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) 2012-2018 Eric Biggers
10 *
11 * This file is free software; you can redistribute it and/or modify it under
12 * the terms of the GNU Lesser General Public License as published by the Free
13 * Software Foundation; either version 3 of the License, or (at your option) any
14 * later version.
15 *
16 * This file is distributed in the hope that it will be useful, but WITHOUT
17 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
18 * FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
19 * details.
20 *
21 * You should have received a copy of the GNU Lesser General Public License
22 * along with this file; if not, see http://www.gnu.org/licenses/.
23 */
24
25 /*
26 * This file provides the API functions wimlib_extract_image(),
27 * wimlib_extract_image_from_pipe(), wimlib_extract_paths(), and
28 * wimlib_extract_pathlist(). Internally, all end up calling
29 * do_wimlib_extract_paths() and extract_trees().
30 *
31 * Although wimlib supports multiple extraction modes/backends (NTFS-3G, UNIX,
32 * Win32), this file does not itself have code to extract files or directories
33 * to any specific target; instead, it handles generic functionality and relies
34 * on lower-level callback functions declared in `struct apply_operations' to do
35 * the actual extraction.
36 */
37
38 #ifdef HAVE_CONFIG_H
39 # include "config.h"
40 #endif
41
42 #include <errno.h>
43 #include <fcntl.h>
44 #include <stdlib.h>
45 #include <sys/stat.h>
46 #include <unistd.h>
47
48 #include "wimlib/apply.h"
49 #include "wimlib/assert.h"
50 #include "wimlib/blob_table.h"
51 #include "wimlib/dentry.h"
52 #include "wimlib/encoding.h"
53 #include "wimlib/endianness.h"
54 #include "wimlib/error.h"
55 #include "wimlib/metadata.h"
56 #include "wimlib/object_id.h"
57 #include "wimlib/pathlist.h"
58 #include "wimlib/paths.h"
59 #include "wimlib/pattern.h"
60 #include "wimlib/reparse.h"
61 #include "wimlib/resource.h"
62 #include "wimlib/security.h"
63 #include "wimlib/unix_data.h"
64 #include "wimlib/wim.h"
65 #include "wimlib/win32.h" /* for realpath() equivalent */
66 #include "wimlib/xattr.h"
67 #include "wimlib/xml.h"
68
69 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x80000000
70 #define WIMLIB_EXTRACT_FLAG_IMAGEMODE 0x40000000
71
72 /* Keep in sync with wimlib.h */
73 #define WIMLIB_EXTRACT_MASK_PUBLIC \
74 (WIMLIB_EXTRACT_FLAG_NTFS | \
75 WIMLIB_EXTRACT_FLAG_UNIX_DATA | \
76 WIMLIB_EXTRACT_FLAG_NO_ACLS | \
77 WIMLIB_EXTRACT_FLAG_STRICT_ACLS | \
78 WIMLIB_EXTRACT_FLAG_RPFIX | \
79 WIMLIB_EXTRACT_FLAG_NORPFIX | \
80 WIMLIB_EXTRACT_FLAG_TO_STDOUT | \
81 WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES | \
82 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS | \
83 WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS | \
84 WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES | \
85 WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS | \
86 WIMLIB_EXTRACT_FLAG_GLOB_PATHS | \
87 WIMLIB_EXTRACT_FLAG_STRICT_GLOB | \
88 WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES | \
89 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE | \
90 WIMLIB_EXTRACT_FLAG_WIMBOOT | \
91 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K | \
92 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K | \
93 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K | \
94 WIMLIB_EXTRACT_FLAG_COMPACT_LZX \
95 )
96
97 /* Send WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE or
98 * WIMLIB_PROGRESS_MSG_EXTRACT_METADATA. */
99 int
do_file_extract_progress(struct apply_ctx * ctx,enum wimlib_progress_msg msg)100 do_file_extract_progress(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
101 {
102 ctx->count_until_file_progress = 500; /* Arbitrary value to limit calls */
103 return extract_progress(ctx, msg);
104 }
105
106 static int
start_file_phase(struct apply_ctx * ctx,u64 end_file_count,enum wimlib_progress_msg msg)107 start_file_phase(struct apply_ctx *ctx, u64 end_file_count, enum wimlib_progress_msg msg)
108 {
109 ctx->progress.extract.current_file_count = 0;
110 ctx->progress.extract.end_file_count = end_file_count;
111 return do_file_extract_progress(ctx, msg);
112 }
113
114 int
start_file_structure_phase(struct apply_ctx * ctx,u64 end_file_count)115 start_file_structure_phase(struct apply_ctx *ctx, u64 end_file_count)
116 {
117 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
118 }
119
120 int
start_file_metadata_phase(struct apply_ctx * ctx,u64 end_file_count)121 start_file_metadata_phase(struct apply_ctx *ctx, u64 end_file_count)
122 {
123 return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
124 }
125
126 static int
end_file_phase(struct apply_ctx * ctx,enum wimlib_progress_msg msg)127 end_file_phase(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
128 {
129 ctx->progress.extract.current_file_count = ctx->progress.extract.end_file_count;
130 return do_file_extract_progress(ctx, msg);
131 }
132
133 int
end_file_structure_phase(struct apply_ctx * ctx)134 end_file_structure_phase(struct apply_ctx *ctx)
135 {
136 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
137 }
138
139 int
end_file_metadata_phase(struct apply_ctx * ctx)140 end_file_metadata_phase(struct apply_ctx *ctx)
141 {
142 return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
143 }
144
145 /* Are all bytes in the specified buffer zero? */
146 static bool
is_all_zeroes(const u8 * p,const size_t size)147 is_all_zeroes(const u8 *p, const size_t size)
148 {
149 const u8 * const end = p + size;
150
151 for (; (uintptr_t)p % WORDBYTES && p != end; p++)
152 if (*p)
153 return false;
154
155 for (; end - p >= WORDBYTES; p += WORDBYTES)
156 if (*(const machine_word_t *)p)
157 return false;
158
159 for (; p != end; p++)
160 if (*p)
161 return false;
162
163 return true;
164 }
165
166 /*
167 * Sparse regions should be detected at the granularity of the filesystem block
168 * size. For now just assume 4096 bytes, which is the default block size on
169 * NTFS and most Linux filesystems.
170 */
171 #define SPARSE_UNIT 4096
172
173 /*
174 * Detect whether the specified buffer begins with a region of all zero bytes.
175 * Return %true if a zero region was found or %false if a nonzero region was
176 * found, and sets *len_ret to the length of the region. This operates at a
177 * granularity of SPARSE_UNIT bytes, meaning that to extend a zero region, there
178 * must be SPARSE_UNIT zero bytes with no interruption, but to extend a nonzero
179 * region, just one nonzero byte in the next SPARSE_UNIT bytes is sufficient.
180 *
181 * Note: besides compression, the WIM format doesn't yet have a way to
182 * efficiently represent zero regions, so that's why we need to detect them
183 * ourselves. Things will still fall apart badly on extremely large sparse
184 * files, but this is a start...
185 */
186 bool
detect_sparse_region(const void * data,size_t size,size_t * len_ret)187 detect_sparse_region(const void *data, size_t size, size_t *len_ret)
188 {
189 const void *p = data;
190 const void * const end = data + size;
191 size_t len = 0;
192 bool zeroes = false;
193
194 while (p != end) {
195 size_t n = min(end - p, SPARSE_UNIT);
196 bool z = is_all_zeroes(p, n);
197
198 if (len != 0 && z != zeroes)
199 break;
200 zeroes = z;
201 len += n;
202 p += n;
203 }
204
205 *len_ret = len;
206 return zeroes;
207 }
208
209 #define PWM_FOUND_WIM_HDR (-1)
210
211 /* Read the header for a blob in a pipable WIM. If @pwm_hdr_ret is not NULL,
212 * also look for a pipable WIM header and return PWM_FOUND_WIM_HDR if found. */
213 static int
read_pwm_blob_header(WIMStruct * pwm,u8 hash_ret[SHA1_HASH_SIZE],struct wim_reshdr * reshdr_ret,struct wim_header_disk * pwm_hdr_ret)214 read_pwm_blob_header(WIMStruct *pwm, u8 hash_ret[SHA1_HASH_SIZE],
215 struct wim_reshdr *reshdr_ret,
216 struct wim_header_disk *pwm_hdr_ret)
217 {
218 int ret;
219 struct pwm_blob_hdr blob_hdr;
220 u64 magic;
221
222 ret = full_read(&pwm->in_fd, &blob_hdr, sizeof(blob_hdr));
223 if (unlikely(ret))
224 goto read_error;
225
226 magic = le64_to_cpu(blob_hdr.magic);
227
228 if (magic == PWM_MAGIC && pwm_hdr_ret != NULL) {
229 memcpy(pwm_hdr_ret, &blob_hdr, sizeof(blob_hdr));
230 ret = full_read(&pwm->in_fd,
231 (u8 *)pwm_hdr_ret + sizeof(blob_hdr),
232 sizeof(*pwm_hdr_ret) - sizeof(blob_hdr));
233 if (unlikely(ret))
234 goto read_error;
235 return PWM_FOUND_WIM_HDR;
236 }
237
238 if (unlikely(magic != PWM_BLOB_MAGIC)) {
239 ERROR("Data read on pipe is invalid (expected blob header)");
240 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
241 }
242
243 copy_hash(hash_ret, blob_hdr.hash);
244
245 reshdr_ret->size_in_wim = 0; /* Not available */
246 reshdr_ret->flags = le32_to_cpu(blob_hdr.flags);
247 reshdr_ret->offset_in_wim = pwm->in_fd.offset;
248 reshdr_ret->uncompressed_size = le64_to_cpu(blob_hdr.uncompressed_size);
249
250 if (unlikely(reshdr_ret->uncompressed_size == 0)) {
251 ERROR("Data read on pipe is invalid (resource is of 0 size)");
252 return WIMLIB_ERR_INVALID_PIPABLE_WIM;
253 }
254
255 return 0;
256
257 read_error:
258 if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE)
259 ERROR("The pipe ended before all needed data was sent!");
260 else
261 ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
262 return ret;
263 }
264
265 static int
read_blobs_from_pipe(struct apply_ctx * ctx,const struct read_blob_callbacks * cbs)266 read_blobs_from_pipe(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
267 {
268 int ret;
269 u8 hash[SHA1_HASH_SIZE];
270 struct wim_reshdr reshdr;
271 struct wim_header_disk pwm_hdr;
272 struct wim_resource_descriptor rdesc;
273 struct blob_descriptor *blob;
274
275 copy_guid(ctx->progress.extract.guid, ctx->wim->hdr.guid);
276 ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
277 ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
278 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
279 if (ret)
280 return ret;
281
282 while (ctx->num_blobs_remaining) {
283
284 ret = read_pwm_blob_header(ctx->wim, hash, &reshdr, &pwm_hdr);
285
286 if (ret == PWM_FOUND_WIM_HDR) {
287 u16 part_number = le16_to_cpu(pwm_hdr.part_number);
288 u16 total_parts = le16_to_cpu(pwm_hdr.total_parts);
289
290 if (part_number == ctx->progress.extract.part_number &&
291 total_parts == ctx->progress.extract.total_parts &&
292 guids_equal(pwm_hdr.guid, ctx->progress.extract.guid))
293 continue;
294
295 copy_guid(ctx->progress.extract.guid, pwm_hdr.guid);
296 ctx->progress.extract.part_number = part_number;
297 ctx->progress.extract.total_parts = total_parts;
298 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
299 if (ret)
300 return ret;
301
302 continue;
303 }
304
305 if (ret)
306 return ret;
307
308 if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)
309 && (blob = lookup_blob(ctx->wim->blob_table, hash))
310 && (blob->out_refcnt))
311 {
312 wim_reshdr_to_desc_and_blob(&reshdr, ctx->wim, &rdesc, blob);
313 ret = read_blob_with_sha1(blob, cbs);
314 blob_unset_is_located_in_wim_resource(blob);
315 if (ret)
316 return ret;
317 ctx->num_blobs_remaining--;
318 } else {
319 wim_reshdr_to_desc(&reshdr, ctx->wim, &rdesc);
320 ret = skip_wim_resource(&rdesc);
321 if (ret)
322 return ret;
323 }
324 }
325
326 return 0;
327 }
328
329 static int
handle_pwm_metadata_resource(WIMStruct * pwm,int image,bool is_needed)330 handle_pwm_metadata_resource(WIMStruct *pwm, int image, bool is_needed)
331 {
332 struct blob_descriptor *blob;
333 struct wim_reshdr reshdr;
334 struct wim_resource_descriptor *rdesc;
335 int ret;
336
337 ret = WIMLIB_ERR_NOMEM;
338 blob = new_blob_descriptor();
339 if (!blob)
340 goto out;
341
342 ret = read_pwm_blob_header(pwm, blob->hash, &reshdr, NULL);
343 if (ret)
344 goto out;
345
346 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
347 if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)) {
348 ERROR("Expected metadata resource, but found non-metadata "
349 "resource");
350 goto out;
351 }
352
353 ret = WIMLIB_ERR_NOMEM;
354 rdesc = MALLOC(sizeof(*rdesc));
355 if (!rdesc)
356 goto out;
357
358 wim_reshdr_to_desc_and_blob(&reshdr, pwm, rdesc, blob);
359 pwm->refcnt++;
360
361 ret = WIMLIB_ERR_NOMEM;
362 pwm->image_metadata[image - 1] = new_unloaded_image_metadata(blob);
363 if (!pwm->image_metadata[image - 1])
364 goto out;
365 blob = NULL;
366
367 /* If the metadata resource is for the image being extracted, then parse
368 * it and save the metadata in memory. Otherwise, skip over it. */
369 if (is_needed)
370 ret = select_wim_image(pwm, image);
371 else
372 ret = skip_wim_resource(rdesc);
373 out:
374 free_blob_descriptor(blob);
375 return ret;
376 }
377
378 /* Creates a temporary file opened for writing. The open file descriptor is
379 * returned in @fd_ret and its name is returned in @name_ret (dynamically
380 * allocated). */
381 static int
create_temporary_file(struct filedes * fd_ret,tchar ** name_ret)382 create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
383 {
384 tchar *name;
385 int raw_fd;
386
387 #ifdef __WIN32__
388 retry:
389 name = _wtempnam(NULL, L"wimlib");
390 if (!name) {
391 ERROR_WITH_ERRNO("Failed to create temporary filename");
392 return WIMLIB_ERR_NOMEM;
393 }
394 raw_fd = _wopen(name, O_WRONLY | O_CREAT | O_EXCL | O_BINARY |
395 _O_SHORT_LIVED, 0600);
396 if (raw_fd < 0 && errno == EEXIST) {
397 FREE(name);
398 goto retry;
399 }
400 #else /* __WIN32__ */
401 const char *tmpdir = getenv("TMPDIR");
402 if (!tmpdir)
403 tmpdir = P_tmpdir;
404 name = MALLOC(strlen(tmpdir) + 1 + 6 + 6 + 1);
405 if (!name)
406 return WIMLIB_ERR_NOMEM;
407 sprintf(name, "%s/wimlibXXXXXX", tmpdir);
408 raw_fd = mkstemp(name);
409 #endif /* !__WIN32__ */
410
411 if (raw_fd < 0) {
412 ERROR_WITH_ERRNO("Failed to create temporary file "
413 "\"%"TS"\"", name);
414 FREE(name);
415 return WIMLIB_ERR_OPEN;
416 }
417
418 filedes_init(fd_ret, raw_fd);
419 *name_ret = name;
420 return 0;
421 }
422
423 static int
begin_extract_blob(struct blob_descriptor * blob,void * _ctx)424 begin_extract_blob(struct blob_descriptor *blob, void *_ctx)
425 {
426 struct apply_ctx *ctx = _ctx;
427
428 if (unlikely(blob->out_refcnt > MAX_OPEN_FILES))
429 return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
430
431 return call_begin_blob(blob, ctx->saved_cbs);
432 }
433
434 static int
extract_chunk(const struct blob_descriptor * blob,u64 offset,const void * chunk,size_t size,void * _ctx)435 extract_chunk(const struct blob_descriptor *blob, u64 offset,
436 const void *chunk, size_t size, void *_ctx)
437 {
438 struct apply_ctx *ctx = _ctx;
439 union wimlib_progress_info *progress = &ctx->progress;
440 bool last = (offset + size == blob->size);
441 int ret;
442
443 if (likely(ctx->supported_features.hard_links)) {
444 progress->extract.completed_bytes +=
445 (u64)size * blob->out_refcnt;
446 if (last)
447 progress->extract.completed_streams += blob->out_refcnt;
448 } else {
449 const struct blob_extraction_target *targets =
450 blob_extraction_targets(blob);
451 for (u32 i = 0; i < blob->out_refcnt; i++) {
452 const struct wim_inode *inode = targets[i].inode;
453 const struct wim_dentry *dentry;
454
455 inode_for_each_extraction_alias(dentry, inode) {
456 progress->extract.completed_bytes += size;
457 if (last)
458 progress->extract.completed_streams++;
459 }
460 }
461 }
462 if (progress->extract.completed_bytes >= ctx->next_progress) {
463
464 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
465 if (ret)
466 return ret;
467
468 set_next_progress(progress->extract.completed_bytes,
469 progress->extract.total_bytes,
470 &ctx->next_progress);
471 }
472
473 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
474 /* Just extracting to temporary file for now. */
475 ret = full_write(&ctx->tmpfile_fd, chunk, size);
476 if (ret) {
477 ERROR_WITH_ERRNO("Error writing data to "
478 "temporary file \"%"TS"\"",
479 ctx->tmpfile_name);
480 }
481 return ret;
482 }
483
484 return call_continue_blob(blob, offset, chunk, size, ctx->saved_cbs);
485 }
486
487 /* Copy the blob's data from the temporary file to each of its targets.
488 *
489 * This is executed only in the very uncommon case that a blob is being
490 * extracted to more than MAX_OPEN_FILES targets! */
491 static int
extract_from_tmpfile(const tchar * tmpfile_name,const struct blob_descriptor * orig_blob,const struct read_blob_callbacks * cbs)492 extract_from_tmpfile(const tchar *tmpfile_name,
493 const struct blob_descriptor *orig_blob,
494 const struct read_blob_callbacks *cbs)
495 {
496 struct blob_descriptor tmpfile_blob;
497 const struct blob_extraction_target *targets = blob_extraction_targets(orig_blob);
498 int ret;
499
500 memcpy(&tmpfile_blob, orig_blob, sizeof(struct blob_descriptor));
501 tmpfile_blob.blob_location = BLOB_IN_FILE_ON_DISK;
502 tmpfile_blob.file_on_disk = (tchar *)tmpfile_name;
503 tmpfile_blob.out_refcnt = 1;
504
505 for (u32 i = 0; i < orig_blob->out_refcnt; i++) {
506 tmpfile_blob.inline_blob_extraction_targets[0] = targets[i];
507 ret = read_blob_with_cbs(&tmpfile_blob, cbs);
508 if (ret)
509 return ret;
510 }
511 return 0;
512 }
513
514 static int
end_extract_blob(struct blob_descriptor * blob,int status,void * _ctx)515 end_extract_blob(struct blob_descriptor *blob, int status, void *_ctx)
516 {
517 struct apply_ctx *ctx = _ctx;
518
519 if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
520 filedes_close(&ctx->tmpfile_fd);
521 if (!status)
522 status = extract_from_tmpfile(ctx->tmpfile_name, blob,
523 ctx->saved_cbs);
524 filedes_invalidate(&ctx->tmpfile_fd);
525 tunlink(ctx->tmpfile_name);
526 FREE(ctx->tmpfile_name);
527 return status;
528 }
529
530 return call_end_blob(blob, status, ctx->saved_cbs);
531 }
532
533 /*
534 * Read the list of blobs to extract and feed their data into the specified
535 * callback functions.
536 *
537 * This handles checksumming each blob.
538 *
539 * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
540 *
541 * This also works if the WIM is being read from a pipe.
542 *
543 * This also will split up blobs that will need to be extracted to more than
544 * MAX_OPEN_FILES locations, as measured by the 'out_refcnt' of each blob.
545 * Therefore, the apply_operations implementation need not worry about running
546 * out of file descriptors, unless it might open more than one file descriptor
547 * per 'blob_extraction_target' (e.g. Win32 currently might because the
548 * destination file system might not support hard links).
549 */
550 int
extract_blob_list(struct apply_ctx * ctx,const struct read_blob_callbacks * cbs)551 extract_blob_list(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
552 {
553 struct read_blob_callbacks wrapper_cbs = {
554 .begin_blob = begin_extract_blob,
555 .continue_blob = extract_chunk,
556 .end_blob = end_extract_blob,
557 .ctx = ctx,
558 };
559 ctx->saved_cbs = cbs;
560 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
561 return read_blobs_from_pipe(ctx, &wrapper_cbs);
562 } else {
563 return read_blob_list(&ctx->blob_list,
564 offsetof(struct blob_descriptor,
565 extraction_list),
566 &wrapper_cbs, VERIFY_BLOB_HASHES);
567 }
568 }
569
570 /* Extract a WIM dentry to standard output.
571 *
572 * This obviously doesn't make sense in all cases. We return an error if the
573 * dentry does not correspond to a regular file. Otherwise we extract the
574 * unnamed data stream only. */
575 static int
extract_dentry_to_stdout(struct wim_dentry * dentry,const struct blob_table * blob_table)576 extract_dentry_to_stdout(struct wim_dentry *dentry,
577 const struct blob_table *blob_table)
578 {
579 struct wim_inode *inode = dentry->d_inode;
580 struct blob_descriptor *blob;
581 struct filedes _stdout;
582
583 if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
584 FILE_ATTRIBUTE_DIRECTORY |
585 FILE_ATTRIBUTE_ENCRYPTED))
586 {
587 ERROR("\"%"TS"\" is not a regular file and therefore cannot be "
588 "extracted to standard output", dentry_full_path(dentry));
589 return WIMLIB_ERR_NOT_A_REGULAR_FILE;
590 }
591
592 blob = inode_get_blob_for_unnamed_data_stream(inode, blob_table);
593 if (!blob) {
594 const u8 *hash = inode_get_hash_of_unnamed_data_stream(inode);
595 if (!is_zero_hash(hash))
596 return blob_not_found_error(inode, hash);
597 return 0;
598 }
599
600 filedes_init(&_stdout, STDOUT_FILENO);
601 return extract_blob_to_fd(blob, &_stdout);
602 }
603
604 static int
extract_dentries_to_stdout(struct wim_dentry ** dentries,size_t num_dentries,const struct blob_table * blob_table)605 extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
606 const struct blob_table *blob_table)
607 {
608 for (size_t i = 0; i < num_dentries; i++) {
609 int ret = extract_dentry_to_stdout(dentries[i], blob_table);
610 if (ret)
611 return ret;
612 }
613 return 0;
614 }
615
616 /**********************************************************************/
617
618 /*
619 * Removes duplicate dentries from the array.
620 *
621 * Returns the new number of dentries, packed at the front of the array.
622 */
623 static size_t
remove_duplicate_trees(struct wim_dentry ** trees,size_t num_trees)624 remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees)
625 {
626 size_t i, j = 0;
627 for (i = 0; i < num_trees; i++) {
628 if (!trees[i]->d_tmp_flag) {
629 /* Found distinct dentry. */
630 trees[i]->d_tmp_flag = 1;
631 trees[j++] = trees[i];
632 }
633 }
634 for (i = 0; i < j; i++)
635 trees[i]->d_tmp_flag = 0;
636 return j;
637 }
638
639 /*
640 * Remove dentries that are descendants of other dentries in the array.
641 *
642 * Returns the new number of dentries, packed at the front of the array.
643 */
644 static size_t
remove_contained_trees(struct wim_dentry ** trees,size_t num_trees)645 remove_contained_trees(struct wim_dentry **trees, size_t num_trees)
646 {
647 size_t i, j = 0;
648 for (i = 0; i < num_trees; i++)
649 trees[i]->d_tmp_flag = 1;
650 for (i = 0; i < num_trees; i++) {
651 struct wim_dentry *d = trees[i];
652 while (!dentry_is_root(d)) {
653 d = d->d_parent;
654 if (d->d_tmp_flag)
655 goto tree_contained;
656 }
657 trees[j++] = trees[i];
658 continue;
659
660 tree_contained:
661 trees[i]->d_tmp_flag = 0;
662 }
663
664 for (i = 0; i < j; i++)
665 trees[i]->d_tmp_flag = 0;
666 return j;
667 }
668
669 static int
dentry_append_to_list(struct wim_dentry * dentry,void * _dentry_list)670 dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
671 {
672 struct list_head *dentry_list = _dentry_list;
673 list_add_tail(&dentry->d_extraction_list_node, dentry_list);
674 return 0;
675 }
676
677 static void
dentry_reset_extraction_list_node(struct wim_dentry * dentry)678 dentry_reset_extraction_list_node(struct wim_dentry *dentry)
679 {
680 dentry->d_extraction_list_node = (struct list_head){NULL, NULL};
681 }
682
683 static int
dentry_delete_from_list(struct wim_dentry * dentry,void * _ignore)684 dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore)
685 {
686 if (will_extract_dentry(dentry)) {
687 list_del(&dentry->d_extraction_list_node);
688 dentry_reset_extraction_list_node(dentry);
689 }
690 return 0;
691 }
692
693 /*
694 * Build the preliminary list of dentries to be extracted.
695 *
696 * The list maintains the invariant that if d1 and d2 are in the list and d1 is
697 * an ancestor of d2, then d1 appears before d2 in the list.
698 */
699 static void
build_dentry_list(struct list_head * dentry_list,struct wim_dentry ** trees,size_t num_trees,bool add_ancestors)700 build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees,
701 size_t num_trees, bool add_ancestors)
702 {
703 INIT_LIST_HEAD(dentry_list);
704
705 /* Add the trees recursively. */
706 for (size_t i = 0; i < num_trees; i++)
707 for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list);
708
709 /* If requested, add ancestors of the trees. */
710 if (add_ancestors) {
711 for (size_t i = 0; i < num_trees; i++) {
712 struct wim_dentry *dentry = trees[i];
713 struct wim_dentry *ancestor;
714 struct list_head *place_after;
715
716 if (dentry_is_root(dentry))
717 continue;
718
719 place_after = dentry_list;
720 ancestor = dentry;
721 do {
722 ancestor = ancestor->d_parent;
723 if (will_extract_dentry(ancestor)) {
724 place_after = &ancestor->d_extraction_list_node;
725 break;
726 }
727 } while (!dentry_is_root(ancestor));
728
729 ancestor = dentry;
730 do {
731 ancestor = ancestor->d_parent;
732 if (will_extract_dentry(ancestor))
733 break;
734 list_add(&ancestor->d_extraction_list_node, place_after);
735 } while (!dentry_is_root(ancestor));
736 }
737 }
738 }
739
740 static void
destroy_dentry_list(struct list_head * dentry_list)741 destroy_dentry_list(struct list_head *dentry_list)
742 {
743 struct wim_dentry *dentry, *tmp;
744 struct wim_inode *inode;
745
746 list_for_each_entry_safe(dentry, tmp, dentry_list, d_extraction_list_node) {
747 inode = dentry->d_inode;
748 dentry_reset_extraction_list_node(dentry);
749 inode->i_visited = 0;
750 inode->i_can_externally_back = 0;
751 if ((void *)dentry->d_extraction_name != (void *)dentry->d_name)
752 FREE(dentry->d_extraction_name);
753 dentry->d_extraction_name = NULL;
754 dentry->d_extraction_name_nchars = 0;
755 }
756 }
757
758 static void
destroy_blob_list(struct list_head * blob_list)759 destroy_blob_list(struct list_head *blob_list)
760 {
761 struct blob_descriptor *blob;
762
763 list_for_each_entry(blob, blob_list, extraction_list)
764 if (blob->out_refcnt > ARRAY_LEN(blob->inline_blob_extraction_targets))
765 FREE(blob->blob_extraction_targets);
766 }
767
768 #ifdef __WIN32__
769 static const utf16lechar replacement_char = cpu_to_le16(0xfffd);
770 #else
771 static const utf16lechar replacement_char = cpu_to_le16('?');
772 #endif
773
774 static bool
file_name_valid(utf16lechar * name,size_t num_chars,bool fix)775 file_name_valid(utf16lechar *name, size_t num_chars, bool fix)
776 {
777 size_t i;
778
779 if (num_chars == 0)
780 return true;
781 for (i = 0; i < num_chars; i++) {
782 switch (le16_to_cpu(name[i])) {
783 #ifdef __WIN32__
784 case '\x01'...'\x1F':
785 case '\\':
786 case ':':
787 case '*':
788 case '?':
789 case '"':
790 case '<':
791 case '>':
792 case '|':
793 #endif
794 case '/':
795 case '\0':
796 if (fix)
797 name[i] = replacement_char;
798 else
799 return false;
800 }
801 }
802
803 return true;
804 }
805
806 static int
dentry_calculate_extraction_name(struct wim_dentry * dentry,struct apply_ctx * ctx)807 dentry_calculate_extraction_name(struct wim_dentry *dentry,
808 struct apply_ctx *ctx)
809 {
810 int ret;
811
812 if (dentry_is_root(dentry))
813 return 0;
814
815 #ifdef WITH_NTFS_3G
816 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
817 dentry->d_extraction_name = dentry->d_name;
818 dentry->d_extraction_name_nchars = dentry->d_name_nbytes /
819 sizeof(utf16lechar);
820 return 0;
821 }
822 #endif
823
824 if (!ctx->supported_features.case_sensitive_filenames) {
825 struct wim_dentry *other;
826 dentry_for_each_ci_match(other, dentry) {
827 if (will_extract_dentry(other)) {
828 if (ctx->extract_flags &
829 WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) {
830 WARNING("\"%"TS"\" has the same "
831 "case-insensitive name as "
832 "\"%"TS"\"; extracting "
833 "dummy name instead",
834 dentry_full_path(dentry),
835 dentry_full_path(other));
836 goto out_replace;
837 } else {
838 WARNING("Not extracting \"%"TS"\": "
839 "has same case-insensitive "
840 "name as \"%"TS"\"",
841 dentry_full_path(dentry),
842 dentry_full_path(other));
843 goto skip_dentry;
844 }
845 }
846 }
847 }
848
849 if (file_name_valid(dentry->d_name, dentry->d_name_nbytes / 2, false)) {
850 size_t nbytes = 0;
851 ret = utf16le_get_tstr(dentry->d_name,
852 dentry->d_name_nbytes,
853 (const tchar **)&dentry->d_extraction_name,
854 &nbytes);
855 dentry->d_extraction_name_nchars = nbytes / sizeof(tchar);
856 return ret;
857 } else {
858 if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES)
859 {
860 WARNING("\"%"TS"\" has an invalid filename "
861 "that is not supported on this platform; "
862 "extracting dummy name instead",
863 dentry_full_path(dentry));
864 goto out_replace;
865 } else {
866 WARNING("Not extracting \"%"TS"\": has an invalid filename "
867 "that is not supported on this platform",
868 dentry_full_path(dentry));
869 goto skip_dentry;
870 }
871 }
872
873 out_replace:
874 {
875 utf16lechar utf16_name_copy[dentry->d_name_nbytes / 2];
876
877 memcpy(utf16_name_copy, dentry->d_name, dentry->d_name_nbytes);
878 file_name_valid(utf16_name_copy, dentry->d_name_nbytes / 2, true);
879
880 const tchar *tchar_name;
881 size_t tchar_nchars;
882
883 ret = utf16le_get_tstr(utf16_name_copy,
884 dentry->d_name_nbytes,
885 &tchar_name, &tchar_nchars);
886 if (ret)
887 return ret;
888
889 tchar_nchars /= sizeof(tchar);
890
891 size_t fixed_name_num_chars = tchar_nchars;
892 tchar fixed_name[tchar_nchars + 50];
893
894 tmemcpy(fixed_name, tchar_name, tchar_nchars);
895 fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars,
896 T(" (invalid filename #%lu)"),
897 ++ctx->invalid_sequence);
898
899 utf16le_put_tstr(tchar_name);
900
901 dentry->d_extraction_name = TSTRDUP(fixed_name);
902 if (!dentry->d_extraction_name)
903 return WIMLIB_ERR_NOMEM;
904 dentry->d_extraction_name_nchars = fixed_name_num_chars;
905 }
906 return 0;
907
908 skip_dentry:
909 for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
910 return 0;
911 }
912
913 /*
914 * Calculate the actual filename component at which each WIM dentry will be
915 * extracted, with special handling for dentries that are unsupported by the
916 * extraction backend or have invalid names.
917 *
918 * ctx->supported_features must be filled in.
919 *
920 * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING
921 */
922 static int
dentry_list_calculate_extraction_names(struct list_head * dentry_list,struct apply_ctx * ctx)923 dentry_list_calculate_extraction_names(struct list_head *dentry_list,
924 struct apply_ctx *ctx)
925 {
926 struct list_head *prev, *cur;
927
928 /* Can't use list_for_each_entry() because a call to
929 * dentry_calculate_extraction_name() may delete the current dentry and
930 * its children from the list. */
931
932 prev = dentry_list;
933 for (;;) {
934 struct wim_dentry *dentry;
935 int ret;
936
937 cur = prev->next;
938 if (cur == dentry_list)
939 break;
940
941 dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node);
942
943 ret = dentry_calculate_extraction_name(dentry, ctx);
944 if (ret)
945 return ret;
946
947 if (prev->next == cur)
948 prev = cur;
949 else
950 ; /* Current dentry and its children (which follow in
951 the list) were deleted. prev stays the same. */
952 }
953 return 0;
954 }
955
956 static int
dentry_resolve_streams(struct wim_dentry * dentry,int extract_flags,struct blob_table * blob_table)957 dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags,
958 struct blob_table *blob_table)
959 {
960 struct wim_inode *inode = dentry->d_inode;
961 struct blob_descriptor *blob;
962 int ret;
963 bool force = false;
964
965 /* Special case: when extracting from a pipe, the WIM blob table is
966 * initially empty, so "resolving" an inode's streams is initially not
967 * possible. However, we still need to keep track of which blobs,
968 * identified by SHA-1 message digests, need to be extracted, so we
969 * "resolve" the inode's streams anyway by allocating a 'struct
970 * blob_descriptor' for each one. */
971 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
972 force = true;
973 ret = inode_resolve_streams(inode, blob_table, force);
974 if (ret)
975 return ret;
976 for (unsigned i = 0; i < inode->i_num_streams; i++) {
977 blob = stream_blob_resolved(&inode->i_streams[i]);
978 if (blob)
979 blob->out_refcnt = 0;
980 }
981 return 0;
982 }
983
984 /*
985 * For each dentry to be extracted, resolve all streams in the corresponding
986 * inode and set 'out_refcnt' in all referenced blob_descriptors to 0.
987 *
988 * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM.
989 */
990 static int
dentry_list_resolve_streams(struct list_head * dentry_list,struct apply_ctx * ctx)991 dentry_list_resolve_streams(struct list_head *dentry_list,
992 struct apply_ctx *ctx)
993 {
994 struct wim_dentry *dentry;
995 int ret;
996
997 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
998 ret = dentry_resolve_streams(dentry,
999 ctx->extract_flags,
1000 ctx->wim->blob_table);
1001 if (ret)
1002 return ret;
1003 }
1004 return 0;
1005 }
1006
1007 static int
ref_stream(struct wim_inode_stream * strm,struct wim_dentry * dentry,struct apply_ctx * ctx)1008 ref_stream(struct wim_inode_stream *strm, struct wim_dentry *dentry,
1009 struct apply_ctx *ctx)
1010 {
1011 struct wim_inode *inode = dentry->d_inode;
1012 struct blob_descriptor *blob = stream_blob_resolved(strm);
1013 struct blob_extraction_target *targets;
1014
1015 if (!blob)
1016 return 0;
1017
1018 /* Tally the size only for each actual extraction of the stream (not
1019 * additional hard links to the inode). */
1020 if (inode->i_visited && ctx->supported_features.hard_links)
1021 return 0;
1022
1023 ctx->progress.extract.total_bytes += blob->size;
1024 ctx->progress.extract.total_streams++;
1025
1026 if (inode->i_visited)
1027 return 0;
1028
1029 /* Add each blob to 'ctx->blob_list' only one time, regardless of how
1030 * many extraction targets it will have. */
1031 if (blob->out_refcnt == 0) {
1032 list_add_tail(&blob->extraction_list, &ctx->blob_list);
1033 ctx->num_blobs_remaining++;
1034 }
1035
1036 /* Set this stream as an extraction target of 'blob'. */
1037
1038 if (blob->out_refcnt < ARRAY_LEN(blob->inline_blob_extraction_targets)) {
1039 targets = blob->inline_blob_extraction_targets;
1040 } else {
1041 struct blob_extraction_target *prev_targets;
1042 size_t alloc_blob_extraction_targets;
1043
1044 if (blob->out_refcnt == ARRAY_LEN(blob->inline_blob_extraction_targets)) {
1045 prev_targets = NULL;
1046 alloc_blob_extraction_targets = ARRAY_LEN(blob->inline_blob_extraction_targets);
1047 } else {
1048 prev_targets = blob->blob_extraction_targets;
1049 alloc_blob_extraction_targets = blob->alloc_blob_extraction_targets;
1050 }
1051
1052 if (blob->out_refcnt == alloc_blob_extraction_targets) {
1053 alloc_blob_extraction_targets *= 2;
1054 targets = REALLOC(prev_targets,
1055 alloc_blob_extraction_targets *
1056 sizeof(targets[0]));
1057 if (!targets)
1058 return WIMLIB_ERR_NOMEM;
1059 if (!prev_targets) {
1060 memcpy(targets,
1061 blob->inline_blob_extraction_targets,
1062 sizeof(blob->inline_blob_extraction_targets));
1063 }
1064 blob->blob_extraction_targets = targets;
1065 blob->alloc_blob_extraction_targets = alloc_blob_extraction_targets;
1066 }
1067 targets = blob->blob_extraction_targets;
1068 }
1069 targets[blob->out_refcnt].inode = inode;
1070 targets[blob->out_refcnt].stream = strm;
1071 blob->out_refcnt++;
1072 return 0;
1073 }
1074
1075 static int
ref_stream_if_needed(struct wim_dentry * dentry,struct wim_inode * inode,struct wim_inode_stream * strm,struct apply_ctx * ctx)1076 ref_stream_if_needed(struct wim_dentry *dentry, struct wim_inode *inode,
1077 struct wim_inode_stream *strm, struct apply_ctx *ctx)
1078 {
1079 bool need_stream = false;
1080 switch (strm->stream_type) {
1081 case STREAM_TYPE_DATA:
1082 if (stream_is_named(strm)) {
1083 /* Named data stream */
1084 if (ctx->supported_features.named_data_streams)
1085 need_stream = true;
1086 } else if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
1087 FILE_ATTRIBUTE_ENCRYPTED))
1088 && !(inode_is_symlink(inode)
1089 && !ctx->supported_features.reparse_points
1090 && ctx->supported_features.symlink_reparse_points))
1091 {
1092 /*
1093 * Unnamed data stream. Skip if any of the following is true:
1094 *
1095 * - file is a directory
1096 * - file is encrypted
1097 * - backend needs to create the file as UNIX symlink
1098 * - backend will extract the stream as externally
1099 * backed from the WIM archive itself
1100 */
1101 if (ctx->apply_ops->will_back_from_wim) {
1102 int ret = (*ctx->apply_ops->will_back_from_wim)(dentry, ctx);
1103 if (ret > 0) /* Error? */
1104 return ret;
1105 if (ret < 0) /* Won't externally back? */
1106 need_stream = true;
1107 } else {
1108 need_stream = true;
1109 }
1110 }
1111 break;
1112 case STREAM_TYPE_REPARSE_POINT:
1113 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT);
1114 if (ctx->supported_features.reparse_points ||
1115 (inode_is_symlink(inode) &&
1116 ctx->supported_features.symlink_reparse_points))
1117 need_stream = true;
1118 break;
1119 case STREAM_TYPE_EFSRPC_RAW_DATA:
1120 wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED);
1121 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
1122 if (ctx->supported_features.encrypted_directories)
1123 need_stream = true;
1124 } else {
1125 if (ctx->supported_features.encrypted_files)
1126 need_stream = true;
1127 }
1128 break;
1129 }
1130 if (need_stream)
1131 return ref_stream(strm, dentry, ctx);
1132 return 0;
1133 }
1134
1135 static int
dentry_ref_streams(struct wim_dentry * dentry,struct apply_ctx * ctx)1136 dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
1137 {
1138 struct wim_inode *inode = dentry->d_inode;
1139 for (unsigned i = 0; i < inode->i_num_streams; i++) {
1140 int ret = ref_stream_if_needed(dentry, inode,
1141 &inode->i_streams[i], ctx);
1142 if (ret)
1143 return ret;
1144 }
1145 inode->i_visited = 1;
1146 return 0;
1147 }
1148
1149 /*
1150 * Given a list of dentries to be extracted, build the list of blobs that need
1151 * to be extracted, and for each blob determine the streams to which that blob
1152 * will be extracted.
1153 *
1154 * This also initializes the extract progress info with byte and blob
1155 * information.
1156 *
1157 * ctx->supported_features must be filled in.
1158 */
1159 static int
dentry_list_ref_streams(struct list_head * dentry_list,struct apply_ctx * ctx)1160 dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx)
1161 {
1162 struct wim_dentry *dentry;
1163 int ret;
1164
1165 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1166 ret = dentry_ref_streams(dentry, ctx);
1167 if (ret)
1168 return ret;
1169 }
1170 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1171 dentry->d_inode->i_visited = 0;
1172 return 0;
1173 }
1174
1175 static void
dentry_list_build_inode_alias_lists(struct list_head * dentry_list)1176 dentry_list_build_inode_alias_lists(struct list_head *dentry_list)
1177 {
1178 struct wim_dentry *dentry;
1179
1180 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1181 dentry->d_inode->i_first_extraction_alias = NULL;
1182
1183 list_for_each_entry(dentry, dentry_list, d_extraction_list_node) {
1184 dentry->d_next_extraction_alias = dentry->d_inode->i_first_extraction_alias;
1185 dentry->d_inode->i_first_extraction_alias = dentry;
1186 }
1187 }
1188
1189 static void
inode_tally_features(const struct wim_inode * inode,struct wim_features * features)1190 inode_tally_features(const struct wim_inode *inode,
1191 struct wim_features *features)
1192 {
1193 if (inode->i_attributes & FILE_ATTRIBUTE_READONLY)
1194 features->readonly_files++;
1195 if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN)
1196 features->hidden_files++;
1197 if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM)
1198 features->system_files++;
1199 if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE)
1200 features->archive_files++;
1201 if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED)
1202 features->compressed_files++;
1203 if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) {
1204 if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY)
1205 features->encrypted_directories++;
1206 else
1207 features->encrypted_files++;
1208 }
1209 if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED)
1210 features->not_context_indexed_files++;
1211 if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE)
1212 features->sparse_files++;
1213 if (inode_has_named_data_stream(inode))
1214 features->named_data_streams++;
1215 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1216 features->reparse_points++;
1217 if (inode_is_symlink(inode))
1218 features->symlink_reparse_points++;
1219 else
1220 features->other_reparse_points++;
1221 }
1222 if (inode_has_security_descriptor(inode))
1223 features->security_descriptors++;
1224 if (inode_has_unix_data(inode))
1225 features->unix_data++;
1226 if (inode_has_object_id(inode))
1227 features->object_ids++;
1228 if (inode_has_xattrs(inode))
1229 features->xattrs++;
1230 }
1231
1232 /* Tally features necessary to extract a dentry and the corresponding inode. */
1233 static void
dentry_tally_features(struct wim_dentry * dentry,struct wim_features * features)1234 dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features)
1235 {
1236 struct wim_inode *inode = dentry->d_inode;
1237
1238 if (dentry_has_short_name(dentry))
1239 features->short_names++;
1240
1241 if (inode->i_visited) {
1242 features->hard_links++;
1243 } else {
1244 inode_tally_features(inode, features);
1245 inode->i_visited = 1;
1246 }
1247 }
1248
1249 /* Tally the features necessary to extract the specified dentries. */
1250 static void
dentry_list_get_features(struct list_head * dentry_list,struct wim_features * features)1251 dentry_list_get_features(struct list_head *dentry_list,
1252 struct wim_features *features)
1253 {
1254 struct wim_dentry *dentry;
1255
1256 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1257 dentry_tally_features(dentry, features);
1258
1259 list_for_each_entry(dentry, dentry_list, d_extraction_list_node)
1260 dentry->d_inode->i_visited = 0;
1261 }
1262
1263 static int
do_feature_check(const struct wim_features * required_features,const struct wim_features * supported_features,int extract_flags)1264 do_feature_check(const struct wim_features *required_features,
1265 const struct wim_features *supported_features,
1266 int extract_flags)
1267 {
1268 /* Encrypted files. */
1269 if (required_features->encrypted_files &&
1270 !supported_features->encrypted_files)
1271 WARNING("Ignoring EFS-encrypted data of %lu files",
1272 required_features->encrypted_files);
1273
1274 /* Named data streams. */
1275 if (required_features->named_data_streams &&
1276 !supported_features->named_data_streams)
1277 WARNING("Ignoring named data streams of %lu files",
1278 required_features->named_data_streams);
1279
1280 /* File attributes. */
1281 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) {
1282
1283 if (required_features->readonly_files &&
1284 !supported_features->readonly_files)
1285 WARNING("Ignoring FILE_ATTRIBUTE_READONLY of %lu files",
1286 required_features->readonly_files);
1287
1288 if (required_features->hidden_files &&
1289 !supported_features->hidden_files)
1290 WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files",
1291 required_features->hidden_files);
1292
1293 if (required_features->system_files &&
1294 !supported_features->system_files)
1295 WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files",
1296 required_features->system_files);
1297
1298 /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE.
1299 * We're an archive program, so theoretically we can do what we
1300 * want with it. */
1301
1302 if (required_features->compressed_files &&
1303 !supported_features->compressed_files)
1304 WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files",
1305 required_features->compressed_files);
1306
1307 if (required_features->not_context_indexed_files &&
1308 !supported_features->not_context_indexed_files)
1309 WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files",
1310 required_features->not_context_indexed_files);
1311
1312 if (required_features->sparse_files &&
1313 !supported_features->sparse_files)
1314 WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files",
1315 required_features->sparse_files);
1316
1317 if (required_features->encrypted_directories &&
1318 !supported_features->encrypted_directories)
1319 WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories",
1320 required_features->encrypted_directories);
1321 }
1322
1323 /* Hard links. */
1324 if (required_features->hard_links && !supported_features->hard_links)
1325 WARNING("Extracting %lu hard links as independent files",
1326 required_features->hard_links);
1327
1328 /* Symbolic links and reparse points. */
1329 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) &&
1330 required_features->symlink_reparse_points &&
1331 !supported_features->symlink_reparse_points &&
1332 !supported_features->reparse_points)
1333 {
1334 ERROR("Extraction backend does not support symbolic links!");
1335 return WIMLIB_ERR_UNSUPPORTED;
1336 }
1337 if (required_features->reparse_points &&
1338 !supported_features->reparse_points)
1339 {
1340 if (supported_features->symlink_reparse_points) {
1341 if (required_features->other_reparse_points) {
1342 WARNING("Ignoring reparse data of %lu non-symlink/junction files",
1343 required_features->other_reparse_points);
1344 }
1345 } else {
1346 WARNING("Ignoring reparse data of %lu files",
1347 required_features->reparse_points);
1348 }
1349 }
1350
1351 /* Security descriptors. */
1352 if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS |
1353 WIMLIB_EXTRACT_FLAG_UNIX_DATA))
1354 == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) &&
1355 required_features->security_descriptors &&
1356 !supported_features->security_descriptors)
1357 {
1358 ERROR("Extraction backend does not support security descriptors!");
1359 return WIMLIB_ERR_UNSUPPORTED;
1360 }
1361 if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) &&
1362 required_features->security_descriptors &&
1363 !supported_features->security_descriptors)
1364 WARNING("Ignoring Windows NT security descriptors of %lu files",
1365 required_features->security_descriptors);
1366
1367 /* Standard UNIX metadata */
1368 if (required_features->unix_data &&
1369 (!supported_features->unix_data ||
1370 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA)))
1371 {
1372 if (extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) {
1373 ERROR("Requested UNIX metadata extraction, but "
1374 "extraction backend does not support it!");
1375 return WIMLIB_ERR_UNSUPPORTED;
1376 }
1377 WARNING("Ignoring UNIX metadata (uid/gid/mode/rdev) of %lu files%"TS,
1378 required_features->unix_data,
1379 (supported_features->unix_data ?
1380 T("\n (use --unix-data mode to extract these)") : T("")));
1381 }
1382
1383 /* Extended attributes */
1384 if (required_features->xattrs &&
1385 (!supported_features->xattrs ||
1386 (supported_features->unix_data &&
1387 !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))))
1388 {
1389 WARNING("Ignoring extended attributes of %lu files%"TS,
1390 required_features->xattrs,
1391 (supported_features->xattrs ?
1392 T("\n (use --unix-data mode to extract these)") : T("")));
1393 }
1394
1395 /* Object IDs. */
1396 if (required_features->object_ids && !supported_features->object_ids) {
1397 WARNING("Ignoring object IDs of %lu files",
1398 required_features->object_ids);
1399 }
1400
1401 /* DOS Names. */
1402 if (required_features->short_names &&
1403 !supported_features->short_names)
1404 {
1405 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) {
1406 ERROR("Extraction backend does not support DOS names!");
1407 return WIMLIB_ERR_UNSUPPORTED;
1408 }
1409 WARNING("Ignoring DOS names of %lu files",
1410 required_features->short_names);
1411 }
1412
1413 /* Timestamps. */
1414 if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) &&
1415 !supported_features->timestamps)
1416 {
1417 ERROR("Extraction backend does not support timestamps!");
1418 return WIMLIB_ERR_UNSUPPORTED;
1419 }
1420
1421 return 0;
1422 }
1423
1424 static const struct apply_operations *
select_apply_operations(int extract_flags)1425 select_apply_operations(int extract_flags)
1426 {
1427 #ifdef WITH_NTFS_3G
1428 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)
1429 return &ntfs_3g_apply_ops;
1430 #endif
1431 #ifdef __WIN32__
1432 return &win32_apply_ops;
1433 #else
1434 return &unix_apply_ops;
1435 #endif
1436 }
1437
1438 static int
extract_trees(WIMStruct * wim,struct wim_dentry ** trees,size_t num_trees,const tchar * target,int extract_flags)1439 extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees,
1440 const tchar *target, int extract_flags)
1441 {
1442 const struct apply_operations *ops;
1443 struct apply_ctx *ctx;
1444 int ret;
1445 LIST_HEAD(dentry_list);
1446
1447 if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) {
1448 ret = extract_dentries_to_stdout(trees, num_trees,
1449 wim->blob_table);
1450 goto out;
1451 }
1452
1453 num_trees = remove_duplicate_trees(trees, num_trees);
1454 num_trees = remove_contained_trees(trees, num_trees);
1455
1456 ops = select_apply_operations(extract_flags);
1457
1458 if (num_trees > 1 && ops->single_tree_only) {
1459 ERROR("Extracting multiple directory trees "
1460 "at once is not supported in %s extraction mode!",
1461 ops->name);
1462 ret = WIMLIB_ERR_UNSUPPORTED;
1463 goto out;
1464 }
1465
1466 ctx = CALLOC(1, ops->context_size);
1467 if (!ctx) {
1468 ret = WIMLIB_ERR_NOMEM;
1469 goto out;
1470 }
1471
1472 ctx->wim = wim;
1473 ctx->target = target;
1474 ctx->target_nchars = tstrlen(target);
1475 ctx->extract_flags = extract_flags;
1476 if (ctx->wim->progfunc) {
1477 ctx->progfunc = ctx->wim->progfunc;
1478 ctx->progctx = ctx->wim->progctx;
1479 ctx->progress.extract.image = wim->current_image;
1480 ctx->progress.extract.extract_flags = (extract_flags &
1481 WIMLIB_EXTRACT_MASK_PUBLIC);
1482 ctx->progress.extract.wimfile_name = wim->filename;
1483 ctx->progress.extract.image_name = wimlib_get_image_name(wim,
1484 wim->current_image);
1485 ctx->progress.extract.target = target;
1486 }
1487 INIT_LIST_HEAD(&ctx->blob_list);
1488 filedes_invalidate(&ctx->tmpfile_fd);
1489 ctx->apply_ops = ops;
1490
1491 ret = (*ops->get_supported_features)(target, &ctx->supported_features);
1492 if (ret)
1493 goto out_cleanup;
1494
1495 build_dentry_list(&dentry_list, trees, num_trees,
1496 !(extract_flags &
1497 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE));
1498
1499 dentry_list_get_features(&dentry_list, &ctx->required_features);
1500
1501 ret = do_feature_check(&ctx->required_features, &ctx->supported_features,
1502 ctx->extract_flags);
1503 if (ret)
1504 goto out_cleanup;
1505
1506 ret = dentry_list_calculate_extraction_names(&dentry_list, ctx);
1507 if (ret)
1508 goto out_cleanup;
1509
1510 if (unlikely(list_empty(&dentry_list))) {
1511 WARNING("There is nothing to extract!");
1512 goto out_cleanup;
1513 }
1514
1515 ret = dentry_list_resolve_streams(&dentry_list, ctx);
1516 if (ret)
1517 goto out_cleanup;
1518
1519 dentry_list_build_inode_alias_lists(&dentry_list);
1520
1521 ret = dentry_list_ref_streams(&dentry_list, ctx);
1522 if (ret)
1523 goto out_cleanup;
1524
1525 if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
1526 /* When extracting from a pipe, the number of bytes of data to
1527 * extract can't be determined in the normal way (examining the
1528 * blob table), since at this point all we have is a set of
1529 * SHA-1 message digests of blobs that need to be extracted.
1530 * However, we can get a reasonably accurate estimate by taking
1531 * <TOTALBYTES> from the corresponding <IMAGE> in the WIM XML
1532 * data. This does assume that a full image is being extracted,
1533 * but currently there is no API for doing otherwise. (Also,
1534 * subtract <HARDLINKBYTES> from this if hard links are
1535 * supported by the extraction mode.) */
1536 ctx->progress.extract.total_bytes =
1537 xml_get_image_total_bytes(wim->xml_info,
1538 wim->current_image);
1539 if (ctx->supported_features.hard_links) {
1540 ctx->progress.extract.total_bytes -=
1541 xml_get_image_hard_link_bytes(wim->xml_info,
1542 wim->current_image);
1543 }
1544 }
1545
1546 ret = extract_progress(ctx,
1547 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1548 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN :
1549 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN));
1550 if (ret)
1551 goto out_cleanup;
1552
1553 ret = (*ops->extract)(&dentry_list, ctx);
1554 if (ret)
1555 goto out_cleanup;
1556
1557 if (ctx->progress.extract.completed_bytes <
1558 ctx->progress.extract.total_bytes)
1559 {
1560 ctx->progress.extract.completed_bytes =
1561 ctx->progress.extract.total_bytes;
1562 ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
1563 if (ret)
1564 goto out_cleanup;
1565 }
1566
1567 ret = extract_progress(ctx,
1568 ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ?
1569 WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END :
1570 WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END));
1571 out_cleanup:
1572 destroy_blob_list(&ctx->blob_list);
1573 destroy_dentry_list(&dentry_list);
1574 FREE(ctx);
1575 out:
1576 return ret;
1577 }
1578
1579 static int
mkdir_if_needed(const tchar * target)1580 mkdir_if_needed(const tchar *target)
1581 {
1582 if (!tmkdir(target, 0755))
1583 return 0;
1584
1585 if (errno == EEXIST)
1586 return 0;
1587
1588 #ifdef __WIN32__
1589 /* _wmkdir() fails with EACCES if called on a drive root directory. */
1590 if (errno == EACCES)
1591 return 0;
1592 #endif
1593
1594 ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target);
1595 return WIMLIB_ERR_MKDIR;
1596 }
1597
1598 /* Make sure the extraction flags make sense, and update them if needed. */
1599 static int
check_extract_flags(const WIMStruct * wim,int * extract_flags_p)1600 check_extract_flags(const WIMStruct *wim, int *extract_flags_p)
1601 {
1602 int extract_flags = *extract_flags_p;
1603
1604 /* Check for invalid flag combinations */
1605
1606 if ((extract_flags &
1607 (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1608 WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS |
1609 WIMLIB_EXTRACT_FLAG_STRICT_ACLS))
1610 return WIMLIB_ERR_INVALID_PARAM;
1611
1612 if ((extract_flags &
1613 (WIMLIB_EXTRACT_FLAG_RPFIX |
1614 WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX |
1615 WIMLIB_EXTRACT_FLAG_NORPFIX))
1616 return WIMLIB_ERR_INVALID_PARAM;
1617
1618 #ifndef WITH_NTFS_3G
1619 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1620 ERROR("wimlib was compiled without support for NTFS-3G, so\n"
1621 " it cannot apply a WIM image directly to an NTFS volume.");
1622 return WIMLIB_ERR_UNSUPPORTED;
1623 }
1624 #endif
1625
1626 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1627 #ifdef __WIN32__
1628 if (!wim->filename)
1629 return WIMLIB_ERR_NO_FILENAME;
1630 #else
1631 ERROR("WIMBoot extraction is only supported on Windows!");
1632 return WIMLIB_ERR_UNSUPPORTED;
1633 #endif
1634 }
1635
1636 if (extract_flags & (WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K |
1637 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K |
1638 WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K |
1639 WIMLIB_EXTRACT_FLAG_COMPACT_LZX))
1640 {
1641 #ifdef __WIN32__
1642 int count = 0;
1643 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K) != 0);
1644 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K) != 0);
1645 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K) != 0);
1646 count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_LZX) != 0);
1647 if (count != 1) {
1648 ERROR("Only one compression format can be specified "
1649 "for compact-mode extraction!");
1650 return WIMLIB_ERR_INVALID_PARAM;
1651 }
1652 if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) {
1653 ERROR("Compact-mode extraction and WIMBoot-mode "
1654 "extraction are mutually exclusive!");
1655 return WIMLIB_ERR_INVALID_PARAM;
1656 }
1657 #else
1658 ERROR("Compact-mode extraction (System Compression) "
1659 "is only supported on Windows!");
1660 return WIMLIB_ERR_UNSUPPORTED;
1661 #endif
1662 }
1663
1664
1665 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX |
1666 WIMLIB_EXTRACT_FLAG_NORPFIX |
1667 WIMLIB_EXTRACT_FLAG_IMAGEMODE)) ==
1668 WIMLIB_EXTRACT_FLAG_IMAGEMODE)
1669 {
1670 /* For full-image extraction, do reparse point fixups by default
1671 * if the WIM header says they are enabled. */
1672 if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX)
1673 extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX;
1674 }
1675
1676 *extract_flags_p = extract_flags;
1677 return 0;
1678 }
1679
1680 struct append_dentry_ctx {
1681 struct wim_dentry **dentries;
1682 size_t num_dentries;
1683 size_t num_alloc_dentries;
1684 };
1685
1686 static int
append_dentry_cb(struct wim_dentry * dentry,void * _ctx)1687 append_dentry_cb(struct wim_dentry *dentry, void *_ctx)
1688 {
1689 struct append_dentry_ctx *ctx = _ctx;
1690
1691 if (ctx->num_dentries == ctx->num_alloc_dentries) {
1692 struct wim_dentry **new_dentries;
1693 size_t new_length;
1694
1695 new_length = max(ctx->num_alloc_dentries + 8,
1696 ctx->num_alloc_dentries * 3 / 2);
1697 new_dentries = REALLOC(ctx->dentries,
1698 new_length * sizeof(ctx->dentries[0]));
1699 if (new_dentries == NULL)
1700 return WIMLIB_ERR_NOMEM;
1701 ctx->dentries = new_dentries;
1702 ctx->num_alloc_dentries = new_length;
1703 }
1704 ctx->dentries[ctx->num_dentries++] = dentry;
1705 return 0;
1706 }
1707
1708 /* Append dentries matched by a path which can contain wildcard characters. */
1709 static int
append_matched_dentries(WIMStruct * wim,const tchar * orig_pattern,int extract_flags,struct append_dentry_ctx * ctx)1710 append_matched_dentries(WIMStruct *wim, const tchar *orig_pattern,
1711 int extract_flags, struct append_dentry_ctx *ctx)
1712 {
1713 const size_t count_before = ctx->num_dentries;
1714 tchar *pattern;
1715 int ret;
1716
1717 pattern = canonicalize_wim_path(orig_pattern);
1718 if (!pattern)
1719 return WIMLIB_ERR_NOMEM;
1720 ret = expand_path_pattern(wim_get_current_root_dentry(wim), pattern,
1721 append_dentry_cb, ctx);
1722 FREE(pattern);
1723 if (ret || ctx->num_dentries > count_before)
1724 return ret;
1725 if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB) {
1726 ERROR("No matches for path pattern \"%"TS"\"", orig_pattern);
1727 return WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1728 }
1729 WARNING("No matches for path pattern \"%"TS"\"", orig_pattern);
1730 return 0;
1731 }
1732
1733 static int
do_wimlib_extract_paths(WIMStruct * wim,int image,const tchar * target,const tchar * const * paths,size_t num_paths,int extract_flags)1734 do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1735 const tchar * const *paths, size_t num_paths,
1736 int extract_flags)
1737 {
1738 int ret;
1739 struct wim_dentry **trees;
1740 size_t num_trees;
1741
1742 if (wim == NULL || target == NULL || target[0] == T('\0') ||
1743 (num_paths != 0 && paths == NULL))
1744 return WIMLIB_ERR_INVALID_PARAM;
1745
1746 ret = check_extract_flags(wim, &extract_flags);
1747 if (ret)
1748 return ret;
1749
1750 ret = select_wim_image(wim, image);
1751 if (ret)
1752 return ret;
1753
1754 ret = wim_checksum_unhashed_blobs(wim);
1755 if (ret)
1756 return ret;
1757
1758 if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS |
1759 WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) ==
1760 (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE))
1761 {
1762 ret = mkdir_if_needed(target);
1763 if (ret)
1764 return ret;
1765 }
1766
1767 if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) {
1768
1769 struct append_dentry_ctx append_dentry_ctx = {
1770 .dentries = NULL,
1771 .num_dentries = 0,
1772 .num_alloc_dentries = 0,
1773 };
1774
1775 for (size_t i = 0; i < num_paths; i++) {
1776 ret = append_matched_dentries(wim, paths[i],
1777 extract_flags,
1778 &append_dentry_ctx);
1779 if (ret) {
1780 trees = append_dentry_ctx.dentries;
1781 goto out_free_trees;
1782 }
1783 }
1784 trees = append_dentry_ctx.dentries;
1785 num_trees = append_dentry_ctx.num_dentries;
1786 } else {
1787 trees = MALLOC(num_paths * sizeof(trees[0]));
1788 if (trees == NULL)
1789 return WIMLIB_ERR_NOMEM;
1790
1791 for (size_t i = 0; i < num_paths; i++) {
1792
1793 tchar *path = canonicalize_wim_path(paths[i]);
1794 if (path == NULL) {
1795 ret = WIMLIB_ERR_NOMEM;
1796 goto out_free_trees;
1797 }
1798
1799 trees[i] = get_dentry(wim, path,
1800 WIMLIB_CASE_PLATFORM_DEFAULT);
1801 FREE(path);
1802 if (trees[i] == NULL) {
1803 ERROR("Path \"%"TS"\" does not exist "
1804 "in WIM image %d",
1805 paths[i], wim->current_image);
1806 ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST;
1807 goto out_free_trees;
1808 }
1809 }
1810 num_trees = num_paths;
1811 }
1812
1813 if (num_trees == 0) {
1814 ret = 0;
1815 goto out_free_trees;
1816 }
1817
1818 ret = extract_trees(wim, trees, num_trees, target, extract_flags);
1819 out_free_trees:
1820 FREE(trees);
1821 return ret;
1822 }
1823
1824 static int
extract_single_image(WIMStruct * wim,int image,const tchar * target,int extract_flags)1825 extract_single_image(WIMStruct *wim, int image,
1826 const tchar *target, int extract_flags)
1827 {
1828 const tchar *path = WIMLIB_WIM_ROOT_PATH;
1829 extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE;
1830 return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags);
1831 }
1832
1833 static const tchar * const filename_forbidden_chars =
1834 #ifdef __WIN32__
1835 T("<>:\"/\\|?*");
1836 #else
1837 T("/");
1838 #endif
1839
1840 /* This function checks if it is okay to use a WIM image's name as a directory
1841 * name. */
1842 static bool
image_name_ok_as_dir(const tchar * image_name)1843 image_name_ok_as_dir(const tchar *image_name)
1844 {
1845 return image_name && *image_name &&
1846 !tstrpbrk(image_name, filename_forbidden_chars) &&
1847 tstrcmp(image_name, T(".")) &&
1848 tstrcmp(image_name, T("..")) &&
1849 tstrlen(image_name) <= 128;
1850 }
1851
1852 /* Extracts all images from the WIM to the directory @target, with the images
1853 * placed in subdirectories named by their image names. */
1854 static int
extract_all_images(WIMStruct * wim,const tchar * target,int extract_flags)1855 extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags)
1856 {
1857 size_t output_path_len = tstrlen(target);
1858 tchar buf[output_path_len + 1 + 128 + 1];
1859 int ret;
1860 int image;
1861 const tchar *image_name;
1862
1863 if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
1864 ERROR("Cannot extract multiple images in NTFS extraction mode.");
1865 return WIMLIB_ERR_INVALID_PARAM;
1866 }
1867
1868 ret = mkdir_if_needed(target);
1869 if (ret)
1870 return ret;
1871 tmemcpy(buf, target, output_path_len);
1872 buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR;
1873 for (image = 1; image <= wim->hdr.image_count; image++) {
1874 image_name = wimlib_get_image_name(wim, image);
1875 if (image_name_ok_as_dir(image_name)) {
1876 tstrcpy(buf + output_path_len + 1, image_name);
1877 } else {
1878 /* Image name is empty or contains forbidden characters.
1879 * Use image number instead. */
1880 tsprintf(buf + output_path_len + 1, T("%d"), image);
1881 }
1882 ret = extract_single_image(wim, image, buf, extract_flags);
1883 if (ret)
1884 return ret;
1885 }
1886 return 0;
1887 }
1888
1889 static int
do_wimlib_extract_image(WIMStruct * wim,int image,const tchar * target,int extract_flags)1890 do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
1891 int extract_flags)
1892 {
1893 if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE |
1894 WIMLIB_EXTRACT_FLAG_TO_STDOUT |
1895 WIMLIB_EXTRACT_FLAG_GLOB_PATHS))
1896 return WIMLIB_ERR_INVALID_PARAM;
1897
1898 if (image == WIMLIB_ALL_IMAGES)
1899 return extract_all_images(wim, target, extract_flags);
1900 else
1901 return extract_single_image(wim, image, target, extract_flags);
1902 }
1903
1904
1905 /****************************************************************************
1906 * Extraction API *
1907 ****************************************************************************/
1908
1909 WIMLIBAPI int
wimlib_extract_paths(WIMStruct * wim,int image,const tchar * target,const tchar * const * paths,size_t num_paths,int extract_flags)1910 wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target,
1911 const tchar * const *paths, size_t num_paths,
1912 int extract_flags)
1913 {
1914 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1915 return WIMLIB_ERR_INVALID_PARAM;
1916
1917 return do_wimlib_extract_paths(wim, image, target, paths, num_paths,
1918 extract_flags);
1919 }
1920
1921 WIMLIBAPI int
wimlib_extract_pathlist(WIMStruct * wim,int image,const tchar * target,const tchar * path_list_file,int extract_flags)1922 wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target,
1923 const tchar *path_list_file, int extract_flags)
1924 {
1925 int ret;
1926 tchar **paths;
1927 size_t num_paths;
1928 void *mem;
1929
1930 ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem);
1931 if (ret) {
1932 ERROR("Failed to read path list file \"%"TS"\"",
1933 path_list_file ? path_list_file : T("<stdin>"));
1934 return ret;
1935 }
1936
1937 ret = wimlib_extract_paths(wim, image, target,
1938 (const tchar * const *)paths, num_paths,
1939 extract_flags);
1940 FREE(paths);
1941 FREE(mem);
1942 return ret;
1943 }
1944
1945 WIMLIBAPI int
wimlib_extract_image_from_pipe_with_progress(int pipe_fd,const tchar * image_num_or_name,const tchar * target,int extract_flags,wimlib_progress_func_t progfunc,void * progctx)1946 wimlib_extract_image_from_pipe_with_progress(int pipe_fd,
1947 const tchar *image_num_or_name,
1948 const tchar *target,
1949 int extract_flags,
1950 wimlib_progress_func_t progfunc,
1951 void *progctx)
1952 {
1953 int ret;
1954 WIMStruct *pwm;
1955 struct filedes *in_fd;
1956 int image;
1957 unsigned i;
1958
1959 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
1960 return WIMLIB_ERR_INVALID_PARAM;
1961
1962 /* Read the WIM header from the pipe and get a WIMStruct to represent
1963 * the pipable WIM. Caveats: Unlike getting a WIMStruct with
1964 * wimlib_open_wim(), getting a WIMStruct in this way will result in an
1965 * empty blob table, no XML data read, and no filename set. */
1966 ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm,
1967 progfunc, progctx);
1968 if (ret)
1969 return ret;
1970
1971 /* Sanity check to make sure this is a pipable WIM. */
1972 if (pwm->hdr.magic != PWM_MAGIC) {
1973 ERROR("The WIM being read from file descriptor %d "
1974 "is not pipable!", pipe_fd);
1975 ret = WIMLIB_ERR_NOT_PIPABLE;
1976 goto out_wimlib_free;
1977 }
1978
1979 /* Sanity check to make sure the first part of a pipable split WIM is
1980 * sent over the pipe first. */
1981 if (pwm->hdr.part_number != 1) {
1982 ERROR("The first part of the split WIM must be "
1983 "sent over the pipe first.");
1984 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
1985 goto out_wimlib_free;
1986 }
1987
1988 in_fd = &pwm->in_fd;
1989 wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE);
1990
1991 /* As mentioned, the WIMStruct we created from the pipe does not have
1992 * XML data yet. Fix this by reading the extra copy of the XML data
1993 * that directly follows the header in pipable WIMs. (Note: see
1994 * write_pipable_wim() for more details about the format of pipable
1995 * WIMs.) */
1996 {
1997 u8 hash[SHA1_HASH_SIZE];
1998
1999 ret = read_pwm_blob_header(pwm, hash,
2000 &pwm->hdr.xml_data_reshdr, NULL);
2001 if (ret)
2002 goto out_wimlib_free;
2003
2004 if (!(pwm->hdr.xml_data_reshdr.flags & WIM_RESHDR_FLAG_METADATA)) {
2005 ERROR("Expected XML data, but found non-metadata resource.");
2006 ret = WIMLIB_ERR_INVALID_PIPABLE_WIM;
2007 goto out_wimlib_free;
2008 }
2009
2010 ret = read_wim_xml_data(pwm);
2011 if (ret)
2012 goto out_wimlib_free;
2013
2014 if (xml_get_image_count(pwm->xml_info) != pwm->hdr.image_count) {
2015 ERROR("Image count in XML data is not the same as in WIM header.");
2016 ret = WIMLIB_ERR_IMAGE_COUNT;
2017 goto out_wimlib_free;
2018 }
2019 }
2020
2021 /* Get image index (this may use the XML data that was just read to
2022 * resolve an image name). */
2023 if (image_num_or_name) {
2024 image = wimlib_resolve_image(pwm, image_num_or_name);
2025 if (image == WIMLIB_NO_IMAGE) {
2026 ERROR("\"%"TS"\" is not a valid image in the pipable WIM!",
2027 image_num_or_name);
2028 ret = WIMLIB_ERR_INVALID_IMAGE;
2029 goto out_wimlib_free;
2030 } else if (image == WIMLIB_ALL_IMAGES) {
2031 ERROR("Applying all images from a pipe is not supported!");
2032 ret = WIMLIB_ERR_INVALID_IMAGE;
2033 goto out_wimlib_free;
2034 }
2035 } else {
2036 if (pwm->hdr.image_count != 1) {
2037 ERROR("No image was specified, but the pipable WIM "
2038 "did not contain exactly 1 image");
2039 ret = WIMLIB_ERR_INVALID_IMAGE;
2040 goto out_wimlib_free;
2041 }
2042 image = 1;
2043 }
2044
2045 /* Load the needed metadata resource. */
2046 for (i = 1; i <= pwm->hdr.image_count; i++) {
2047 ret = handle_pwm_metadata_resource(pwm, i, i == image);
2048 if (ret)
2049 goto out_wimlib_free;
2050 }
2051 /* Extract the image. */
2052 extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;
2053 ret = do_wimlib_extract_image(pwm, image, target, extract_flags);
2054 /* Clean up and return. */
2055 out_wimlib_free:
2056 wimlib_free(pwm);
2057 return ret;
2058 }
2059
2060
2061 WIMLIBAPI int
wimlib_extract_image_from_pipe(int pipe_fd,const tchar * image_num_or_name,const tchar * target,int extract_flags)2062 wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
2063 const tchar *target, int extract_flags)
2064 {
2065 return wimlib_extract_image_from_pipe_with_progress(pipe_fd,
2066 image_num_or_name,
2067 target,
2068 extract_flags,
2069 NULL,
2070 NULL);
2071 }
2072
2073 WIMLIBAPI int
wimlib_extract_image(WIMStruct * wim,int image,const tchar * target,int extract_flags)2074 wimlib_extract_image(WIMStruct *wim, int image, const tchar *target,
2075 int extract_flags)
2076 {
2077 if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC)
2078 return WIMLIB_ERR_INVALID_PARAM;
2079 return do_wimlib_extract_image(wim, image, target, extract_flags);
2080 }
2081