ops->path_max;
159         tchar target[bufsize];
160         tchar *buf = target;
161         tchar *fixed_target;
162         ssize_t sret;
163         int ret;
164
165         /* If absolute symbolic link fixups requested, reserve space in the link
166          * target buffer for the absolute path of the target directory.  */
167         if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX)
168         {
169                 buf += ctx->realtarget_nchars;
170                 bufsize -= ctx->realtarget_nchars;
171         }
172
173         /* Translate the WIM inode's reparse data into the link target.  */
174         sret = wim_inode_readlink(inode, buf, bufsize - 1, lte_override);
175         if (sret < 0) {
176                 errno = -sret;
177                 return WIMLIB_ERR_READLINK;
178         }
179         buf[sret] = '\0';
180
181         if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
182             buf[0] == '/')
183         {
184                 /* Fix absolute symbolic link target to point into the
185                  * actual extraction destination.  */
186                 tmemcpy(target, ctx->realtarget, ctx->realtarget_nchars);
187                 fixed_target = target;
188         } else {
189                 /* Keep same link target.  */
190                 fixed_target = buf;
191         }
192
193         /* Call into the apply_operations to create the symbolic link.  */
194         DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"",
195               path, fixed_target);
196         ret = ctx->ops->create_symlink(fixed_target, path, ctx);
197         if (ret) {
198                 ERROR_WITH_ERRNO("Failed to create symlink "
199                                  "\"%"TS"\" => \"%"TS"\"", path, fixed_target);
200                 return ret;
201         }
202
203         /* Account for reparse data consumed.  */
204         update_extract_progress(ctx,
205                                 (lte_override ? lte_override :
206                                       inode_unnamed_lte_resolved(inode)));
207         return 0;
208 }
209 #endif /* !__WIN32__ */
210
211 /* Create a file, directory, or symbolic link.  */
212 static int
213 extract_inode(const tchar *path, struct apply_ctx *ctx, struct wim_inode *inode)
214 {
215         int ret;
216
217 #ifndef __WIN32__
218         if (ctx->supported_features.symlink_reparse_points &&
219             !ctx->supported_features.reparse_points &&
220             inode_is_symlink(inode))
221         {
222                 ret = extract_symlink(path, ctx, inode, NULL);
223         } else
224 #endif /* !__WIN32__ */
225         if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
226                 ret = ctx->ops->create_directory(path, ctx, &inode->extract_cookie);
227                 if (ret) {
228                         ERROR_WITH_ERRNO("Failed to create the directory "
229                                          "\"%"TS"\"", path);
230                 }
231         } else if ((inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) &&
232                     ctx->ops->extract_encrypted_stream_creates_file &&
233                     ctx->supported_features.encrypted_files) {
234                 ret = ctx->ops->extract_encrypted_stream(
235                                 path, inode_unnamed_lte_resolved(inode), ctx);
236                 if (ret) {
237                         ERROR_WITH_ERRNO("Failed to create and extract "
238                                          "encrypted file \"%"TS"\"", path);
239                 }
240         } else {
241                 ret = ctx->ops->create_file(path, ctx, &inode->extract_cookie);
242                 if (ret) {
243                         ERROR_WITH_ERRNO("Failed to create the file "
244                                          "\"%"TS"\"", path);
245                 }
246         }
247         return ret;
248 }
249
250 static int
251 extract_hardlink(const tchar *oldpath, const tchar *newpath,
252                  struct apply_ctx *ctx)
253 {
254         int ret;
255
256         DEBUG("Creating hardlink \"%"TS"\" => \"%"TS"\"", newpath, oldpath);
257         ret = ctx->ops->create_hardlink(oldpath, newpath, ctx);
258         if (ret) {
259                 ERROR_WITH_ERRNO("Failed to create hardlink "
260                                  "\"%"TS"\" => \"%"TS"\"",
261                                  newpath, oldpath);
262         }
263         return ret;
264 }
265
266 #ifdef __WIN32__
267 static int
268 try_extract_rpfix(u8 *rpbuf,
269                   u16 *rpbuflen_p,
270                   const wchar_t *extract_root_realpath,
271                   unsigned extract_root_realpath_nchars)
272 {
273         struct reparse_data rpdata;
274         wchar_t *target;
275         size_t target_nchars;
276         size_t stripped_nchars;
277         wchar_t *stripped_target;
278         wchar_t stripped_target_nchars;
279         int ret;
280
281         utf16lechar *new_target;
282         utf16lechar *new_print_name;
283         size_t new_target_nchars;
284         size_t new_print_name_nchars;
285         utf16lechar *p;
286
287         ret = parse_reparse_data(rpbuf, *rpbuflen_p, &rpdata);
288         if (ret)
289                 return ret;
290
291         if (extract_root_realpath[0] == L'\0' ||
292             extract_root_realpath[1] != L':' ||
293             extract_root_realpath[2] != L'\\')
294                 return WIMLIB_ERR_REPARSE_POINT_FIXUP_FAILED;
295
296         ret = parse_substitute_name(rpdata.substitute_name,
297                                     rpdata.substitute_name_nbytes,
298                                     rpdata.rptag);
299         if (ret < 0)
300                 return 0;
301         stripped_nchars = ret;
302         target = rpdata.substitute_name;
303         target_nchars = rpdata.substitute_name_nbytes / sizeof(utf16lechar);
304         stripped_target = target + stripped_nchars;
305         stripped_target_nchars = target_nchars - stripped_nchars;
306
307         new_target = alloca((6 + extract_root_realpath_nchars +
308                              stripped_target_nchars) * sizeof(utf16lechar));
309
310         p = new_target;
311         if (stripped_nchars == 6) {
312                 /* Include \??\ prefix if it was present before */
313                 p = wmempcpy(p, L"\\??\\", 4);
314         }
315
316         /* Print name excludes the \??\ if present. */
317         new_print_name = p;
318         if (stripped_nchars != 0) {
319                 /* Get drive letter from real path to extract root, if a drive
320                  * letter was present before. */
321                 *p++ = extract_root_realpath[0];
322                 *p++ = extract_root_realpath[1];
323         }
324         /* Copy the rest of the extract root */
325         p = wmempcpy(p, extract_root_realpath + 2, extract_root_realpath_nchars - 2);
326
327         /* Append the stripped target */
328         p = wmempcpy(p, stripped_target, stripped_target_nchars);
329         new_target_nchars = p - new_target;
330         new_print_name_nchars = p - new_print_name;
331
332         if (new_target_nchars * sizeof(utf16lechar) >= REPARSE_POINT_MAX_SIZE ||
333             new_print_name_nchars * sizeof(utf16lechar) >= REPARSE_POINT_MAX_SIZE)
334                 return WIMLIB_ERR_REPARSE_POINT_FIXUP_FAILED;
335
336         rpdata.substitute_name = new_target;
337         rpdata.substitute_name_nbytes = new_target_nchars * sizeof(utf16lechar);
338         rpdata.print_name = new_print_name;
339         rpdata.print_name_nbytes = new_print_name_nchars * sizeof(utf16lechar);
340         return make_reparse_buffer(&rpdata, rpbuf, rpbuflen_p);
341 }
342 #endif /* __WIN32__ */
343
344 /* Set reparse data on extracted file or directory that has
345  * FILE_ATTRIBUTE_REPARSE_POINT set.  */
346 static int
347 extract_reparse_data(const tchar *path, struct apply_ctx *ctx,
348                      struct wim_inode *inode,
349                      struct wim_lookup_table_entry *lte_override)
350 {
351         int ret;
352         u8 rpbuf[REPARSE_POINT_MAX_SIZE];
353         u16 rpbuflen;
354
355         ret = wim_inode_get_reparse_data(inode, rpbuf, &rpbuflen, lte_override);
356         if (ret)
357                 goto error;
358
359 #ifdef __WIN32__
360         /* Fix up target of absolute symbolic link or junction points so
361          * that they point into the actual extraction target.  */
362         if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
363             (inode->i_reparse_tag == WIM_IO_REPARSE_TAG_SYMLINK ||
364              inode->i_reparse_tag == WIM_IO_REPARSE_TAG_MOUNT_POINT) &&
365             !inode->i_not_rpfixed)
366         {
367                 ret = try_extract_rpfix(rpbuf, &rpbuflen, ctx->realtarget,
368                                         ctx->realtarget_nchars);
369                 if (ret && !(ctx->extract_flags &
370                              WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS))
371                 {
372                         WARNING("Reparse point fixup of \"%"TS"\" "
373                                 "failed", path);
374                         ret = 0;
375                 }
376                 if (ret)
377                         goto error;
378         }
379 #endif
380
381         ret = ctx->ops->set_reparse_data(path, rpbuf, rpbuflen, ctx);
382
383         /* On Windows, the SeCreateSymbolicLink privilege is required to create
384          * symbolic links.  To be more friendly towards non-Administrator users,
385          * we merely warn the user if symbolic links cannot be created due to
386          * insufficient permissions or privileges, unless
387          * WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS was provided.  */
388 #ifdef __WIN32__
389         if (ret && inode_is_symlink(inode) &&
390             (errno == EACCES || errno == EPERM) &&
391             !(ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS))
392         {
393                 WARNING("Can't set reparse data on \"%"TS"\": "
394                         "Access denied!\n"
395                         "          You may be trying to "
396                         "extract a symbolic link without the\n"
397                         "          SeCreateSymbolicLink privilege, "
398                         "which by default non-Administrator\n"
399                         "          accounts do not have.",
400                         path);
401                 ret = 0;
402         }
403 #endif
404         if (ret)
405                 goto error;
406
407         /* Account for reparse data consumed.  */
408         update_extract_progress(ctx,
409                                 (lte_override ? lte_override :
410                                       inode_unnamed_lte_resolved(inode)));
411         return 0;
412
413 error:
414         ERROR_WITH_ERRNO("Failed to set reparse data on \"%"TS"\"", path);
415         return ret;
416 }
417
418 /*
419  * Extract zero or more streams to a file.
420  *
421  * This function operates slightly differently depending on whether @lte_spec is
422  * NULL or not.  When @lte_spec is NULL, the behavior is to extract the default
423  * file contents (unnamed stream), and, if named data streams are supported in
424  * the extract mode and volume, any named data streams.  When @lte_spec is not
425  * NULL, the behavior is to extract only all copies of the stream @lte_spec, and
426  * in addition use @lte_spec to set the reparse data or create the symbolic link
427  * if appropriate.
428  *
429  * @path
430  *      Path to file to extract (as can be passed to apply_operations
431  *      functions).
432  * @ctx
433  *      Apply context.
434  * @dentry
435  *      WIM dentry that corresponds to the file being extracted.
436  * @lte_spec
437  *      If non-NULL, specifies the lookup table entry for a stream to extract,
438  *      and only that stream will be extracted (although there may be more than
439  *      one instance of it).
440  * @lte_override
441  *      Used only if @lte_spec != NULL; it is passed to the extraction functions
442  *      rather than @lte_spec, allowing the location of the stream to be
443  *      overridden.  (This is used when the WIM is being read from a nonseekable
444  *      file, such as a pipe, when streams need to be used more than once; each
445  *      such stream is extracted to a temporary file.)
446  */
447 static int
448 extract_streams(const tchar *path, struct apply_ctx *ctx,
449                 struct wim_dentry *dentry,
450                 struct wim_lookup_table_entry *lte_spec,
451                 struct wim_lookup_table_entry *lte_override)
452 {
453         struct wim_inode *inode = dentry->d_inode;
454         struct wim_lookup_table_entry *lte;
455         file_spec_t file_spec;
456         int ret;
457
458         if (dentry->was_linked)
459                 return 0;
460
461 #ifdef ENABLE_DEBUG
462         if (lte_spec) {
463                 char sha1_str[100];
464                 char *p = sha1_str;
465                 for (unsigned i = 0; i < SHA1_HASH_SIZE; i++)
466                         p += sprintf(p, "%02x", lte_override->hash[i]);
467                 DEBUG("Extracting stream SHA1=%s to \"%"TS"\"",
468                       sha1_str, path, inode->i_ino);
469         } else {
470                 DEBUG("Extracting streams to \"%"TS"\"", path, inode->i_ino);
471         }
472 #endif
473
474         if (ctx->ops->uses_cookies)
475                 file_spec.cookie = inode->extract_cookie;
476         else
477                 file_spec.path = path;
478
479         /* Unnamed data stream.  */
480         lte = inode_unnamed_lte_resolved(inode);
481         if (lte && (!lte_spec || lte == lte_spec)) {
482                 if (lte_spec)
483                         lte = lte_override;
484                 if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
485                                              FILE_ATTRIBUTE_REPARSE_POINT)))
486                 {
487                         if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED &&
488                             ctx->supported_features.encrypted_files) {
489                                 if (!ctx->ops->extract_encrypted_stream_creates_file) {
490                                         ret = ctx->ops->extract_encrypted_stream(
491                                                                 path, lte, ctx);
492                                         if (ret)
493                                                 goto error;
494                                 }
495                         } else {
496                                 ret = ctx->ops->extract_unnamed_stream(
497                                                         file_spec, lte, ctx);
498                                 if (ret)
499                                         goto error;
500                         }
501                         update_extract_progress(ctx, lte);
502                 }
503                 else if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT)
504                 {
505                         ret = 0;
506                         if (ctx->supported_features.reparse_points)
507                                 ret = extract_reparse_data(path, ctx, inode, lte);
508                 #ifndef __WIN32__
509                         else if ((inode_is_symlink(inode) &&
510                                   ctx->supported_features.symlink_reparse_points))
511                                 ret = extract_symlink(path, ctx, inode, lte);
512                 #endif
513                         if (ret)
514                                 return ret;
515                 }
516         }
517
518         /* Named data streams.  */
519         if (can_extract_named_data_streams(ctx)) {
520                 for (u16 i = 0; i < inode->i_num_ads; i++) {
521                         struct wim_ads_entry *entry = &inode->i_ads_entries[i];
522
523                         if (!ads_entry_is_named_stream(entry))
524                                 continue;
525                         lte = entry->lte;
526                         if (!lte)
527                                 continue;
528                         if (lte_spec && lte_spec != lte)
529                                 continue;
530                         if (lte_spec)
531                                 lte = lte_override;
532                         ret = ctx->ops->extract_named_stream(file_spec, entry->stream_name,
533                                                              entry->stream_name_nbytes / 2,
534                                                              lte, ctx);
535                         if (ret)
536                                 goto error;
537                         update_extract_progress(ctx, lte);
538                 }
539         }
540         return 0;
541
542 error:
543         ERROR_WITH_ERRNO("Failed to extract data of \"%"TS"\"", path);
544         return ret;
545 }
546
547 /* Set attributes on an extracted file or directory if supported by the
548  * extraction mode.  */
549 static int
550 extract_file_attributes(const tchar *path, struct apply_ctx *ctx,
551                         struct wim_dentry *dentry, unsigned pass)
552 {
553         int ret;
554
555         if (ctx->ops->set_file_attributes &&
556             !(ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES) &&
557             !(dentry == ctx->target_dentry && ctx->root_dentry_is_special)) {
558                 u32 attributes = dentry->d_inode->i_attributes;
559
560                 /* Clear unsupported attributes.  */
561                 attributes &= ctx->supported_attributes_mask;
562
563                 if ((attributes & FILE_ATTRIBUTE_DIRECTORY &&
564                      !ctx->supported_features.encrypted_directories) ||
565                     (!(attributes & FILE_ATTRIBUTE_DIRECTORY) &&
566                      !ctx->supported_features.encrypted_files))
567                 {
568                         attributes &= ~FILE_ATTRIBUTE_ENCRYPTED;
569                 }
570
571                 if (attributes == 0)
572                         attributes = FILE_ATTRIBUTE_NORMAL;
573
574                 ret = ctx->ops->set_file_attributes(path, attributes, ctx, pass);
575                 if (ret) {
576                         ERROR_WITH_ERRNO("Failed to set attributes on "
577                                          "\"%"TS"\"", path);
578                         return ret;
579                 }
580         }
581         return 0;
582 }
583
584
585 /* Set or remove the short (DOS) name on an extracted file or directory if
586  * supported by the extraction mode.  Since DOS names are unimportant and it's
587  * easy to run into problems setting them on Windows (SetFileShortName()
588  * requires SE_RESTORE privilege, which only the Administrator can request, and
589  * also requires DELETE access to the file), failure is ignored unless
590  * WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES is set.  */
591 static int
592 extract_short_name(const tchar *path, struct apply_ctx *ctx,
593                    struct wim_dentry *dentry)
594 {
595         int ret;
596
597         /* The root of the dentry tree being extracted may not be extracted to
598          * its original name, so its short name should be ignored.  */
599         if (dentry == ctx->target_dentry)
600                 return 0;
601
602         if (ctx->supported_features.short_names) {
603                 ret = ctx->ops->set_short_name(path,
604                                                dentry->short_name,
605                                                dentry->short_name_nbytes / 2,
606                                                ctx);
607                 if (ret && (ctx->extract_flags &
608                             WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES))
609                 {
610                         ERROR_WITH_ERRNO("Failed to set short name of "
611                                          "\"%"TS"\"", path);
612                         return ret;
613                 }
614         }
615         return 0;
616 }
617
618 /* Set security descriptor, UNIX data, or neither on an extracted file, taking
619  * into account the current extraction mode and flags.  */
620 static int
621 extract_security(const tchar *path, struct apply_ctx *ctx,
622                  struct wim_dentry *dentry)
623 {
624         int ret;
625         struct wim_inode *inode = dentry->d_inode;
626
627         if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS)
628                 return 0;
629
630         if ((ctx->target_dentry == dentry) && ctx->root_dentry_is_special)
631                 return 0;
632
633 #ifndef __WIN32__
634         if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
635                 struct wimlib_unix_data data;
636
637                 ret = inode_get_unix_data(inode, &data, NULL);
638                 if (ret < 0)
639                         ret = 0;
640                 else if (ret == 0)
641                         ret = ctx->ops->set_unix_data(path, &data, ctx);
642                 if (ret) {
643                         if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) {
644                                 ERROR_WITH_ERRNO("Failed to set UNIX owner, "
645                                                  "group, and/or mode on "
646                                                  "\"%"TS"\"", path);
647                                 return ret;
648                         } else {
649                                 WARNING_WITH_ERRNO("Failed to set UNIX owner, "
650                                                    "group, and/or/mode on "
651                                                    "\"%"TS"\"", path);
652                         }
653                 }
654         }
655         else
656 #endif /* __WIN32__ */
657         if (ctx->supported_features.security_descriptors &&
658             inode->i_security_id != -1)
659         {
660                 const struct wim_security_data *sd;
661                 const u8 *desc;
662                 size_t desc_size;
663
664                 sd = wim_const_security_data(ctx->wim);
665                 desc = sd->descriptors[inode->i_security_id];
666                 desc_size = sd->sizes[inode->i_security_id];
667
668                 ret = ctx->ops->set_security_descriptor(path, desc,
669                                                         desc_size, ctx);
670                 if (ret) {
671                         if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) {
672                                 ERROR_WITH_ERRNO("Failed to set security "
673                                                  "descriptor on \"%"TS"\"", path);
674                                 return ret;
675                         } else {
676                         #if 0
677                                 if (errno != EACCES) {
678                                         WARNING_WITH_ERRNO("Failed to set "
679                                                            "security descriptor "
680                                                            "on \"%"TS"\"", path);
681                                 }
682                         #endif
683                                 ctx->no_security_descriptors++;
684                         }
685                 }
686         }
687         return 0;
688 }
689
690 /* Set timestamps on an extracted file.  Failure is warning-only unless
691  * WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS is set.  */
692 static int
693 extract_timestamps(const tchar *path, struct apply_ctx *ctx,
694                    struct wim_dentry *dentry)
695 {
696         struct wim_inode *inode = dentry->d_inode;
697         int ret;
698
699         if ((ctx->target_dentry == dentry) && ctx->root_dentry_is_special)
700                 return 0;
701
702         if (ctx->ops->set_timestamps) {
703                 ret = ctx->ops->set_timestamps(path,
704                                                inode->i_creation_time,
705                                                inode->i_last_write_time,
706                                                inode->i_last_access_time,
707                                                ctx);
708                 if (ret) {
709                         if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) {
710                                 ERROR_WITH_ERRNO("Failed to set timestamps "
711                                                  "on \"%"TS"\"", path);
712                                 return ret;
713                         } else {
714                                 WARNING_WITH_ERRNO("Failed to set timestamps "
715                                                    "on \"%"TS"\"", path);
716                         }
717                 }
718         }
719         return 0;
720 }
721
722 /* Check whether the extraction of a dentry should be skipped completely.  */
723 static bool
724 dentry_is_supported(struct wim_dentry *dentry,
725                     const struct wim_features *supported_features)
726 {
727         struct wim_inode *inode = dentry->d_inode;
728
729         if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
730                 return supported_features->reparse_points ||
731                         (inode_is_symlink(inode) &&
732                          supported_features->symlink_reparse_points);
733         }
734         if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
735                 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
736                         return supported_features->encrypted_directories != 0;
737                 else
738                         return supported_features->encrypted_files != 0;
739         }
740         return true;
741 }
742
743 /* Given a WIM dentry to extract, build the path to which to extract it, in the
744  * format understood by the callbacks in the apply_operations being used.
745  *
746  * Write the resulting path into @path, which must have room for at least
747  * ctx->ops->path_max characters.
748  *
749  * Return %true if successful; %false if this WIM dentry doesn't actually need
750  * to be extracted or if the calculated path exceeds ctx->ops->max_path
751  * characters.
752  *
753  * This function clobbers the tmp_list member of @dentry and its ancestors up
754  * until the extraction root.  */
755 static bool
756 build_extraction_path(tchar path[], struct wim_dentry *dentry,
757                       const struct apply_ctx *ctx)
758 {
759         size_t path_nchars;
760         LIST_HEAD(ancestor_list);
761         tchar *p = path;
762         const tchar *target_prefix;
763         size_t target_prefix_nchars;
764         struct wim_dentry *d;
765
766         path_nchars = ctx->ops->path_prefix_nchars;
767
768         if (ctx->ops->requires_realtarget_in_paths) {
769                 target_prefix        = ctx->realtarget;
770                 target_prefix_nchars = ctx->realtarget_nchars;
771         } else if (ctx->ops->requires_target_in_paths) {
772                 target_prefix        = ctx->target;
773                 target_prefix_nchars = ctx->target_nchars;
774         } else {
775                 target_prefix        = NULL;
776                 target_prefix_nchars = 0;
777         }
778         path_nchars += target_prefix_nchars;
779
780         for (d = dentry; d != ctx->target_dentry; d = d->parent) {
781                 if (!dentry_in_list(d))
782                         break;
783
784                 path_nchars += d->extraction_name_nchars + 1;
785                 list_add(&d->tmp_list, &ancestor_list);
786         }
787
788         path_nchars++; /* null terminator */
789
790         if (path_nchars > ctx->ops->path_max) {
791                 WARNING("\"%"TS"\": Path too long to extract",
792                         dentry_full_path(dentry));
793                 return false;
794         }
795
796         p = tmempcpy(p, ctx->ops->path_prefix, ctx->ops->path_prefix_nchars);
797         p = tmempcpy(p, target_prefix, target_prefix_nchars);
798         list_for_each_entry(d, &ancestor_list, tmp_list) {
799                 *p++ = ctx->ops->path_separator;
800                 p = tmempcpy(p, d->extraction_name, d->extraction_name_nchars);
801         }
802         *p++ = T('\0');
803         wimlib_assert(p - path == path_nchars);
804         return true;
805 }
806
807 static unsigned
808 get_num_path_components(const tchar *path, tchar path_separator)
809 {
810         unsigned num_components = 0;
811 #ifdef __WIN32__
812         /* Ignore drive letter.  */
813         if (path[0] != L'\0' && path[1] == L':')
814                 path += 2;
815 #endif
816
817         while (*path) {
818                 while (*path == path_separator)
819                         path++;
820                 if (*path)
821                         num_components++;
822                 while (*path && *path != path_separator)
823                         path++;
824         }
825         return num_components;
826 }
827
828 static int
829 extract_multiimage_symlink(const tchar *oldpath, const tchar *newpath,
830                            struct apply_ctx *ctx, struct wim_dentry *dentry)
831 {
832         size_t num_raw_path_components;
833         const struct wim_dentry *d;
834         size_t num_target_path_components;
835         tchar *p;
836         const tchar *p_old;
837         int ret;
838
839         num_raw_path_components = 0;
840         for (d = dentry; d != ctx->target_dentry; d = d->parent)
841                 num_raw_path_components++;
842
843         if (ctx->ops->requires_realtarget_in_paths)
844                 num_target_path_components = get_num_path_components(ctx->realtarget,
845                                                                      ctx->ops->path_separator);
846         else if (ctx->ops->requires_target_in_paths)
847                 num_target_path_components = get_num_path_components(ctx->target,
848                                                                      ctx->ops->path_separator);
849         else
850                 num_target_path_components = 0;
851
852         if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE) {
853                 wimlib_assert(num_target_path_components > 0);
854                 num_raw_path_components++;
855                 num_target_path_components--;
856         }
857
858         p_old = oldpath + ctx->ops->path_prefix_nchars;
859 #ifdef __WIN32__
860         if (p_old[0] != L'\0' && p_old[1] == ':')
861                 p_old += 2;
862 #endif
863         while (*p_old == ctx->ops->path_separator)
864                 p_old++;
865         while (--num_target_path_components) {
866                 while (*p_old != ctx->ops->path_separator)
867                         p_old++;
868                 while (*p_old == ctx->ops->path_separator)
869                         p_old++;
870         }
871
872         tchar symlink_target[tstrlen(p_old) + 3 * num_raw_path_components + 1];
873
874         p = &symlink_target[0];
875         while (num_raw_path_components--) {
876                 *p++ = '.';
877                 *p++ = '.';
878                 *p++ = ctx->ops->path_separator;
879         }
880         tstrcpy(p, p_old);
881         DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"",
882               newpath, symlink_target);
883         ret = ctx->ops->create_symlink(symlink_target, newpath, ctx);
884         if (ret) {
885                 ERROR_WITH_ERRNO("Failed to create symlink "
886                                  "\"%"TS"\" => \"%"TS"\"",
887                                  newpath, symlink_target);
888         }
889         return ret;
890 }
891
892 /* Create the "skeleton" of an extracted file or directory.  Don't yet extract
893  * data streams, reparse data (including symbolic links), timestamps, and
894  * security descriptors.  Basically, everything that doesn't require reading
895  * non-metadata resources from the WIM file and isn't delayed until the final
896  * pass.  */
897 static int
898 do_dentry_extract_skeleton(tchar path[], struct wim_dentry *dentry,
899                            struct apply_ctx *ctx)
900 {
901         struct wim_inode *inode = dentry->d_inode;
902         int ret;
903         const tchar *oldpath;
904
905         if (unlikely(is_linked_extraction(ctx))) {
906                 struct wim_lookup_table_entry *unnamed_lte;
907
908                 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
909                 if (unnamed_lte && unnamed_lte->extracted_file) {
910                         oldpath = unnamed_lte->extracted_file;
911                         if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK)
912                                 goto hardlink;
913                         else
914                                 goto symlink;
915                 }
916         }
917
918         /* Create hard link if this dentry corresponds to an already-extracted
919          * inode.  */
920         if (inode->i_extracted_file) {
921                 oldpath = inode->i_extracted_file;
922                 goto hardlink;
923         }
924
925         /* Skip symlinks unless they can be extracted as reparse points rather
926          * than created directly.  */
927         if (inode_is_symlink(inode) && !ctx->supported_features.reparse_points)
928                 return 0;
929
930         /* Create this file or directory unless it's the extraction root, which
931          * was already created if necessary.  */
932         if (dentry != ctx->target_dentry) {
933                 ret = extract_inode(path, ctx, inode);
934                 if (ret)
935                         return ret;
936         }
937
938         /* Create empty named data streams.  */
939         if (can_extract_named_data_streams(ctx)) {
940                 for (u16 i = 0; i < inode->i_num_ads; i++) {
941                         file_spec_t file_spec;
942                         struct wim_ads_entry *entry = &inode->i_ads_entries[i];
943
944                         if (!ads_entry_is_named_stream(entry))
945                                 continue;
946                         if (entry->lte)
947                                 continue;
948                         if (ctx->ops->uses_cookies)
949                                 file_spec.cookie = inode->extract_cookie;
950                         else
951                                 file_spec.path = path;
952                         ret = ctx->ops->extract_named_stream(file_spec,
953                                                              entry->stream_name,
954                                                              entry->stream_name_nbytes / 2,
955                                                              entry->lte, ctx);
956                         if (ret) {
957                                 ERROR_WITH_ERRNO("\"%"TS"\": failed to create "
958                                                  "empty named data stream",
959                                                  path);
960                                 return ret;
961                         }
962                 }
963         }
964
965         /* Set file attributes (if supported).  */
966         ret = extract_file_attributes(path, ctx, dentry, 0);
967         if (ret)
968                 return ret;
969
970         /* Set or remove file short name (if supported).  */
971         ret = extract_short_name(path, ctx, dentry);
972         if (ret)
973                 return ret;
974
975         /* If inode has multiple links and hard links are supported in this
976          * extraction mode and volume, save the path to the extracted file in
977          * case it's needed to create a hard link.  */
978         if (unlikely(is_linked_extraction(ctx))) {
979                 struct wim_lookup_table_entry *unnamed_lte;
980
981                 unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode);
982                 if (unnamed_lte) {
983                         unnamed_lte->extracted_file = TSTRDUP(path);
984                         if (!unnamed_lte->extracted_file)
985                                 return WIMLIB_ERR_NOMEM;
986                 }
987         } else if (inode->i_nlink > 1 && ctx->supported_features.hard_links) {
988                 inode->i_extracted_file = TSTRDUP(path);
989                 if (!inode->i_extracted_file)
990                         return WIMLIB_ERR_NOMEM;
991         }
992         return 0;
993
994 symlink:
995         ret = extract_multiimage_symlink(oldpath, path, ctx, dentry);
996         if (ret)
997                 return ret;
998         dentry->was_linked = 1;
999         return 0;
1000
1001 hardlink:
1002         ret = extract_hardlink(oldpath, path, ctx);
1003         if (ret)
1004                 return ret;
1005         dentry->was_linked = 1;
1006         return 0;
1007 }
1008
1009 /* This is a wrapper around do_dentry_extract_skeleton() that handles building
1010  * the path, doing short name reordering.  This is also idempotent; dentries
1011  * already processed have skeleton_extracted set and no action is taken.  See
1012  * apply_operations.requires_short_name_reordering for more details about short
1013  * name reordering.  */
1014 static int
1015 dentry_extract_skeleton(struct wim_dentry *dentry, struct apply_ctx *ctx)
1016 {
1017         tchar path[ctx->ops->path_max];
1018         struct wim_dentry *orig_dentry;
1019         struct wim_dentry *other_dentry;
1020         int ret;
1021
1022         if (dentry->skeleton_extracted)
1023                 return 0;
1024
1025         orig_dentry = NULL;
1026         if (ctx->supported_features.short_names
1027             && ctx->ops->requires_short_name_reordering
1028             && !dentry_has_short_name(dentry)
1029             && !dentry->d_inode->i_dos_name_extracted)
1030         {
1031                 inode_for_each_dentry(other_dentry, dentry->d_inode) {
1032                         if (dentry_has_short_name(other_dentry)
1033                             && !other_dentry->skeleton_extracted
1034                             && dentry_in_list(other_dentry))
1035                         {
1036                                 DEBUG("Creating %"TS" before %"TS" "
1037                                       "to guarantee correct DOS name extraction",
1038                                       dentry_full_path(other_dentry),
1039                                       dentry_full_path(dentry));
1040                                 orig_dentry = dentry;
1041                                 dentry = other_dentry;
1042                                 break;
1043                         }
1044                 }
1045         }
1046 again:
1047         if (!build_extraction_path(path, dentry, ctx))
1048                 return 0;
1049         ret = do_dentry_extract_skeleton(path, dentry, ctx);
1050         if (ret)
1051                 return ret;
1052
1053         dentry->skeleton_extracted = 1;
1054
1055         if (orig_dentry) {
1056                 dentry = orig_dentry;
1057                 orig_dentry = NULL;
1058                 goto again;
1059         }
1060         dentry->d_inode->i_dos_name_extracted = 1;
1061         return 0;
1062 }
1063
1064 /* Create a file or directory, then immediately extract all streams.  The WIM
1065  * may not be read sequentially by this function.  */
1066 static int
1067 dentry_extract(struct wim_dentry *dentry, struct apply_ctx *ctx)
1068 {
1069         tchar path[ctx->ops->path_max];
1070         int ret;
1071
1072         ret = dentry_extract_skeleton(dentry, ctx);
1073         if (ret)
1074                 return ret;
1075
1076         if (!build_extraction_path(path, dentry, ctx))
1077                 return 0;
1078
1079         return extract_streams(path, ctx, dentry, NULL, NULL);
1080 }
1081
1082 /* Finish extracting a file, directory, or symbolic link by setting file
1083  * security and timestamps.  */
1084 static int
1085 dentry_extract_final(struct wim_dentry *dentry, struct apply_ctx *ctx)
1086 {
1087         int ret;
1088         tchar path[ctx->ops->path_max];
1089
1090         if (!build_extraction_path(path, dentry, ctx))
1091                 return 0;
1092
1093         ret = extract_security(path, ctx, dentry);
1094         if (ret)
1095                 return ret;
1096
1097         if (ctx->ops->requires_final_set_attributes_pass) {
1098                 /* Set file attributes (if supported).  */
1099                 ret = extract_file_attributes(path, ctx, dentry, 1);
1100                 if (ret)
1101                         return ret;
1102         }
1103
1104         return extract_timestamps(path, ctx, dentry);
1105 }
1106
1107 static int
1108 extract_structure(struct list_head *dentry_list, struct apply_ctx *ctx)
1109 {
1110         struct wim_dentry *dentry;
1111         int ret;
1112
1113         list_for_each_entry(dentry, dentry_list, extraction_list) {
1114                 ret = dentry_extract_skeleton(dentry, ctx);
1115                 if (ret)
1116                         return ret;
1117         }
1118         return 0;
1119 }
1120
1121 static int
1122 extract_dir_structure(struct list_head *dentry_list, struct apply_ctx *ctx)
1123 {
1124         struct wim_dentry *dentry;
1125         int ret;
1126
1127         list_for_each_entry(dentry, dentry_list, extraction_list) {
1128                 if (dentry_is_directory(dentry)) {
1129                         ret = dentry_extract_skeleton(dentry, ctx);
1130                         if (ret)
1131                                 return ret;
1132                 }
1133         }
1134         return 0;
1135 }
1136
1137 static int
1138 extract_dentries(struct list_head *dentry_list, struct apply_ctx *ctx)
1139 {
1140         struct wim_dentry *dentry;
1141         int ret;
1142
1143         list_for_each_entry(dentry, dentry_list, extraction_list) {
1144                 ret = dentry_extract(dentry, ctx);
1145                 if (ret)
1146                         return ret;
1147         }
1148         return 0;
1149 }
1150
1151 static int
1152 extract_final_metadata(struct list_head *dentry_list, struct apply_ctx *ctx)
1153 {
1154         struct wim_dentry *dentry;
1155         int ret;
1156
1157         list_for_each_entry_reverse(dentry, dentry_list, extraction_list) {
1158                 ret = dentry_extract_final(dentry, ctx);
1159                 if (ret)
1160                         return ret;
1161         }
1162         return 0;
1163 }
1164
1165 /* Creates a temporary file opened for writing.  The open file descriptor is
1166  * returned in @fd_ret and its name is returned in @name_ret (dynamically
1167  * allocated).  */
1168 static int
1169 create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
1170 {
1171         tchar *name;
1172         int raw_fd;
1173
1174 retry:
1175         name = ttempnam(NULL, T("wimlib"));
1176         if (name == NULL) {
1177                 ERROR_WITH_ERRNO("Failed to create temporary filename");
1178                 return WIMLIB_ERR_NOMEM;
1179         }
1180
1181         raw_fd = topen(name, O_WRONLY | O_CREAT | O_EXCL | O_BINARY, 0600);
1182
1183         if (raw_fd < 0) {
1184                 if (errno == EEXIST) {
1185                         FREE(name);
1186                         goto retry;
1187                 }
1188                 ERROR_WITH_ERRNO("Failed to open temporary file \"%"TS"\"", name);
1189                 FREE(name);
1190                 return WIMLIB_ERR_OPEN;
1191         }
1192
1193         filedes_init(fd_ret, raw_fd);
1194         *name_ret = name;
1195         return 0;
1196 }
1197
1198 /* Extract all instances of the stream @lte that are being extracted in this
1199  * call of extract_tree(), but actually read the stream data from @lte_override.
1200  */
1201 static int
1202 extract_stream_instances(struct wim_lookup_table_entry *lte,
1203                          struct wim_lookup_table_entry *lte_override,
1204                          struct apply_ctx *ctx)
1205 {
1206         struct wim_dentry **lte_dentries;
1207         tchar path[ctx->ops->path_max];
1208         size_t i;
1209         int ret;
1210
1211         if (lte->out_refcnt <= ARRAY_LEN(lte->inline_lte_dentries))
1212                 lte_dentries = lte->inline_lte_dentries;
1213         else
1214                 lte_dentries = lte->lte_dentries;
1215
1216         for (i = 0; i < lte->out_refcnt; i++) {
1217                 struct wim_dentry *dentry = lte_dentries[i];
1218
1219                 if (dentry->tmp_flag)
1220                         continue;
1221                 if (!build_extraction_path(path, dentry, ctx))
1222                         continue;
1223                 ret = extract_streams(path, ctx, dentry, lte, lte_override);
1224                 if (ret)
1225                         goto out_clear_tmp_flags;
1226                 dentry->tmp_flag = 1;
1227         }
1228         ret = 0;
1229 out_clear_tmp_flags:
1230         for (i = 0; i < lte->out_refcnt; i++)
1231                 lte_dentries[i]->tmp_flag = 0;
1232         return ret;
1233 }
1234
1235 /* Determine whether the specified stream needs to be extracted to a temporary
1236  * file or not.
1237  *
1238  * @lte->out_refcnt specifies the number of instances of this stream that must
1239  * be extracted.
1240  *
1241  * @is_partial_res is %true if this stream is just one of multiple in a single
1242  * WIM resource being extracted.  */
1243 static bool
1244 need_tmpfile_to_extract(struct wim_lookup_table_entry *lte,
1245                         bool is_partial_res)
1246 {
1247         /* Temporary file is always required when reading a partial resource,
1248          * since in that case we retrieve all the contained streams in one pass.
1249          * */
1250         if (is_partial_res)
1251                 return true;
1252
1253         /* Otherwise we don't need a temporary file if only a single instance of
1254          * the stream is needed.  */
1255         if (lte->out_refcnt == 1)
1256                 return false;
1257
1258         wimlib_assert(lte->out_refcnt >= 2);
1259
1260         /* We also don't need a temporary file if random access to the stream is
1261          * allowed.  */
1262         if (lte->resource_location != RESOURCE_IN_WIM ||
1263             filedes_is_seekable(&lte->rspec->wim->in_fd))
1264                 return false;
1265
1266         return true;
1267 }
1268
1269 static int
1270 begin_extract_stream_to_tmpfile(struct wim_lookup_table_entry *lte,
1271                                 bool is_partial_res,
1272                                 void *_ctx)
1273 {
1274         struct apply_ctx *ctx = _ctx;
1275         int ret;
1276
1277         if (!need_tmpfile_to_extract(lte, is_partial_res)) {
1278                 DEBUG("Temporary file not needed "
1279                       "for stream (size=%"PRIu64")", lte->size);
1280                 ret = extract_stream_instances(lte, lte, ctx);
1281                 if (ret)
1282                         return ret;
1283
1284                 return BEGIN_STREAM_STATUS_SKIP_STREAM;
1285         }
1286
1287         DEBUG("Temporary file needed for stream (size=%"PRIu64")", lte->size);
1288         return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
1289 }
1290
1291 static int
1292 end_extract_stream_to_tmpfile(struct wim_lookup_table_entry *lte,
1293                               int status, void *_ctx)
1294 {
1295         struct apply_ctx *ctx = _ctx;
1296         struct wim_lookup_table_entry lte_override;
1297         int ret;
1298         int errno_save = errno;
1299
1300         ret = filedes_close(&ctx->tmpfile_fd);
1301
1302         if (status) {
1303                 ret = status;
1304                 errno = errno_save;
1305                 goto out_delete_tmpfile;
1306         }
1307
1308         if (ret) {
1309                 ERROR_WITH_ERRNO("Error writing temporary file %"TS, ctx->tmpfile_name);
1310                 ret = WIMLIB_ERR_WRITE;
1311                 goto out_delete_tmpfile;
1312         }
1313
1314         /* Now that a full stream has been extracted to a temporary file,
1315          * extract all instances of it to the actual target.  */
1316
1317         memcpy(&lte_override, lte, sizeof(struct wim_lookup_table_entry));
1318         lte_override.resource_location = RESOURCE_IN_FILE_ON_DISK;
1319         lte_override.file_on_disk = ctx->tmpfile_name;
1320
1321         ret = extract_stream_instances(lte, &lte_override, ctx);
1322
1323 out_delete_tmpfile:
1324         errno_save = errno;
1325         tunlink(ctx->tmpfile_name);
1326         FREE(ctx->tmpfile_name);
1327         errno = errno_save;
1328         return ret;
1329 }
1330
1331 /* Extracts a list of streams (ctx.stream_list), assuming that the directory
1332  * structure and empty files were already created.  This relies on the
1333  * per-`struct wim_lookup_table_entry' list of dentries that reference each
1334  * stream that was constructed earlier.  */
1335 static int
1336 extract_stream_list(struct apply_ctx *ctx)
1337 {
1338         struct read_stream_list_callbacks cbs = {
1339                 .begin_stream           = begin_extract_stream_to_tmpfile,
1340                 .begin_stream_ctx       = ctx,
1341                 .consume_chunk          = extract_chunk_to_fd,
1342                 .consume_chunk_ctx      = &ctx->tmpfile_fd,
1343                 .end_stream             = end_extract_stream_to_tmpfile,
1344                 .end_stream_ctx         = ctx,
1345         };
1346         return read_stream_list(&ctx->stream_list,
1347                                 offsetof(struct wim_lookup_table_entry, extraction_list),
1348                                 &cbs, VERIFY_STREAM_HASHES);
1349 }
1350
1351 #define PWM_ALLOW_WIM_HDR 0x00001
1352 #define PWM_SILENT_EOF    0x00002
1353
1354 /* Read the header from a stream in a pipable WIM.  */
1355 static int
1356 read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte,
1357                        struct wim_resource_spec *rspec,
1358                        int flags, struct wim_header_disk *hdr_ret)
1359 {
1360         union {
1361                 struct pwm_stream_hdr stream_hdr;
1362                 struct wim_header_disk pwm_hdr;
1363         } buf;
1364         struct wim_reshdr reshdr;
1365         int ret;
1366
1367         ret = full_read(&pwm->in_fd, &buf.stream_hdr, sizeof(buf.stream_hdr));
1368         if (ret)
1369                 goto read_error;
1370
1371         if ((flags & PWM_ALLOW_WIM_HDR) && buf.stream_hdr.magic == PWM_MAGIC) {
1372                 BUILD_BUG_ON(sizeof(buf.pwm_hdr) < sizeof(buf.stream_hdr));
1373                 ret = full_read(&pwm->in_fd, &buf.stream_hdr + 1,
1374                                 sizeof(buf.pwm_hdr) - sizeof(buf.stream_hdr));
1375
1376                 if (ret)
1377                         goto read_error;
1378                 lte->resource_location = RESOURCE_NONEXISTENT;
1379                 memcpy(hdr_ret, &buf.pwm_hdr, sizeof(buf.pwm_hdr));
1380                 return 0;
1381         }
1382
1383         if (le64_to_cpu(buf.stream_hdr.magic) != PWM_STREAM_MAGIC) {
1384                 ERROR("Data read on pipe is invalid (expected stream header).");
1385                 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
1386         }
1387
1388         copy_hash(lte->hash, buf.stream_hdr.hash);
1389
1390         reshdr.size_in_wim = 0;
1391         reshdr.flags = le32_to_cpu(buf.stream_hdr.flags);
1392         reshdr.offset_in_wim = pwm->in_fd.offset;
1393         reshdr.uncompressed_size = le64_to_cpu(buf.stream_hdr.uncompressed_size);
1394         wim_res_hdr_to_spec(&reshdr, pwm, rspec);
1395         lte_bind_wim_resource_spec(lte, rspec);
1396         lte->flags = rspec->flags;
1397         lte->size = rspec->uncompressed_size;
1398         lte->offset_in_res = 0;
1399         return 0;
1400
1401 read_error:
1402         if (ret != WIMLIB_ERR_UNEXPECTED_END_OF_FILE || !(flags & PWM_SILENT_EOF))
1403                 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
1404         return ret;
1405 }
1406
1407 static int
1408 extract_streams_from_pipe(struct apply_ctx *ctx)
1409 {
1410         struct wim_lookup_table_entry *found_lte;
1411         struct wim_resource_spec *rspec;
1412         struct wim_lookup_table_entry *needed_lte;
1413         struct wim_lookup_table *lookup_table;
1414         struct wim_header_disk pwm_hdr;
1415         int ret;
1416         int pwm_flags;
1417
1418         ret = WIMLIB_ERR_NOMEM;
1419         found_lte = new_lookup_table_entry();
1420         if (found_lte == NULL)
1421                 goto out;
1422
1423         rspec = MALLOC(sizeof(struct wim_resource_spec));
1424         if (rspec == NULL)
1425                 goto out_free_found_lte;
1426
1427         lookup_table = ctx->wim->lookup_table;
1428         pwm_flags = PWM_ALLOW_WIM_HDR;
1429         if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME))
1430                 pwm_flags |= PWM_SILENT_EOF;
1431         memcpy(ctx->progress.extract.guid, ctx->wim->hdr.guid, WIM_GID_LEN);
1432         ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
1433         ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
1434         if (ctx->progress_func)
1435                 ctx->progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN,
1436                                    &ctx->progress);
1437         while (ctx->num_streams_remaining) {
1438                 if (found_lte->resource_location != RESOURCE_NONEXISTENT)
1439                         lte_unbind_wim_resource_spec(found_lte);
1440                 ret = read_pwm_stream_header(ctx->wim, found_lte, rspec,
1441                                              pwm_flags, &pwm_hdr);
1442                 if (ret) {
1443                         if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE &&
1444                             (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME))
1445                         {
1446                                 goto resume_done;
1447                         }
1448                         goto out_free_found_lte;
1449                 }
1450
1451                 if ((found_lte->resource_location != RESOURCE_NONEXISTENT)
1452                     && !(found_lte->flags & WIM_RESHDR_FLAG_METADATA)
1453                     && (needed_lte = lookup_stream(lookup_table, found_lte->hash))
1454                     && (needed_lte->out_refcnt))
1455                 {
1456                         tchar *tmpfile_name = NULL;
1457                         struct wim_lookup_table_entry *lte_override;
1458                         struct wim_lookup_table_entry tmpfile_lte;
1459
1460                         needed_lte->offset_in_res = found_lte->offset_in_res;
1461                         needed_lte->flags = found_lte->flags;
1462                         needed_lte->size = found_lte->size;
1463
1464                         lte_unbind_wim_resource_spec(found_lte);
1465                         lte_bind_wim_resource_spec(needed_lte, rspec);
1466
1467                         if (needed_lte->out_refcnt > 1) {
1468
1469                                 struct filedes tmpfile_fd;
1470
1471                                 /* Extract stream to temporary file.  */
1472                                 ret = create_temporary_file(&tmpfile_fd, &tmpfile_name);
1473                                 if (ret) {
1474                                         lte_unbind_wim_resource_spec(needed_lte);
1475                                         goto out_free_found_lte;
1476                                 }
1477
1478                                 ret = extract_full_stream_to_fd(needed_lte,
1479                                                                 &tmpfile_fd);
1480                                 if (ret) {
1481                                         filedes_close(&tmpfile_fd);
1482                                         goto delete_tmpfile;
1483                                 }
1484
1485                                 if (filedes_close(&tmpfile_fd)) {
1486                                         ERROR_WITH_ERRNO("Error writing to temporary "
1487                                                          "file \"%"TS"\"", tmpfile_name);
1488                                         ret = WIMLIB_ERR_WRITE;
1489                                         goto delete_tmpfile;
1490                                 }
1491                                 memcpy(&tmpfile_lte, needed_lte,
1492                                        sizeof(struct wim_lookup_table_entry));
1493                                 tmpfile_lte.resource_location = RESOURCE_IN_FILE_ON_DISK;
1494                                 tmpfile_lte.file_on_disk = tmpfile_name;
1495                                 lte_override = &tmpfile_lte;
1496                         } else {
1497                                 lte_override = needed_lte;
1498                         }
1499
1500                         ret = extract_stream_instances(needed_lte, lte_override, ctx);
1501                 delete_tmpfile:
1502                         lte_unbind_wim_resource_spec(needed_lte);
1503                         if (tmpfile_name) {
1504                                 tunlink(tmpfile_name);
1505                                 FREE(tmpfile_name);
1506                         }
1507                         if (ret)
1508                                 goto out_free_found_lte;
1509                         ctx->num_streams_remaining--;
1510                 } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) {
1511                         ret = skip_wim_stream(found_lte);
1512                         if (ret)
1513                                 goto out_free_found_lte;
1514                 } else {
1515                         u16 part_number = le16_to_cpu(pwm_hdr.part_number);
1516                         u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
1517
1518                         if (part_number != ctx->progress.extract.part_number ||
1519                             total_parts != ctx->progress.extract.total_parts ||
1520                             memcmp(pwm_hdr.guid, ctx->progress.extract.guid,
1521                                    WIM_GID_LEN))
1522                         {
1523                                 ctx->progress.extract.part_number = part_number;
1524                                 ctx->progress.extract.total_parts = total_parts;
1525                                 memcpy(ctx->progress.extract.guid,
1526                                        pwm_hdr.guid, WIM_GID_LEN);
1527                                 if (ctx->progress_func) {
1528                                         ctx->progress_func(
1529                                                 WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN,
1530                                                            &ctx->progress);
1531                                 }
1532
1533                         }
1534                 }
1535         }
1536         ret = 0;
1537 out_free_found_lte:
1538         if (found_lte->resource_location != RESOURCE_IN_WIM)
1539                 FREE(rspec);
1540         free_lookup_table_entry(found_lte);
1541 out:
1542         return ret;
1543
1544 resume_done:
1545         /* TODO */
1546         return 0;
1547 }
1548
1549 /* Extract a WIM dentry to standard output.
1550  *
1551  * This obviously doesn't make sense in all cases.  We return an error if the
1552  * dentry does not correspond to a regular file.  Otherwise we extract the
1553  * unnamed data stream only.  */
1554 static int
1555 extract_dentry_to_stdout(struct wim_dentry *dentry,
1556                          const struct wim_lookup_table *lookup_table)
1557 {
1558         struct wim_inode *inode = dentry->d_inode;
1559         struct wim_lookup_table_entry *lte;
1560         struct filedes _stdout;
1561
1562         if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
1563                                    FILE_ATTRIBUTE_DIRECTORY))
1564         {
1565                 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
1566                       "extracted to standard output", dentry_full_path(dentry));
1567                 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
1568         }
1569
1570         lte = inode_unnamed_lte(inode, lookup_table);
1571         if (!lte) {
1572                 const u8 *hash = inode_unnamed_stream_hash(inode);
1573                 if (!is_zero_hash(hash))
1574                         return stream_not_found_error(inode, hash);
1575                 return 0;
1576         }
1577
1578         filedes_init(&_stdout, STDOUT_FILENO);
1579         return extract_full_stream_to_fd(lte, &_stdout);
1580 }
1581
1582 static int
1583 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
1584                            const struct wim_lookup_table *lookup_table)
1585 {
1586         for (size_t i = 0; i < num_dentries; i++) {
1587                 int ret = extract_dentry_to_stdout(dentries[i], lookup_table);
1588                 if (ret)
1589                         return ret;
1590         }
1591         return 0;
1592 }
1593
1594 /**********************************************************************/
1595
1596 /*
1597  * Removes duplicate dentries from the array.
1598  *
1599  * Returns the new number of dentries, packed at the front of the array.
1600  */
1601 static size_t
1602 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
1603 {
1604         size_t i, j = 0;
1605         for (i = 0; i < num_trees; i++) {
1606                 if (!trees[i]->tmp_flag) {
1607                         /* Found distinct dentry.  */
1608                         trees[i]->tmp_flag = 1;
1609                         trees[j++] = trees[i];
1610                 }
1611         }
1612         for (i = 0; i < j; i++)
1613                 trees[i]->tmp_flag = 0;
1614         return j;
1615 }
1616
1617 /*
1618  * Remove dentries that are descendants of other dentries in the array.
1619  *
1620  * Returns the new number of dentries, packed at the front of the array.
1621  */
1622 static size_t
1623 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
1624 {
1625         size_t i, j = 0;
1626         for (i = 0; i < num_trees; i++)
1627                 trees[i]->tmp_flag = 1;
1628         for (i = 0; i < num_trees; i++) {
1629                 struct wim_dentry *d = trees[i];
1630                 while (!dentry_is_root(d)) {
1631                         d = d->parent;
1632                         if (d->tmp_flag)
1633                                 goto tree_contained;
1634                 }
1635                 trees[j++] = trees[i];
1636                 continue;
1637
1638         tree_contained:
1639                 trees[i]->tmp_flag = 0;
1640         }
1641
1642         for (i = 0; i < j; i++)
1643                 trees[i]->tmp_flag = 0;
1644         return j;
1645 }
1646
1647 static int
1648 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
1649 {
1650         struct list_head *dentry_list = _dentry_list;
1651         list_add_tail(&dentry->extraction_list, dentry_list);
1652         return 0;
1653 }
1654
1655 static void
1656 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
1657 {
1658         dentry->extraction_list = (struct list_head){NULL, NULL};
1659 }
1660
1661 static void
1662 dentry_delete_from_list(struct wim_dentry *dentry)
1663 {
1664         list_del(&dentry->extraction_list);
1665         dentry_reset_extraction_list_node(dentry);
1666 }
1667
1668 static int
1669 do_dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
1670 {
1671         dentry_delete_from_list(dentry);
1672         return 0;
1673 }
1674
1675 /*
1676  * Build the preliminary list of dentries to be extracted.
1677  *
1678  * The list maintains the invariant that if d1 and d2 are in the list and d1 is
1679  * an ancestor of d2, then d1 appears before d2 in the list.
1680  */
1681 static void
1682 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
1683                   size_t num_trees, bool add_ancestors)
1684 {
1685         INIT_LIST_HEAD(dentry_list);
1686
1687         /* Add the trees recursively.  */
1688         for (size_t i = 0; i < num_trees; i++)
1689                 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
1690
1691         /* If requested, add ancestors of the trees.  */
1692         if (add_ancestors) {
1693                 for (size_t i = 0; i < num_trees; i++) {
1694                         struct wim_dentry *dentry = trees[i];
1695                         struct wim_dentry *ancestor;
1696                         struct list_head *place_after;
1697
1698                         if (dentry_is_root(dentry))
1699                                 continue;
1700
1701                         place_after = dentry_list;
1702                         ancestor = dentry;
1703                         do {
1704                                 ancestor = ancestor->parent;
1705                                 if (dentry_in_list(ancestor)) {
1706                                         place_after = &ancestor->extraction_list;
1707                                         break;
1708                                 }
1709                         } while (!dentry_is_root(ancestor));
1710
1711                         ancestor = dentry;
1712                         do {
1713                                 ancestor = ancestor->parent;
1714                                 if (dentry_in_list(ancestor))
1715                                         break;
1716                                 list_add(&ancestor->extraction_list, place_after);
1717                         } while (!dentry_is_root(ancestor));
1718                 }
1719         }
1720 }
1721
1722 static const struct apply_operations *
1723 select_apply_operations(int extract_flags)
1724 {
1725 #ifdef WITH_NTFS_3G
1726         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1727                 return &ntfs_3g_apply_ops;
1728 #endif
1729 #ifdef __WIN32__
1730         return &win32_apply_ops;
1731 #else
1732         return &unix_apply_ops;
1733 #endif
1734 }
1735
1736 #ifdef __WIN32__
1737 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
1738 #else
1739 static const utf16lechar replacement_char = cpu_to_le16('?');
1740 #endif
1741
1742 static bool
1743 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
1744 {
1745         size_t i;
1746
1747         if (num_chars == 0)
1748                 return true;
1749         for (i = 0; i < num_chars; i++) {
1750                 switch (name[i]) {
1751         #ifdef __WIN32__
1752                 case cpu_to_le16('\\'):
1753                 case cpu_to_le16(':'):
1754                 case cpu_to_le16('*'):
1755                 case cpu_to_le16('?'):
1756                 case cpu_to_le16('"'):
1757                 case cpu_to_le16('<'):
1758                 case cpu_to_le16('>'):
1759                 case cpu_to_le16('|'):
1760         #endif
1761                 case cpu_to_le16('/'):
1762                 case cpu_to_le16('\0'):
1763                         if (fix)
1764                                 name[i] = replacement_char;
1765                         else
1766                                 return false;
1767                 }
1768         }
1769
1770 #ifdef __WIN32__
1771         if (name[num_chars - 1] == cpu_to_le16(' ') ||
1772             name[num_chars - 1] == cpu_to_le16('.'))
1773         {
1774                 if (fix)
1775                         name[num_chars - 1] = replacement_char;
1776                 else
1777                         return false;
1778         }
1779 #endif
1780         return true;
1781 }
1782
1783 static int
1784 dentry_calculate_extraction_name(struct wim_dentry *dentry,
1785                                  struct apply_ctx *ctx)
1786 {
1787         int ret;
1788
1789         if (dentry == ctx->target_dentry)
1790                 return 0;
1791
1792         if (!dentry_is_supported(dentry, &ctx->supported_features))
1793                 goto skip_dentry;
1794
1795         if (!ctx->ops->supports_case_sensitive_filenames)
1796         {
1797                 struct wim_dentry *other;
1798                 list_for_each_entry(other, &dentry->case_insensitive_conflict_list,
1799                                     case_insensitive_conflict_list)
1800                 {
1801                         if (dentry_in_list(other)) {
1802                                 if (ctx->extract_flags &
1803                                     WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
1804                                         WARNING("\"%"TS"\" has the same "
1805                                                 "case-insensitive name as "
1806                                                 "\"%"TS"\"; extracting "
1807                                                 "dummy name instead",
1808                                                 dentry_full_path(dentry),
1809                                                 dentry_full_path(other));
1810                                         goto out_replace;
1811                                 } else {
1812                                         WARNING("Not extracting \"%"TS"\": "
1813                                                 "has same case-insensitive "
1814                                                 "name as \"%"TS"\"",
1815                                                 dentry_full_path(dentry),
1816                                                 dentry_full_path(other));
1817                                         goto skip_dentry;
1818                                 }
1819                         }
1820                 }
1821         }
1822
1823         if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
1824 #if TCHAR_IS_UTF16LE
1825                 dentry->extraction_name = dentry->file_name;
1826                 dentry->extraction_name_nchars = dentry->file_name_nbytes / 2;
1827                 return 0;
1828 #else
1829                 return utf16le_to_tstr(dentry->file_name,
1830                                        dentry->file_name_nbytes,
1831                                        &dentry->extraction_name,
1832                                        &dentry->extraction_name_nchars);
1833 #endif
1834         } else {
1835                 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
1836                 {
1837                         WARNING("\"%"TS"\" has an invalid filename "
1838                                 "that is not supported on this platform; "
1839                                 "extracting dummy name instead",
1840                                 dentry_full_path(dentry));
1841                         goto out_replace;
1842                 } else {
1843                         WARNING("Not extracting \"%"TS"\": has an invalid filename "
1844                                 "that is not supported on this platform",
1845                                 dentry_full_path(dentry));
1846                         goto skip_dentry;
1847                 }
1848         }
1849
1850 out_replace:
1851         {
1852                 utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2];
1853
1854                 memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes);
1855                 file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true);
1856
1857                 tchar *tchar_name;
1858                 size_t tchar_nchars;
1859         #if TCHAR_IS_UTF16LE
1860                 tchar_name = utf16_name_copy;
1861                 tchar_nchars = dentry->file_name_nbytes / 2;
1862         #else
1863                 ret = utf16le_to_tstr(utf16_name_copy,
1864                                       dentry->file_name_nbytes,
1865                                       &tchar_name, &tchar_nchars);
1866                 if (ret)
1867                         return ret;
1868         #endif
1869                 size_t fixed_name_num_chars = tchar_nchars;
1870                 tchar fixed_name[tchar_nchars + 50];
1871
1872                 tmemcpy(fixed_name, tchar_name, tchar_nchars);
1873                 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
1874                                                  T(" (invalid filename #%lu)"),
1875                                                  ++ctx->invalid_sequence);
1876         #if !TCHAR_IS_UTF16LE
1877                 FREE(tchar_name);
1878         #endif
1879                 dentry->extraction_name = memdup(fixed_name,
1880                                                  2 * fixed_name_num_chars + 2);
1881                 if (!dentry->extraction_name)
1882                         return WIMLIB_ERR_NOMEM;
1883                 dentry->extraction_name_nchars = fixed_name_num_chars;
1884         }
1885         return 0;
1886
1887 skip_dentry:
1888         for_dentry_in_tree(dentry, do_dentry_delete_from_list, NULL);
1889         return 0;
1890 }
1891
1892 /*
1893  * Calculate the actual filename component at which each WIM dentry will be
1894  * extracted, with special handling for dentries that are unsupported by the
1895  * extraction backend or have invalid names.
1896  *
1897  * Note: this has a dependency on start_extract() being called because
1898  * ctx.supported_features must be filled in in order to determine whether each
1899  * dentry is supported.
1900  *
1901  * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
1902  */
1903 static int
1904 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
1905                                        struct apply_ctx *ctx)
1906 {
1907         struct list_head *prev, *cur;
1908
1909         /* Can't use list_for_each_entry() because a call to
1910          * dentry_calculate_extraction_name() may delete the current dentry and
1911          * its children from the list.  */
1912
1913         prev = dentry_list;
1914         for (;;) {
1915                 struct wim_dentry *dentry;
1916                 int ret;
1917
1918                 cur = prev->next;
1919                 if (cur == dentry_list)
1920                         break;
1921
1922                 dentry = list_entry(cur, struct wim_dentry, extraction_list);
1923
1924                 ret = dentry_calculate_extraction_name(dentry, ctx);
1925                 if (ret)
1926                         return ret;
1927
1928                 if (prev->next == cur)
1929                         prev = cur;
1930                 else
1931                         ; /* Current dentry and its children (which follow in
1932                              the list) were deleted.  prev stays the same.  */
1933         }
1934         return 0;
1935 }
1936
1937 static int
1938 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
1939                        struct wim_lookup_table *lookup_table)
1940 {
1941         struct wim_inode *inode = dentry->d_inode;
1942         struct wim_lookup_table_entry *lte;
1943         int ret;
1944         bool force = false;
1945
1946         /* Special case:  when extracting from a pipe, the WIM lookup table is
1947          * initially empty, so "resolving" an inode's streams is initially not
1948          * possible.  However, we still need to keep track of which streams,
1949          * identified by SHA1 message digests, need to be extracted, so we
1950          * "resolve" the inode's streams anyway by allocating new entries.  */
1951         if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
1952                 force = true;
1953         ret = inode_resolve_streams(inode, lookup_table, force);
1954         if (ret)
1955                 return ret;
1956         for (u32 i = 0; i <= inode->i_num_ads; i++) {
1957                 lte = inode_stream_lte_resolved(inode, i);
1958                 if (lte)
1959                         lte->out_refcnt = 0;
1960         }
1961         return 0;
1962 }
1963
1964 /*
1965  * For each dentry to be extracted, resolve all streams in the corresponding
1966  * inode and set 'out_refcnt' in each to 0.
1967  *
1968  * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
1969  */
1970 static int
1971 dentry_list_resolve_streams(struct list_head *dentry_list,
1972                             struct apply_ctx *ctx)
1973 {
1974         struct wim_dentry *dentry;
1975         int ret;
1976
1977         list_for_each_entry(dentry, dentry_list, extraction_list) {
1978                 ret = dentry_resolve_streams(dentry,
1979                                              ctx->extract_flags,
1980                                              ctx->wim->lookup_table);
1981                 if (ret)
1982                         return ret;
1983         }
1984         return 0;
1985 }
1986
1987 static int
1988 ref_stream(struct wim_lookup_table_entry *lte,
1989            struct wim_dentry *dentry, struct apply_ctx *ctx)
1990 {
1991         if (!lte)
1992                 return 0;
1993
1994         /* Tally the size only for each extraction of the stream (not hard
1995          * links).  */
1996         if (!(dentry->d_inode->i_visited &&
1997               ctx->supported_features.hard_links) &&
1998             (!is_linked_extraction(ctx) || (lte->out_refcnt == 0 &&
1999                                             lte->extracted_file == NULL)))
2000         {
2001                 ctx->progress.extract.total_bytes += lte->size;
2002                 ctx->progress.extract.num_streams++;
2003         }
2004
2005         /* Add stream to the dentry_list only one time, even if it's going
2006          * to be extracted to multiple locations.  */
2007         if (lte->out_refcnt == 0) {
2008                 list_add_tail(&lte->extraction_list, &ctx->stream_list);
2009                 ctx->num_streams_remaining++;
2010         }
2011
2012         if (!(ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FILE_ORDER)) {
2013                 struct wim_dentry **lte_dentries;
2014
2015                 /* Append dentry to this stream's array of dentries referencing
2016                  * it.  Use inline array to avoid memory allocation until the
2017                  * number of dentries becomes too large.  */
2018                 if (lte->out_refcnt < ARRAY_LEN(lte->inline_lte_dentries)) {
2019                         lte_dentries = lte->inline_lte_dentries;
2020                 } else {
2021                         struct wim_dentry **prev_lte_dentries;
2022                         size_t alloc_lte_dentries;
2023
2024                         if (lte->out_refcnt == ARRAY_LEN(lte->inline_lte_dentries)) {
2025                                 prev_lte_dentries = NULL;
2026                                 alloc_lte_dentries = ARRAY_LEN(lte->inline_lte_dentries);
2027                         } else {
2028                                 prev_lte_dentries = lte->lte_dentries;
2029                                 alloc_lte_dentries = lte->alloc_lte_dentries;
2030                         }
2031
2032                         if (lte->out_refcnt == alloc_lte_dentries) {
2033                                 alloc_lte_dentries *= 2;
2034                                 lte_dentries = REALLOC(prev_lte_dentries,
2035                                                        alloc_lte_dentries *
2036                                                         sizeof(lte_dentries[0]));
2037                                 if (lte_dentries == NULL)
2038                                         return WIMLIB_ERR_NOMEM;
2039                                 if (prev_lte_dentries == NULL) {
2040                                         memcpy(lte_dentries,
2041                                                lte->inline_lte_dentries,
2042                                                sizeof(lte->inline_lte_dentries));
2043                                 }
2044                                 lte->lte_dentries = lte_dentries;
2045                                 lte->alloc_lte_dentries = alloc_lte_dentries;
2046                         }
2047                         lte_dentries = lte->lte_dentries;
2048                 }
2049                 lte_dentries[lte->out_refcnt] = dentry;
2050         }
2051         lte->out_refcnt++;
2052         return 0;
2053 }
2054
2055 static int
2056 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
2057 {
2058         struct wim_inode *inode = dentry->d_inode;
2059         int ret;
2060
2061         /* The unnamed data stream will always be extracted, except in an
2062          * unlikely case.  */
2063         if (!inode_is_encrypted_directory(inode)) {
2064                 ret = ref_stream(inode_unnamed_lte_resolved(inode),
2065                                  dentry, ctx);
2066                 if (ret)
2067                         return ret;
2068         }
2069
2070         /* Named data streams will be extracted only if supported in the current
2071          * extraction mode and volume, and to avoid complications, if not doing
2072          * a linked extraction.  */
2073         if (can_extract_named_data_streams(ctx)) {
2074                 for (u16 i = 0; i < inode->i_num_ads; i++) {
2075                         if (!ads_entry_is_named_stream(&inode->i_ads_entries[i]))
2076                                 continue;
2077                         ret = ref_stream(inode->i_ads_entries[i].lte,
2078                                          dentry, ctx);
2079                         if (ret)
2080                                 return ret;
2081                 }
2082         }
2083         inode->i_visited = 1;
2084         return 0;
2085 }
2086
2087 /*
2088  * For each dentry to be extracted, iterate through the data streams of the
2089  * corresponding inode.  For each such stream that is not to be ignored due to
2090  * the supported features or extraction flags, add it to the list of streams to
2091  * be extracted (ctx->stream_list) if not already done so.
2092  *
2093  * Also, if doing a sequential extraction, build a mapping from each stream to
2094  * the dentries referencing it.
2095  *
2096  * This also initializes the extract progress info with byte and stream
2097  * information.
2098  *
2099  * Note: This has a dependency on start_extract being called because
2100  * ctx.supported_features must be filled in in order to determine whether named
2101  * data streams are supported.
2102  *
2103  * Note: this uses the i_visited member of the inodes (assumed to be 0
2104  * initially), but does not reset it.
2105  *
2106  * Possible error codes: WIMLIB_ERR_NOMEM.
2107  */
2108 static int
2109 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
2110 {
2111         struct wim_dentry *dentry;
2112         int ret;
2113
2114         list_for_each_entry(dentry, dentry_list, extraction_list) {
2115                 ret = dentry_ref_streams(dentry, ctx);
2116                 if (ret)
2117                         return ret;
2118         }
2119         return 0;
2120 }
2121
2122 /* Tally features necessary to extract a dentry and the corresponding inode.  */
2123 static void
2124 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
2125 {
2126         struct wim_inode *inode = dentry->d_inode;
2127
2128         if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
2129                 features->archive_files++;
2130         if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
2131                 features->hidden_files++;
2132         if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
2133                 features->system_files++;
2134         if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
2135                 features->compressed_files++;
2136         if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
2137                 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
2138                         features->encrypted_directories++;
2139                 else
2140                         features->encrypted_files++;
2141         }
2142         if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
2143                 features->not_context_indexed_files++;
2144         if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
2145                 features->sparse_files++;
2146         if (inode_has_named_stream(inode))
2147                 features->named_data_streams++;
2148         if (inode->i_visited)
2149                 features->hard_links++;
2150         if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
2151                 features->reparse_points++;
2152                 if (inode_is_symlink(inode))
2153                         features->symlink_reparse_points++;
2154                 else
2155                         features->other_reparse_points++;
2156         }
2157         if (inode->i_security_id != -1)
2158                 features->security_descriptors++;
2159         if (dentry->short_name_nbytes)
2160                 features->short_names++;
2161         if (inode_has_unix_data(inode))
2162                 features->unix_data++;
2163         inode->i_visited = 1;
2164 }
2165
2166 /* Tally the features necessary to extract the specified dentries.  */
2167 static void
2168 dentry_list_get_features(struct list_head *dentry_list,
2169                          struct wim_features *features)
2170 {
2171         struct wim_dentry *dentry;
2172
2173         memset(features, 0, sizeof(struct wim_features));
2174
2175         list_for_each_entry(dentry, dentry_list, extraction_list)
2176                 dentry_tally_features(dentry, features);
2177
2178         list_for_each_entry(dentry, dentry_list, extraction_list)
2179                 dentry->d_inode->i_visited = 0;
2180 }
2181
2182 static u32
2183 compute_supported_attributes_mask(const struct wim_features *supported_features)
2184 {
2185         u32 mask = (u32)~0UL;
2186
2187         if (!supported_features->archive_files)
2188                 mask &= ~FILE_ATTRIBUTE_ARCHIVE;
2189
2190         if (!supported_features->hidden_files)
2191                 mask &= ~FILE_ATTRIBUTE_HIDDEN;
2192
2193         if (!supported_features->system_files)
2194                 mask &= ~FILE_ATTRIBUTE_SYSTEM;
2195
2196         if (!supported_features->not_context_indexed_files)
2197                 mask &= ~FILE_ATTRIBUTE_NOT_CONTENT_INDEXED;
2198
2199         if (!supported_features->compressed_files)
2200                 mask &= ~FILE_ATTRIBUTE_COMPRESSED;
2201
2202         if (!supported_features->sparse_files)
2203                 mask &= ~FILE_ATTRIBUTE_SPARSE_FILE;
2204
2205         if (!supported_features->reparse_points)
2206                 mask &= ~FILE_ATTRIBUTE_REPARSE_POINT;
2207
2208         return mask;
2209 }
2210
2211 static int
2212 do_feature_check(const struct wim_features *required_features,
2213                  const struct wim_features *supported_features,
2214                  int extract_flags, const struct apply_operations *ops)
2215 {
2216         /* File attributes.  */
2217         if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
2218                 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
2219                  * We're an archive program, so theoretically we can do what we
2220                  * want with it.  */
2221
2222                 if (required_features->hidden_files &&
2223                     !supported_features->hidden_files)
2224                         WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
2225                                 required_features->hidden_files);
2226
2227                 if (required_features->system_files &&
2228                     !supported_features->system_files)
2229                         WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
2230                                 required_features->system_files);
2231
2232                 if (required_features->compressed_files &&
2233                     !supported_features->compressed_files)
2234                         WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
2235                                 required_features->compressed_files);
2236
2237                 if (required_features->not_context_indexed_files &&
2238                     !supported_features->not_context_indexed_files)
2239                         WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
2240                                 required_features->not_context_indexed_files);
2241
2242                 if (required_features->sparse_files &&
2243                     !supported_features->sparse_files)
2244                         WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
2245                                 required_features->sparse_files);
2246
2247                 if (required_features->encrypted_directories &&
2248                     !supported_features->encrypted_directories)
2249                         WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
2250                                 required_features->encrypted_directories);
2251         }
2252
2253         /* Encrypted files.  */
2254         if (required_features->encrypted_files &&
2255             !supported_features->encrypted_files)
2256                 WARNING("Ignoring %lu encrypted files",
2257                         required_features->encrypted_files);
2258
2259         /* Named data streams.  */
2260         if (required_features->named_data_streams &&
2261             (!supported_features->named_data_streams ||
2262              (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
2263                                WIMLIB_EXTRACT_FLAG_HARDLINK))))
2264                 WARNING("Ignoring named data streams of %lu files",
2265                         required_features->named_data_streams);
2266
2267         /* Hard links.  */
2268         if ((extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) &&
2269             !supported_features->hard_links)
2270         {
2271                 ERROR("Extraction backend does not support hard links!");
2272                 return WIMLIB_ERR_UNSUPPORTED;
2273         }
2274         if (required_features->hard_links && !supported_features->hard_links)
2275                 WARNING("Extracting %lu hard links as independent files",
2276                         required_features->hard_links);
2277
2278         /* Symbolic links and reparse points.  */
2279         if ((extract_flags & WIMLIB_EXTRACT_FLAG_SYMLINK) &&
2280             !supported_features->symlink_reparse_points)
2281         {
2282                 ERROR("Extraction backend does not support symbolic links!");
2283                 return WIMLIB_ERR_UNSUPPORTED;
2284         }
2285         if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
2286             required_features->symlink_reparse_points &&
2287             !supported_features->symlink_reparse_points &&
2288             !supported_features->reparse_points)
2289         {
2290                 ERROR("Extraction backend does not support symbolic links!");
2291                 return WIMLIB_ERR_UNSUPPORTED;
2292         }
2293         if (required_features->reparse_points &&
2294             !supported_features->reparse_points)
2295         {
2296                 if (supported_features->symlink_reparse_points) {
2297                         if (required_features->other_reparse_points) {
2298                                 WARNING("Ignoring %lu non-symlink/junction "
2299                                         "reparse point files",
2300                                         required_features->other_reparse_points);
2301                         }
2302                 } else {
2303                         WARNING("Ignoring %lu reparse point files",
2304                                 required_features->reparse_points);
2305                 }
2306         }
2307
2308         /* Security descriptors.  */
2309         if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
2310                                WIMLIB_EXTRACT_FLAG_UNIX_DATA))
2311              == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
2312             required_features->security_descriptors &&
2313             !supported_features->security_descriptors)
2314         {
2315                 ERROR("Extraction backend does not support security descriptors!");
2316                 return WIMLIB_ERR_UNSUPPORTED;
2317         }
2318         if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
2319             required_features->security_descriptors &&
2320             !supported_features->security_descriptors)
2321                 WARNING("Ignoring Windows NT security descriptors of %lu files",
2322                         required_features->security_descriptors);
2323
2324         /* UNIX data.  */
2325         if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) &&
2326             required_features->unix_data && !supported_features->unix_data)
2327         {
2328                 ERROR("Extraction backend does not support UNIX data!");
2329                 return WIMLIB_ERR_UNSUPPORTED;
2330         }
2331
2332         /* DOS Names.  */
2333         if (required_features->short_names &&
2334             !supported_features->short_names)
2335         {
2336                 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
2337                         ERROR("Extraction backend does not support DOS names!");
2338                         return WIMLIB_ERR_UNSUPPORTED;
2339                 }
2340                 WARNING("Ignoring DOS names of %lu files",
2341                         required_features->short_names);
2342         }
2343
2344         /* Timestamps.  */
2345         if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
2346             !ops->set_timestamps)
2347         {
2348                 ERROR("Extraction backend does not support timestamps!");
2349                 return WIMLIB_ERR_UNSUPPORTED;
2350         }
2351
2352         return 0;
2353 }
2354
2355 static void
2356 do_extract_warnings(struct apply_ctx *ctx)
2357 {
2358         if (ctx->partial_security_descriptors == 0 &&
2359             ctx->no_security_descriptors == 0)
2360                 return;
2361
2362         WARNING("Extraction to \"%"TS"\" complete, but with one or more warnings:",
2363                 ctx->target);
2364         if (ctx->partial_security_descriptors != 0) {
2365                 WARNING("- Could only partially set the security descriptor\n"
2366                         "            on %lu files or directories.",
2367                         ctx->partial_security_descriptors);
2368         }
2369         if (ctx->no_security_descriptors != 0) {
2370                 WARNING("- Could not set security descriptor at all\n"
2371                         "            on %lu files or directories.",
2372                         ctx->no_security_descriptors);
2373         }
2374 #ifdef __WIN32__
2375         WARNING("To fully restore all security descriptors, run the program\n"
2376                 "          with Administrator rights.");
2377 #endif
2378 }
2379
2380 static void
2381 destroy_dentry_list(struct list_head *dentry_list)
2382 {
2383         struct wim_dentry *dentry, *tmp;
2384         struct wim_inode *inode;
2385
2386         list_for_each_entry_safe(dentry, tmp, dentry_list, extraction_list) {
2387                 inode = dentry->d_inode;
2388                 dentry_reset_extraction_list_node(dentry);
2389                 dentry->was_linked = 0;
2390                 dentry->skeleton_extracted = 0;
2391                 inode->i_visited = 0;
2392                 FREE(inode->i_extracted_file);
2393                 inode->i_extracted_file = NULL;
2394                 inode->i_dos_name_extracted = 0;
2395                 if ((void*)dentry->extraction_name != (void*)dentry->file_name)
2396                         FREE(dentry->extraction_name);
2397                 dentry->extraction_name = NULL;
2398         }
2399 }
2400
2401 static void
2402 destroy_stream_list(struct list_head *stream_list)
2403 {
2404         struct wim_lookup_table_entry *lte;
2405
2406         list_for_each_entry(lte, stream_list, extraction_list)
2407                 if (lte->out_refcnt > ARRAY_LEN(lte->inline_lte_dentries))
2408                         FREE(lte->lte_dentries);
2409 }
2410
2411 static int
2412 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
2413               const tchar *target, int extract_flags,
2414               wimlib_progress_func_t progress_func)
2415 {
2416         int ret;
2417         struct apply_ctx ctx;
2418         struct list_head dentry_list;
2419         struct wim_features required_features;
2420
2421         /* Handle stdout extraction as a separate case.  */
2422         if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT)
2423                 return extract_dentries_to_stdout(trees, num_trees,
2424                                                   wim->lookup_table);
2425
2426         /* Start initializing the apply_ctx.  */
2427         memset(&ctx, 0, sizeof(struct apply_ctx));
2428         ctx.wim = wim;
2429         ctx.extract_flags = extract_flags;
2430         ctx.target = target;
2431         ctx.target_nchars = tstrlen(target);
2432         ctx.progress_func = progress_func;
2433         if (progress_func) {
2434                 ctx.progress.extract.wimfile_name = wim->filename;
2435                 ctx.progress.extract.image = wim->current_image;
2436                 ctx.progress.extract.extract_flags = (extract_flags &
2437                                                       WIMLIB_EXTRACT_MASK_PUBLIC);
2438                 ctx.progress.extract.image_name = wimlib_get_image_name(wim,
2439                                                                         wim->current_image);
2440                 ctx.progress.extract.target = target;
2441         }
2442
2443         if (extract_flags & WIMLIB_EXTRACT_FLAG_FILEMODE) {
2444                 /* Called from wimlib_extract_files().  There should be only 1
2445                  * tree, and directory structure should not be preserved.  */
2446                 wimlib_assert(num_trees == 1);
2447                 wimlib_assert(extract_flags &
2448                               WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE);
2449                 ret = calculate_dentry_full_path(trees[0]);
2450                 if (ret)
2451                         return ret;
2452                 ctx.progress.extract.extract_root_wim_source_path = trees[0]->_full_path;
2453                 ctx.target_dentry = trees[0];
2454         } else {
2455                 ctx.progress.extract.extract_root_wim_source_path = T("");
2456                 ctx.target_dentry = wim_root_dentry(wim);
2457         }
2458         /* Note: ctx.target_dentry represents the dentry that gets extracted to
2459          * @target.  There may be none, in which case it gets set to the image
2460          * root and never matches any of the dentries actually being extracted.
2461          */
2462
2463         num_trees = remove_duplicate_trees(trees, num_trees);
2464
2465         /* All trees are now distinct.  */
2466
2467         num_trees = remove_contained_trees(trees, num_trees);
2468
2469         /* All trees are now distinct and non-overlapping.  */
2470
2471         /* Build list of dentries to be extracted.  */
2472         build_dentry_list(&dentry_list, trees, num_trees,
2473                           !(extract_flags & WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
2474
2475         /* Select the appropriate apply_operations based on the platform and
2476          * extract_flags.  */
2477         ctx.ops = select_apply_operations(extract_flags);
2478
2479         /* Figure out whether the root dentry is being extracted to the root of
2480          * a volume and therefore needs to be treated "specially", for example
2481          * not being explicitly created and not having attributes set.  */
2482         if (ctx.ops->target_is_root && ctx.ops->root_directory_is_special)
2483                 ctx.root_dentry_is_special = ctx.ops->target_is_root(target);
2484
2485         /* Call the start_extract() callback.  This gives the apply_operations
2486          * implementation a chance to do any setup needed to access the volume.
2487          * Furthermore, start_extract() is expected to set the supported
2488          * features of this extraction mode (ctx.supported_features), which are
2489          * determined at runtime as they may vary depending on the actual
2490          * volume.  */
2491         ret = ctx.ops->start_extract(target, &ctx);
2492         if (ret)
2493                 return ret;
2494
2495         /* Get and check the features required to extract the dentries.  */
2496         dentry_list_get_features(&dentry_list, &required_features);
2497         ret = do_feature_check(&required_features, &ctx.supported_features,
2498                                extract_flags, ctx.ops);
2499         if (ret)
2500                 goto out_finish_or_abort_extract;
2501
2502         ctx.supported_attributes_mask =
2503                 compute_supported_attributes_mask(&ctx.supported_features);
2504
2505         /* Calculate extraction name for each dentry and remove subtrees that
2506          * can't be extracted due to naming problems.  */
2507         ret = dentry_list_calculate_extraction_names(&dentry_list, &ctx);
2508         if (ret)
2509                 goto out_destroy_dentry_list;
2510
2511         /* Build list of streams to extract.  */
2512         ret = dentry_list_resolve_streams(&dentry_list, &ctx);
2513         if (ret)
2514                 goto out_destroy_dentry_list;
2515         INIT_LIST_HEAD(&ctx.stream_list);
2516         ret = dentry_list_ref_streams(&dentry_list, &ctx);
2517         if (ret)
2518                 goto out_destroy_stream_list;
2519
2520         if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2521                 /* When extracting from a pipe, the number of bytes of data to
2522                  * extract can't be determined in the normal way (examining the
2523                  * lookup table), since at this point all we have is a set of
2524                  * SHA1 message digests of streams that need to be extracted.
2525                  * However, we can get a reasonably accurate estimate by taking
2526                  * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
2527                  * data.  This does assume that a full image is being extracted,
2528                  * but currently there is no API for doing otherwise.  (Also,
2529                  * subtract <HARDLINKBYTES> from this if hard links are
2530                  * supported by the extraction mode.)  */
2531                 ctx.progress.extract.total_bytes =
2532                         wim_info_get_image_total_bytes(wim->wim_info,
2533                                                        wim->current_image);
2534                 if (ctx.supported_features.hard_links) {
2535                         ctx.progress.extract.total_bytes -=
2536                                 wim_info_get_image_hard_link_bytes(wim->wim_info,
2537                                                                    wim->current_image);
2538                 }
2539         }
2540
2541         if (ctx.ops->realpath_works_on_nonexisting_files &&
2542             ((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) ||
2543              ctx.ops->requires_realtarget_in_paths))
2544         {
2545                 ctx.realtarget = realpath(target, NULL);
2546                 if (!ctx.realtarget) {
2547                         ret = WIMLIB_ERR_NOMEM;
2548                         goto out_destroy_stream_list;
2549                 }
2550                 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2551         }
2552
2553         if (progress_func) {
2554                 int msg;
2555                 if (extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE)
2556                         msg = WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN;
2557                 else
2558                         msg = WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN;
2559                 progress_func(msg, &ctx.progress);
2560         }
2561
2562         if (!ctx.root_dentry_is_special) {
2563                 tchar path[ctx.ops->path_max];
2564                 if (build_extraction_path(path, ctx.target_dentry, &ctx)) {
2565                         ret = extract_inode(path, &ctx, ctx.target_dentry->d_inode);
2566                         if (ret)
2567                                 goto out_free_realtarget;
2568                 }
2569         }
2570
2571         /* If we need to fix up the targets of absolute symbolic links
2572          * (WIMLIB_EXTRACT_FLAG_RPFIX) or the extraction mode requires paths to
2573          * be absolute, use realpath() (or its replacement on Windows) to get
2574          * the absolute path to the extraction target.  Note that this requires
2575          * the target directory to exist, unless
2576          * realpath_works_on_nonexisting_files is set in the apply_operations.
2577          * */
2578         if (!ctx.realtarget &&
2579             (((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) &&
2580               required_features.symlink_reparse_points) ||
2581              ctx.ops->requires_realtarget_in_paths))
2582         {
2583                 ctx.realtarget = realpath(target, NULL);
2584                 if (!ctx.realtarget) {
2585                         ret = WIMLIB_ERR_NOMEM;
2586                         goto out_free_realtarget;
2587                 }
2588                 ctx.realtarget_nchars = tstrlen(ctx.realtarget);
2589         }
2590
2591         if (ctx.ops->requires_short_name_reordering) {
2592                 if (progress_func)
2593                         progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2594                                       &ctx.progress);
2595                 ret = extract_dir_structure(&dentry_list, &ctx);
2596                 if (ret)
2597                         goto out_free_realtarget;
2598         }
2599
2600         /* Finally, the important part: extract the tree of files.  */
2601         if (!(extract_flags & WIMLIB_EXTRACT_FLAG_FILE_ORDER)) {
2602                 /* Sequential extraction requested, so two passes are needed
2603                  * (one for file structure, one for streams.)  */
2604                 if (progress_func && !ctx.ops->requires_short_name_reordering)
2605                         progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2606                                       &ctx.progress);
2607
2608                 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_RESUME)) {
2609                         ret = extract_structure(&dentry_list, &ctx);
2610                         if (ret)
2611                                 goto out_free_realtarget;
2612                 }
2613                 if (progress_func)
2614                         progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2615                                       &ctx.progress);
2616                 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
2617                         ret = extract_streams_from_pipe(&ctx);
2618                 else
2619                         ret = extract_stream_list(&ctx);
2620                 if (ret)
2621                         goto out_free_realtarget;
2622         } else {
2623                 /* Sequential extraction was not requested, so we can make do
2624                  * with one pass where we both create the files and extract
2625                  * streams.   */
2626                 if (progress_func && !ctx.ops->requires_short_name_reordering)
2627                         progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN,
2628                                       &ctx.progress);
2629                 ret = extract_dentries(&dentry_list, &ctx);
2630                 if (ret)
2631                         goto out_free_realtarget;
2632                 if (progress_func)
2633                         progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END,
2634                                       &ctx.progress);
2635         }
2636
2637         /* If the total number of bytes to extract was miscalculated, just jump
2638          * to the calculated number in order to avoid confusing the progress
2639          * function.  This should only occur when extracting from a pipe.  */
2640         if (ctx.progress.extract.completed_bytes != ctx.progress.extract.total_bytes)
2641         {
2642                 DEBUG("Calculated %"PRIu64" bytes to extract, but actually "
2643                       "extracted %"PRIu64,
2644                       ctx.progress.extract.total_bytes,
2645                       ctx.progress.extract.completed_bytes);
2646         }
2647         if (progress_func &&
2648             ctx.progress.extract.completed_bytes < ctx.progress.extract.total_bytes)
2649         {
2650                 ctx.progress.extract.completed_bytes = ctx.progress.extract.total_bytes;
2651                 progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, &ctx.progress);
2652         }
2653
2654         /* Apply security descriptors and timestamps.  This is done at the end,
2655          * and in a depth-first manner, to prevent timestamps from getting
2656          * changed by subsequent extract operations and to minimize the chance
2657          * of the restored security descriptors getting in our way.  */
2658         if (progress_func)
2659                 progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS,
2660                               &ctx.progress);
2661         ret = extract_final_metadata(&dentry_list, &ctx);
2662         if (ret)
2663                 goto out_free_realtarget;
2664
2665         if (progress_func) {
2666                 int msg;
2667                 if (extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE)
2668                         msg = WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END;
2669                 else
2670                         msg = WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END;
2671                 progress_func(msg, &ctx.progress);
2672         }
2673         do_extract_warnings(&ctx);
2674         ret = 0;
2675 out_free_realtarget:
2676         FREE(ctx.realtarget);
2677 out_destroy_stream_list:
2678         if (!(ctx.extract_flags & WIMLIB_EXTRACT_FLAG_FILE_ORDER))
2679                 destroy_stream_list(&ctx.stream_list);
2680 out_destroy_dentry_list:
2681         destroy_dentry_list(&dentry_list);
2682 out_finish_or_abort_extract:
2683         if (ret) {
2684                 if (ctx.ops->abort_extract)
2685                         ctx.ops->abort_extract(&ctx);
2686         } else {
2687                 if (ctx.ops->finish_extract)
2688                         ret = ctx.ops->finish_extract(&ctx);
2689         }
2690         return ret;
2691 }
2692
2693 static int
2694 mkdir_if_needed(const tchar *target)
2695 {
2696         struct stat stbuf;
2697         if (tstat(target, &stbuf)) {
2698                 if (errno == ENOENT) {
2699                         if (tmkdir(target, 0755)) {
2700                                 ERROR_WITH_ERRNO("Failed to create directory "
2701                                                  "\"%"TS"\"", target);
2702                                 return WIMLIB_ERR_MKDIR;
2703                         }
2704                 } else {
2705                         ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target);
2706                         return WIMLIB_ERR_STAT;
2707                 }
2708         } else if (!S_ISDIR(stbuf.st_mode)) {
2709                 ERROR("\"%"TS"\" is not a directory", target);
2710                 return WIMLIB_ERR_NOTDIR;
2711         }
2712         return 0;
2713 }
2714
2715 /* Make sure the extraction flags make sense, and update them if needed.  */
2716 static int
2717 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
2718 {
2719         int extract_flags = *extract_flags_p;
2720
2721         /* Check for invalid flag combinations  */
2722         if ((extract_flags &
2723              (WIMLIB_EXTRACT_FLAG_SYMLINK |
2724               WIMLIB_EXTRACT_FLAG_HARDLINK)) == (WIMLIB_EXTRACT_FLAG_SYMLINK |
2725                                                  WIMLIB_EXTRACT_FLAG_HARDLINK))
2726                 return WIMLIB_ERR_INVALID_PARAM;
2727
2728         if ((extract_flags &
2729              (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2730               WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
2731                                                     WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
2732                 return WIMLIB_ERR_INVALID_PARAM;
2733
2734         if ((extract_flags &
2735              (WIMLIB_EXTRACT_FLAG_RPFIX |
2736               WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
2737                                                 WIMLIB_EXTRACT_FLAG_NORPFIX))
2738                 return WIMLIB_ERR_INVALID_PARAM;
2739
2740         if ((extract_flags &
2741              (WIMLIB_EXTRACT_FLAG_RESUME |
2742               WIMLIB_EXTRACT_FLAG_FROM_PIPE)) == WIMLIB_EXTRACT_FLAG_RESUME)
2743                 return WIMLIB_ERR_INVALID_PARAM;
2744
2745 #ifndef WITH_NTFS_3G
2746         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2747                 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
2748                       "        it cannot apply a WIM image directly to a NTFS volume.");
2749                 return WIMLIB_ERR_UNSUPPORTED;
2750         }
2751 #endif
2752
2753         if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
2754                               WIMLIB_EXTRACT_FLAG_NORPFIX |
2755                               WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
2756                                         WIMLIB_EXTRACT_FLAG_IMAGEMODE)
2757         {
2758                 /* For full-image extraction, do reparse point fixups by default
2759                  * if the WIM header says they are enabled.  */
2760                 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
2761                         extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
2762         }
2763
2764         /* TODO: Since UNIX data entries are stored in the file resources, in a
2765          * completely sequential extraction they may come up before the
2766          * corresponding file or symbolic link data.  This needs to be handled
2767          * better.  */
2768         if ((extract_flags & (WIMLIB_EXTRACT_FLAG_UNIX_DATA |
2769                               WIMLIB_EXTRACT_FLAG_FILE_ORDER))
2770                                     == WIMLIB_EXTRACT_FLAG_UNIX_DATA)
2771         {
2772                 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
2773                         WARNING("Setting UNIX file/owner group may "
2774                                 "be impossible on some\n"
2775                                 "          symbolic links "
2776                                 "when applying from a pipe.");
2777                 } else {
2778                         extract_flags |= WIMLIB_EXTRACT_FLAG_FILE_ORDER;
2779                         WARNING("Disabling sequential extraction for "
2780                                 "UNIX data mode");
2781                 }
2782         }
2783
2784         *extract_flags_p = extract_flags;
2785         return 0;
2786 }
2787
2788 static u32
2789 get_wildcard_flags(int extract_flags)
2790 {
2791         u32 wildcard_flags = 0;
2792
2793         if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB)
2794                 wildcard_flags |= WILDCARD_FLAG_ERROR_IF_NO_MATCH;
2795         else
2796                 wildcard_flags |= WILDCARD_FLAG_WARN_IF_NO_MATCH;
2797
2798         if (default_ignore_case)
2799                 wildcard_flags |= WILDCARD_FLAG_CASE_INSENSITIVE;
2800
2801         return wildcard_flags;
2802 }
2803
2804 struct append_dentry_ctx {
2805         struct wim_dentry **dentries;
2806         size_t num_dentries;
2807         size_t num_alloc_dentries;
2808 };
2809
2810 static int
2811 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
2812 {
2813         struct append_dentry_ctx *ctx = _ctx;
2814
2815         if (ctx->num_dentries == ctx->num_alloc_dentries) {
2816                 struct wim_dentry **new_dentries;
2817                 size_t new_length;
2818
2819                 new_length = max(ctx->num_alloc_dentries + 8,
2820                                  ctx->num_alloc_dentries * 3 / 2);
2821                 new_dentries = REALLOC(ctx->dentries,
2822                                        new_length * sizeof(ctx->dentries[0]));
2823                 if (new_dentries == NULL)
2824                         return WIMLIB_ERR_NOMEM;
2825                 ctx->dentries = new_dentries;
2826                 ctx->num_alloc_dentries = new_length;
2827         }
2828         ctx->dentries[ctx->num_dentries++] = dentry;
2829         return 0;
2830 }
2831
2832 static int
2833 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
2834                         const tchar * const *paths, size_t num_paths,
2835                         int extract_flags, wimlib_progress_func_t progress_func)
2836 {
2837         int ret;
2838         struct wim_dentry **trees;
2839         size_t num_trees;
2840
2841         if (wim == NULL || target == NULL || target[0] == T('\0') ||
2842             (num_paths != 0 && paths == NULL))
2843                 return WIMLIB_ERR_INVALID_PARAM;
2844
2845         ret = check_extract_flags(wim, &extract_flags);
2846         if (ret)
2847                 return ret;
2848
2849         ret = select_wim_image(wim, image);
2850         if (ret)
2851                 return ret;
2852
2853         ret = wim_checksum_unhashed_streams(wim);
2854         if (ret)
2855                 return ret;
2856
2857         if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
2858                               WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
2859                               WIMLIB_EXTRACT_FLAG_FILEMODE)) ==
2860             (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
2861         {
2862                 ret = mkdir_if_needed(target);
2863                 if (ret)
2864                         return ret;
2865         }
2866
2867         if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
2868
2869                 struct append_dentry_ctx append_dentry_ctx = {
2870                         .dentries = NULL,
2871                         .num_dentries = 0,
2872                         .num_alloc_dentries = 0,
2873                 };
2874
2875                 u32 wildcard_flags = get_wildcard_flags(extract_flags);
2876
2877                 for (size_t i = 0; i < num_paths; i++) {
2878                         tchar *path = canonicalize_wim_path(paths[i]);
2879                         if (path == NULL) {
2880                                 ret = WIMLIB_ERR_NOMEM;
2881                                 trees = append_dentry_ctx.dentries;
2882                                 goto out_free_trees;
2883                         }
2884                         ret = expand_wildcard(wim, path,
2885                                               append_dentry_cb,
2886                                               &append_dentry_ctx,
2887                                               wildcard_flags);
2888                         FREE(path);
2889                         if (ret) {
2890                                 trees = append_dentry_ctx.dentries;
2891                                 goto out_free_trees;
2892                         }
2893                 }
2894                 trees = append_dentry_ctx.dentries;
2895                 num_trees = append_dentry_ctx.num_dentries;
2896         } else {
2897                 trees = MALLOC(num_paths * sizeof(trees[0]));
2898                 if (trees == NULL)
2899                         return WIMLIB_ERR_NOMEM;
2900
2901                 for (size_t i = 0; i < num_paths; i++) {
2902
2903                         tchar *path = canonicalize_wim_path(paths[i]);
2904                         if (path == NULL) {
2905                                 ret = WIMLIB_ERR_NOMEM;
2906                                 goto out_free_trees;
2907                         }
2908
2909                         trees[i] = get_dentry(wim, path,
2910                                               WIMLIB_CASE_PLATFORM_DEFAULT);
2911                         FREE(path);
2912                         if (trees[i] == NULL) {
2913                                   ERROR("Path \"%"TS"\" does not exist "
2914                                         "in WIM image %d",
2915                                         paths[i], wim->current_image);
2916                                   ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
2917                                   goto out_free_trees;
2918                         }
2919                 }
2920                 num_trees = num_paths;
2921         }
2922
2923         if (num_trees == 0) {
2924                 ret = 0;
2925                 goto out_free_trees;
2926         }
2927
2928         ret = extract_trees(wim, trees, num_trees,
2929                             target, extract_flags, progress_func);
2930 out_free_trees:
2931         FREE(trees);
2932         return ret;
2933 }
2934
2935 static int
2936 extract_single_image(WIMStruct *wim, int image,
2937                      const tchar *target, int extract_flags,
2938                      wimlib_progress_func_t progress_func)
2939 {
2940         const tchar *path = T("");
2941         extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
2942         return do_wimlib_extract_paths(wim, image, target, &path, 1,
2943                                        extract_flags, progress_func);
2944 }
2945
2946 static const tchar * const filename_forbidden_chars =
2947 T(
2948 #ifdef __WIN32__
2949 "<>:\"/\\|?*"
2950 #else
2951 "/"
2952 #endif
2953 );
2954
2955 /* This function checks if it is okay to use a WIM image's name as a directory
2956  * name.  */
2957 static bool
2958 image_name_ok_as_dir(const tchar *image_name)
2959 {
2960         return image_name && *image_name &&
2961                 !tstrpbrk(image_name, filename_forbidden_chars) &&
2962                 tstrcmp(image_name, T(".")) &&
2963                 tstrcmp(image_name, T(".."));
2964 }
2965
2966 /* Extracts all images from the WIM to the directory @target, with the images
2967  * placed in subdirectories named by their image names. */
2968 static int
2969 extract_all_images(WIMStruct *wim,
2970                    const tchar *target,
2971                    int extract_flags,
2972                    wimlib_progress_func_t progress_func)
2973 {
2974         size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20);
2975         size_t output_path_len = tstrlen(target);
2976         tchar buf[output_path_len + 1 + image_name_max_len + 1];
2977         int ret;
2978         int image;
2979         const tchar *image_name;
2980
2981         extract_flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE;
2982
2983         if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
2984                 ERROR("Cannot extract multiple images in NTFS extraction mode.");
2985                 return WIMLIB_ERR_INVALID_PARAM;
2986         }
2987
2988         ret = mkdir_if_needed(target);
2989         if (ret)
2990                 return ret;
2991         tmemcpy(buf, target, output_path_len);
2992         buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
2993         for (image = 1; image <= wim->hdr.image_count; image++) {
2994                 image_name = wimlib_get_image_name(wim, image);
2995                 if (image_name_ok_as_dir(image_name)) {
2996                         tstrcpy(buf + output_path_len + 1, image_name);
2997                 } else {
2998                         /* Image name is empty or contains forbidden characters.
2999                          * Use image number instead. */
3000                         tsprintf(buf + output_path_len + 1, T("%d"), image);
3001                 }
3002                 ret = extract_single_image(wim, image, buf, extract_flags,
3003                                            progress_func);
3004                 if (ret)
3005                         return ret;
3006         }
3007         return 0;
3008 }
3009
3010 static void
3011 clear_lte_extracted_file(WIMStruct *wim, int extract_flags)
3012 {
3013         if (unlikely(extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
3014                                       WIMLIB_EXTRACT_FLAG_HARDLINK)))
3015                 for_lookup_table_entry(wim->lookup_table,
3016                                        lte_free_extracted_file, NULL);
3017 }
3018
3019 static int
3020 do_wimlib_extract_image(WIMStruct *wim,
3021                         int image,
3022                         const tchar *target,
3023                         int extract_flags,
3024                         wimlib_progress_func_t progress_func)
3025 {
3026         int ret;
3027
3028         if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
3029                              WIMLIB_EXTRACT_FLAG_TO_STDOUT |
3030                              WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
3031                 return WIMLIB_ERR_INVALID_PARAM;
3032
3033         if (image == WIMLIB_ALL_IMAGES)
3034                 ret = extract_all_images(wim, target, extract_flags,
3035                                          progress_func);
3036         else
3037                 ret = extract_single_image(wim, image, target, extract_flags,
3038                                            progress_func);
3039
3040         clear_lte_extracted_file(wim, extract_flags);
3041         return ret;
3042 }
3043
3044
3045 /****************************************************************************
3046  *                          Extraction API                                  *
3047  ****************************************************************************/
3048
3049 /* Note: new code should use wimlib_extract_paths() instead of
3050  * wimlib_extract_files() if possible.  */
3051 WIMLIBAPI int
3052 wimlib_extract_files(WIMStruct *wim, int image,
3053                      const struct wimlib_extract_command *cmds, size_t num_cmds,
3054                      int default_extract_flags,
3055                      wimlib_progress_func_t progress_func)
3056 {
3057         int all_flags = 0;
3058         int link_flags;
3059         int ret;
3060
3061         if (num_cmds == 0)
3062                 return 0;
3063
3064         default_extract_flags |= WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE;
3065
3066         for (size_t i = 0; i < num_cmds; i++) {
3067                 int cmd_flags = (cmds[i].extract_flags |
3068                                  default_extract_flags);
3069
3070                 if (cmd_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
3071                         return WIMLIB_ERR_INVALID_PARAM;
3072
3073                 int cmd_link_flags = (cmd_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK |
3074                                                    WIMLIB_EXTRACT_FLAG_HARDLINK));
3075                 if (i == 0) {
3076                         link_flags = cmd_link_flags;
3077                 } else {
3078                         if (cmd_link_flags != link_flags) {
3079                                 ERROR("The same symlink or hardlink extraction mode "
3080                                       "must be set on all extraction commands!");
3081                                 return WIMLIB_ERR_INVALID_PARAM;
3082                         }
3083                 }
3084                 all_flags |= cmd_flags;
3085         }
3086         if (all_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
3087                 ERROR("Glob paths not supported for wimlib_extract_files(). "
3088                       "Use wimlib_extract_paths() instead.");
3089                 return WIMLIB_ERR_INVALID_PARAM;
3090         }
3091
3092         for (size_t i = 0; i < num_cmds; i++) {
3093                 int extract_flags = (cmds[i].extract_flags |
3094                                      default_extract_flags);
3095                 const tchar *target = cmds[i].fs_dest_path;
3096                 const tchar *wim_source_path = cmds[i].wim_source_path;
3097
3098                 ret = do_wimlib_extract_paths(wim, image, target,
3099                                               &wim_source_path, 1,
3100                                               extract_flags | WIMLIB_EXTRACT_FLAG_FILEMODE,
3101                                               progress_func);
3102                 if (ret)
3103                         break;
3104         }
3105
3106         clear_lte_extracted_file(wim, all_flags);
3107         return ret;
3108 }
3109
3110 WIMLIBAPI int
3111 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
3112                      const tchar * const *paths, size_t num_paths,
3113                      int extract_flags, wimlib_progress_func_t progress_func)
3114 {
3115         int ret;
3116
3117         if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
3118                 return WIMLIB_ERR_INVALID_PARAM;
3119
3120         ret = do_wimlib_extract_paths(wim, image, target, paths, num_paths,
3121                                       extract_flags, progress_func);
3122         clear_lte_extracted_file(wim, extract_flags);
3123         return ret;
3124 }
3125
3126 WIMLIBAPI int
3127 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
3128                         const tchar *path_list_file, int extract_flags,
3129                         wimlib_progress_func_t progress_func)
3130 {
3131         int ret;
3132         tchar **paths;
3133         size_t num_paths;
3134         void *mem;
3135
3136         ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
3137         if (ret) {
3138                 ERROR("Failed to read path list file \"%"TS"\"",
3139                       path_list_file);
3140                 return ret;
3141         }
3142
3143         ret = wimlib_extract_paths(wim, image, target,
3144                                    (const tchar * const *)paths, num_paths,
3145                                    extract_flags, progress_func);
3146         FREE(paths);
3147         FREE(mem);
3148         return ret;
3149 }
3150
3151 WIMLIBAPI int
3152 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
3153                                const tchar *target, int extract_flags,
3154                                wimlib_progress_func_t progress_func)
3155 {
3156         int ret;
3157         WIMStruct *pwm;
3158         struct filedes *in_fd;
3159         int image;
3160         unsigned i;
3161
3162         if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
3163                 return WIMLIB_ERR_INVALID_PARAM;
3164
3165         if (extract_flags & WIMLIB_EXTRACT_FLAG_FILE_ORDER)
3166                 return WIMLIB_ERR_INVALID_PARAM;
3167
3168         /* Read the WIM header from the pipe and get a WIMStruct to represent
3169          * the pipable WIM.  Caveats:  Unlike getting a WIMStruct with
3170          * wimlib_open_wim(), getting a WIMStruct in this way will result in
3171          * an empty lookup table, no XML data read, and no filename set.  */
3172         ret = open_wim_as_WIMStruct(&pipe_fd,
3173                                     WIMLIB_OPEN_FLAG_FROM_PIPE,
3174                                     &pwm, progress_func);
3175         if (ret)
3176                 return ret;
3177
3178         /* Sanity check to make sure this is a pipable WIM.  */
3179         if (pwm->hdr.magic != PWM_MAGIC) {
3180                 ERROR("The WIM being read from file descriptor %d "
3181                       "is not pipable!", pipe_fd);
3182                 ret = WIMLIB_ERR_NOT_PIPABLE;
3183                 goto out_wimlib_free;
3184         }
3185
3186         /* Sanity check to make sure the first part of a pipable split WIM is
3187          * sent over the pipe first.  */
3188         if (pwm->hdr.part_number != 1) {
3189                 ERROR("The first part of the split WIM must be "
3190                       "sent over the pipe first.");
3191                 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
3192                 goto out_wimlib_free;
3193         }
3194
3195         in_fd = &pwm->in_fd;
3196         wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
3197
3198         /* As mentioned, the WIMStruct we created from the pipe does not have
3199          * XML data yet.  Fix this by reading the extra copy of the XML data
3200          * that directly follows the header in pipable WIMs.  (Note: see
3201          * write_pipable_wim() for more details about the format of pipable
3202          * WIMs.)  */
3203         {
3204                 struct wim_lookup_table_entry xml_lte;
3205                 struct wim_resource_spec xml_rspec;
3206                 ret = read_pwm_stream_header(pwm, &xml_lte, &xml_rspec, 0, NULL);
3207                 if (ret)
3208                         goto out_wimlib_free;
3209
3210                 if (!(xml_lte.flags & WIM_RESHDR_FLAG_METADATA))
3211                 {
3212                         ERROR("Expected XML data, but found non-metadata "
3213                               "stream.");
3214                         ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
3215                         goto out_wimlib_free;
3216                 }
3217
3218                 wim_res_spec_to_hdr(&xml_rspec, &pwm->hdr.xml_data_reshdr);
3219
3220                 ret = read_wim_xml_data(pwm);
3221                 if (ret)
3222                         goto out_wimlib_free;
3223
3224                 if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) {
3225                         ERROR("Image count in XML data is not the same as in WIM header.");
3226                         ret = WIMLIB_ERR_IMAGE_COUNT;
3227                         goto out_wimlib_free;
3228                 }
3229         }
3230
3231         /* Get image index (this may use the XML data that was just read to
3232          * resolve an image name).  */
3233         if (image_num_or_name) {
3234                 image = wimlib_resolve_image(pwm, image_num_or_name);
3235                 if (image == WIMLIB_NO_IMAGE) {
3236                         ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
3237                               image_num_or_name);
3238                         ret = WIMLIB_ERR_INVALID_IMAGE;
3239                         goto out_wimlib_free;
3240                 } else if (image == WIMLIB_ALL_IMAGES) {
3241                         ERROR("Applying all images from a pipe is not supported!");
3242                         ret = WIMLIB_ERR_INVALID_IMAGE;
3243                         goto out_wimlib_free;
3244                 }
3245         } else {
3246                 if (pwm->hdr.image_count != 1) {
3247                         ERROR("No image was specified, but the pipable WIM "
3248                               "did not contain exactly 1 image");
3249                         ret = WIMLIB_ERR_INVALID_IMAGE;
3250                         goto out_wimlib_free;
3251                 }
3252                 image = 1;
3253         }
3254
3255         /* Load the needed metadata resource.  */
3256         for (i = 1; i <= pwm->hdr.image_count; i++) {
3257                 struct wim_lookup_table_entry *metadata_lte;
3258                 struct wim_image_metadata *imd;
3259                 struct wim_resource_spec *metadata_rspec;
3260
3261                 metadata_lte = new_lookup_table_entry();
3262                 if (metadata_lte == NULL) {
3263                         ret = WIMLIB_ERR_NOMEM;
3264                         goto out_wimlib_free;
3265                 }
3266                 metadata_rspec = MALLOC(sizeof(struct wim_resource_spec));
3267                 if (metadata_rspec == NULL) {
3268                         ret = WIMLIB_ERR_NOMEM;
3269                         free_lookup_table_entry(metadata_lte);
3270                         goto out_wimlib_free;
3271                 }
3272
3273                 ret = read_pwm_stream_header(pwm, metadata_lte, metadata_rspec, 0, NULL);
3274                 imd = pwm->image_metadata[i - 1];
3275                 imd->metadata_lte = metadata_lte;
3276                 if (ret) {
3277                         FREE(metadata_rspec);
3278                         goto out_wimlib_free;
3279                 }
3280
3281                 if (!(metadata_lte->flags & WIM_RESHDR_FLAG_METADATA)) {
3282                         ERROR("Expected metadata resource, but found "
3283                               "non-metadata stream.");
3284                         ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
3285                         goto out_wimlib_free;
3286                 }
3287
3288                 if (i == image) {
3289                         /* Metadata resource is for the image being extracted.
3290                          * Parse it and save the metadata in memory.  */
3291                         ret = read_metadata_resource(pwm, imd);
3292                         if (ret)
3293                                 goto out_wimlib_free;
3294                         imd->modified = 1;
3295                 } else {
3296                         /* Metadata resource is not for the image being
3297                          * extracted.  Skip over it.  */
3298                         ret = skip_wim_stream(metadata_lte);
3299                         if (ret)
3300                                 goto out_wimlib_free;
3301                 }
3302         }
3303         /* Extract the image.  */
3304         extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
3305         ret = do_wimlib_extract_image(pwm, image, target,
3306                                       extract_flags, progress_func);
3307         /* Clean up and return.  */
3308 out_wimlib_free:
3309         wimlib_free(pwm);
3310         return ret;
3311 }
3312
3313 WIMLIBAPI int
3314 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
3315                      int extract_flags, wimlib_progress_func_t progress_func)
3316 {
3317         if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
3318                 return WIMLIB_ERR_INVALID_PARAM;
3319         return do_wimlib_extract_image(wim, image, target, extract_flags,
3320                                        progress_func);
3321 }