6 * Copyright (C) 2012 Eric Biggers
8 * This file is part of wimlib, a library for working with WIM files.
10 * wimlib is free software; you can redistribute it and/or modify it under the
11 * terms of the GNU General Public License as published by the Free
12 * Software Foundation; either version 3 of the License, or (at your option)
15 * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
16 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
17 * A PARTICULAR PURPOSE. See the GNU General Public License for more
20 * You should have received a copy of the GNU General Public License
21 * along with wimlib; if not, see http://www.gnu.org/licenses/.
24 #include "wimlib_internal.h"
26 #include "timestamp.h"
27 #include "lookup_table.h"
37 /** Private flag: Used to mark that we currently adding the root directory of
39 #define WIMLIB_ADD_IMAGE_FLAG_ROOT 0x80000000
42 * Adds the dentry tree and security data for a new image to the image metadata
43 * array of the WIMStruct.
45 int add_new_dentry_tree(WIMStruct *w, struct wim_dentry *root_dentry,
46 struct wim_security_data *sd)
48 struct wim_lookup_table_entry *metadata_lte;
49 struct wim_image_metadata *imd;
50 struct wim_image_metadata *new_imd;
52 wimlib_assert(root_dentry != NULL);
54 DEBUG("Reallocating image metadata array for image_count = %u",
55 w->hdr.image_count + 1);
56 imd = CALLOC((w->hdr.image_count + 1), sizeof(struct wim_image_metadata));
59 ERROR("Failed to allocate memory for new image metadata array");
63 memcpy(imd, w->image_metadata,
64 w->hdr.image_count * sizeof(struct wim_image_metadata));
66 metadata_lte = new_lookup_table_entry();
70 metadata_lte->resource_entry.flags = WIM_RESHDR_FLAG_METADATA;
71 random_hash(metadata_lte->hash);
72 lookup_table_insert(w->lookup_table, metadata_lte);
74 new_imd = &imd[w->hdr.image_count];
76 new_imd->root_dentry = root_dentry;
77 new_imd->metadata_lte = metadata_lte;
78 new_imd->security_data = sd;
79 new_imd->modified = 1;
81 FREE(w->image_metadata);
82 w->image_metadata = imd;
88 return WIMLIB_ERR_NOMEM;
94 * build_dentry_tree():
95 * Recursively builds a tree of WIM dentries from an on-disk directory
98 * @root_ret: Place to return a pointer to the root of the dentry tree. Only
99 * modified if successful. Set to NULL if the file or directory was
100 * excluded from capture.
102 * @root_disk_path: The path to the root of the directory tree on disk.
104 * @lookup_table: The lookup table for the WIM file. For each file added to the
105 * dentry tree being built, an entry is added to the lookup table,
106 * unless an identical stream is already in the lookup table.
107 * These lookup table entries that are added point to the path of
110 * @sd: Ignored. (Security data only captured in NTFS mode.)
113 * Configuration for files to be excluded from capture.
115 * @add_flags: Bitwise or of WIMLIB_ADD_IMAGE_FLAG_*
117 * @extra_arg: Ignored. (Only used in NTFS mode.)
119 * @return: 0 on success, nonzero on failure. It is a failure if any of
120 * the files cannot be `stat'ed, or if any of the needed
121 * directories cannot be opened or read. Failure to add the files
122 * to the WIM may still occur later when trying to actually read
123 * the on-disk files during a call to wimlib_write() or
124 * wimlib_overwrite().
126 static int build_dentry_tree(struct wim_dentry **root_ret,
127 const char *root_disk_path,
128 struct wim_lookup_table *lookup_table,
129 struct wim_security_data *sd,
130 const struct capture_config *config,
132 wimlib_progress_func_t progress_func,
135 struct stat root_stbuf;
137 int (*stat_fn)(const char *restrict, struct stat *restrict);
138 struct wim_dentry *root;
139 const char *filename;
140 struct wim_inode *inode;
142 if (exclude_path(root_disk_path, config, true)) {
143 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_ROOT) {
144 ERROR("Cannot exclude the root directory from capture");
145 return WIMLIB_ERR_INVALID_CAPTURE_CONFIG;
147 if ((add_image_flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE)
150 union wimlib_progress_info info;
151 info.scan.cur_path = root_disk_path;
152 info.scan.excluded = true;
153 progress_func(WIMLIB_PROGRESS_MSG_SCAN_DENTRY, &info);
159 if ((add_image_flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE)
162 union wimlib_progress_info info;
163 info.scan.cur_path = root_disk_path;
164 info.scan.excluded = false;
165 progress_func(WIMLIB_PROGRESS_MSG_SCAN_DENTRY, &info);
168 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_DEREFERENCE)
173 ret = (*stat_fn)(root_disk_path, &root_stbuf);
175 ERROR_WITH_ERRNO("Failed to stat `%s'", root_disk_path);
176 return WIMLIB_ERR_STAT;
179 if ((add_image_flags & WIMLIB_ADD_IMAGE_FLAG_ROOT) &&
180 !S_ISDIR(root_stbuf.st_mode))
182 /* Do a dereference-stat in case the root is a symbolic link.
183 * This case is allowed, provided that the symbolic link points
185 ret = stat(root_disk_path, &root_stbuf);
187 ERROR_WITH_ERRNO("Failed to stat `%s'", root_disk_path);
188 return WIMLIB_ERR_STAT;
190 if (!S_ISDIR(root_stbuf.st_mode)) {
191 ERROR("`%s' is not a directory", root_disk_path);
192 return WIMLIB_ERR_NOTDIR;
195 if (!S_ISREG(root_stbuf.st_mode) && !S_ISDIR(root_stbuf.st_mode)
196 && !S_ISLNK(root_stbuf.st_mode)) {
197 ERROR("`%s' is not a regular file, directory, or symbolic link.",
199 return WIMLIB_ERR_SPECIAL_FILE;
202 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_ROOT)
205 filename = path_basename(root_disk_path);
207 root = new_dentry_with_timeless_inode(filename);
210 return WIMLIB_ERR_INVALID_UTF8_STRING;
211 else if (errno == ENOMEM)
212 return WIMLIB_ERR_NOMEM;
214 return WIMLIB_ERR_ICONV_NOT_AVAILABLE;
217 inode = root->d_inode;
219 #ifdef HAVE_STAT_NANOSECOND_PRECISION
220 inode->i_creation_time = timespec_to_wim_timestamp(&root_stbuf.st_mtim);
221 inode->i_last_write_time = timespec_to_wim_timestamp(&root_stbuf.st_mtim);
222 inode->i_last_access_time = timespec_to_wim_timestamp(&root_stbuf.st_atim);
224 inode->i_creation_time = unix_timestamp_to_wim(root_stbuf.st_mtime);
225 inode->i_last_write_time = unix_timestamp_to_wim(root_stbuf.st_mtime);
226 inode->i_last_access_time = unix_timestamp_to_wim(root_stbuf.st_atime);
228 if (sizeof(ino_t) >= 8)
229 inode->i_ino = (u64)root_stbuf.st_ino;
231 inode->i_ino = (u64)root_stbuf.st_ino |
232 ((u64)root_stbuf.st_dev << ((sizeof(ino_t) * 8) & 63));
234 add_image_flags &= ~WIMLIB_ADD_IMAGE_FLAG_ROOT;
235 inode->i_resolved = 1;
237 if (S_ISREG(root_stbuf.st_mode)) { /* Archiving a regular file */
239 struct wim_lookup_table_entry *lte;
240 u8 hash[SHA1_HASH_SIZE];
242 inode->i_attributes = FILE_ATTRIBUTE_NORMAL;
244 /* Empty files do not have to have a lookup table entry. */
245 if (root_stbuf.st_size == 0)
248 /* For each regular file, we must check to see if the file is in
249 * the lookup table already; if it is, we increment its refcnt;
250 * otherwise, we create a new lookup table entry and insert it.
253 ret = sha1sum(root_disk_path, hash);
257 lte = __lookup_resource(lookup_table, hash);
260 DEBUG("Add lte reference %u for `%s'", lte->refcnt,
263 char *file_on_disk = STRDUP(root_disk_path);
265 ERROR("Failed to allocate memory for file path");
266 ret = WIMLIB_ERR_NOMEM;
269 lte = new_lookup_table_entry();
272 ret = WIMLIB_ERR_NOMEM;
275 lte->file_on_disk = file_on_disk;
276 lte->resource_location = RESOURCE_IN_FILE_ON_DISK;
277 lte->resource_entry.original_size = root_stbuf.st_size;
278 lte->resource_entry.size = root_stbuf.st_size;
279 copy_hash(lte->hash, hash);
280 lookup_table_insert(lookup_table, lte);
282 root->d_inode->i_lte = lte;
283 } else if (S_ISDIR(root_stbuf.st_mode)) { /* Archiving a directory */
285 inode->i_attributes = FILE_ATTRIBUTE_DIRECTORY;
288 struct dirent entry, *result;
289 struct wim_dentry *child;
291 dir = opendir(root_disk_path);
293 ERROR_WITH_ERRNO("Failed to open the directory `%s'",
295 ret = WIMLIB_ERR_OPEN;
299 /* Buffer for names of files in directory. */
300 size_t len = strlen(root_disk_path);
301 char name[len + 1 + FILENAME_MAX + 1];
302 memcpy(name, root_disk_path, len);
305 /* Create a dentry for each entry in the directory on disk, and recurse
306 * to any subdirectories. */
309 ret = readdir_r(dir, &entry, &result);
311 ret = WIMLIB_ERR_READ;
312 ERROR_WITH_ERRNO("Error reading the "
319 if (result->d_name[0] == '.' && (result->d_name[1] == '\0'
320 || (result->d_name[1] == '.' && result->d_name[2] == '\0')))
322 strcpy(name + len + 1, result->d_name);
323 ret = build_dentry_tree(&child, name, lookup_table,
324 NULL, config, add_image_flags,
325 progress_func, NULL);
329 dentry_add_child(root, child);
332 } else { /* Archiving a symbolic link */
333 inode->i_attributes = FILE_ATTRIBUTE_REPARSE_POINT;
334 inode->i_reparse_tag = WIM_IO_REPARSE_TAG_SYMLINK;
336 /* The idea here is to call readlink() to get the UNIX target of
337 * the symbolic link, then turn the target into a reparse point
338 * data buffer that contains a relative or absolute symbolic
339 * link (NOT a junction point or *full* path symbolic link with
343 char deref_name_buf[4096];
344 ssize_t deref_name_len;
346 deref_name_len = readlink(root_disk_path, deref_name_buf,
347 sizeof(deref_name_buf) - 1);
348 if (deref_name_len >= 0) {
349 deref_name_buf[deref_name_len] = '\0';
350 DEBUG("Read symlink `%s'", deref_name_buf);
351 ret = inode_set_symlink(root->d_inode, deref_name_buf,
355 * Unfortunately, Windows seems to have the
356 * concept of "file" symbolic links as being
357 * different from "directory" symbolic links...
358 * so FILE_ATTRIBUTE_DIRECTORY needs to be set
359 * on the symbolic link if the *target* of the
360 * symbolic link is a directory.
363 if (stat(root_disk_path, &stbuf) == 0 &&
364 S_ISDIR(stbuf.st_mode))
366 inode->i_attributes |= FILE_ATTRIBUTE_DIRECTORY;
370 ERROR_WITH_ERRNO("Failed to read target of "
371 "symbolic link `%s'", root_disk_path);
372 ret = WIMLIB_ERR_READLINK;
379 free_dentry_tree(root, lookup_table);
388 COMPRESSION_EXCLUSION_LIST,
392 /* Default capture configuration file when none is specified. */
393 static const char *default_config =
398 "\\System Volume Information\n"
402 "[CompressionExclusionList]\n"
406 "\\WINDOWS\\inf\\*.pnf\n";
408 static void destroy_pattern_list(struct pattern_list *list)
413 static void destroy_capture_config(struct capture_config *config)
415 destroy_pattern_list(&config->exclusion_list);
416 destroy_pattern_list(&config->exclusion_exception);
417 destroy_pattern_list(&config->compression_exclusion_list);
418 destroy_pattern_list(&config->alignment_list);
419 FREE(config->config_str);
420 FREE(config->prefix);
421 memset(config, 0, sizeof(*config));
424 static int pattern_list_add_pattern(struct pattern_list *list,
428 if (list->num_pats >= list->num_allocated_pats) {
429 pats = REALLOC(list->pats,
430 sizeof(list->pats[0]) * (list->num_allocated_pats + 8));
432 return WIMLIB_ERR_NOMEM;
433 list->num_allocated_pats += 8;
436 list->pats[list->num_pats++] = pattern;
440 /* Parses the contents of the image capture configuration file and fills in a
441 * `struct capture_config'. */
442 static int init_capture_config(const char *_config_str, size_t config_len,
443 const char *_prefix, struct capture_config *config)
450 size_t bytes_remaining;
451 enum pattern_type type = NONE;
453 unsigned long line_no = 0;
455 DEBUG("config_len = %zu", config_len);
456 bytes_remaining = config_len;
457 memset(config, 0, sizeof(*config));
458 config_str = MALLOC(config_len);
460 ERROR("Could not duplicate capture config string");
461 return WIMLIB_ERR_NOMEM;
463 prefix = STRDUP(_prefix);
466 return WIMLIB_ERR_NOMEM;
469 memcpy(config_str, _config_str, config_len);
471 config->config_str = config_str;
472 config->prefix = prefix;
473 config->prefix_len = strlen(prefix);
474 while (bytes_remaining) {
477 eol = memchr(p, '\n', bytes_remaining);
479 ERROR("Expected end-of-line in capture config file on "
480 "line %lu", line_no);
481 ret = WIMLIB_ERR_INVALID_CAPTURE_CONFIG;
486 bytes_remaining -= (next_p - p);
490 if (*(eol - 1) == '\r')
494 /* Translate backslash to forward slash */
495 for (char *pp = p; pp != eol; pp++)
499 /* Remove drive letter */
500 if (eol - p > 2 && isalpha(*p) && *(p + 1) == ':')
504 if (strcmp(p, "[ExclusionList]") == 0)
505 type = EXCLUSION_LIST;
506 else if (strcmp(p, "[ExclusionException]") == 0)
507 type = EXCLUSION_EXCEPTION;
508 else if (strcmp(p, "[CompressionExclusionList]") == 0)
509 type = COMPRESSION_EXCLUSION_LIST;
510 else if (strcmp(p, "[AlignmentList]") == 0)
511 type = ALIGNMENT_LIST;
512 else if (p[0] == '[' && strrchr(p, ']')) {
513 ERROR("Unknown capture configuration section `%s'", p);
514 ret = WIMLIB_ERR_INVALID_CAPTURE_CONFIG;
515 } else switch (type) {
517 DEBUG("Adding pattern \"%s\" to exclusion list", p);
518 ret = pattern_list_add_pattern(&config->exclusion_list, p);
520 case EXCLUSION_EXCEPTION:
521 DEBUG("Adding pattern \"%s\" to exclusion exception list", p);
522 ret = pattern_list_add_pattern(&config->exclusion_exception, p);
524 case COMPRESSION_EXCLUSION_LIST:
525 DEBUG("Adding pattern \"%s\" to compression exclusion list", p);
526 ret = pattern_list_add_pattern(&config->compression_exclusion_list, p);
529 DEBUG("Adding pattern \"%s\" to alignment list", p);
530 ret = pattern_list_add_pattern(&config->alignment_list, p);
533 ERROR("Line %lu of capture configuration is not "
534 "in a block (such as [ExclusionList])",
536 ret = WIMLIB_ERR_INVALID_CAPTURE_CONFIG;
544 destroy_capture_config(config);
548 static bool match_pattern(const char *path, const char *path_basename,
549 const struct pattern_list *list)
551 for (size_t i = 0; i < list->num_pats; i++) {
552 const char *pat = list->pats[i];
555 /* Absolute path from root of capture */
558 if (strchr(pat, '/'))
559 /* Relative path from root of capture */
562 /* A file name pattern */
563 string = path_basename;
565 if (fnmatch(pat, string, FNM_PATHNAME
571 DEBUG("`%s' matches the pattern \"%s\"",
579 static void print_pattern_list(const struct pattern_list *list)
581 for (size_t i = 0; i < list->num_pats; i++)
582 printf(" %s\n", list->pats[i]);
585 static void print_capture_config(const struct capture_config *config)
587 if (config->exclusion_list.num_pats) {
588 puts("Files or folders excluded from image capture:");
589 print_pattern_list(&config->exclusion_list);
594 /* Return true if the image capture configuration file indicates we should
595 * exclude the filename @path from capture.
597 * If @exclude_prefix is %true, the part of the path up and including the name
598 * of the directory being captured is not included in the path for matching
599 * purposes. This allows, for example, a pattern like /hiberfil.sys to match a
600 * file /mnt/windows7/hiberfil.sys if we are capturing the /mnt/windows7
603 bool exclude_path(const char *path, const struct capture_config *config,
606 const char *basename = path_basename(path);
607 if (exclude_prefix) {
608 wimlib_assert(strlen(path) >= config->prefix_len);
609 if (memcmp(config->prefix, path, config->prefix_len) == 0
610 && path[config->prefix_len] == '/')
611 path += config->prefix_len;
613 return match_pattern(path, basename, &config->exclusion_list) &&
614 !match_pattern(path, basename, &config->exclusion_exception);
618 WIMLIBAPI int wimlib_add_image(WIMStruct *w, const char *source,
619 const char *name, const char *config_str,
620 size_t config_len, int add_image_flags,
621 wimlib_progress_func_t progress_func)
623 int (*capture_tree)(struct wim_dentry **, const char *,
624 struct wim_lookup_table *,
625 struct wim_security_data *,
626 const struct capture_config *,
627 int, wimlib_progress_func_t, void *);
630 struct wim_dentry *root_dentry = NULL;
631 struct wim_security_data *sd;
632 struct capture_config config;
633 struct wim_image_metadata *imd;
636 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_NTFS) {
638 if (add_image_flags & (WIMLIB_ADD_IMAGE_FLAG_DEREFERENCE)) {
639 ERROR("Cannot dereference files when capturing directly from NTFS");
640 return WIMLIB_ERR_INVALID_PARAM;
642 capture_tree = build_dentry_tree_ntfs;
643 extra_arg = &w->ntfs_vol;
645 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
646 " cannot capture a WIM image directly from a NTFS volume!");
647 return WIMLIB_ERR_UNSUPPORTED;
650 capture_tree = build_dentry_tree;
654 DEBUG("Adding dentry tree from directory or NTFS volume `%s'.", source);
656 if (!name || !*name) {
657 ERROR("Must specify a non-empty string for the image name");
658 return WIMLIB_ERR_INVALID_PARAM;
660 if (!source || !*source) {
661 ERROR("Must specify the name of a directory or NTFS volume");
662 return WIMLIB_ERR_INVALID_PARAM;
665 if (w->hdr.total_parts != 1) {
666 ERROR("Cannot add an image to a split WIM");
667 return WIMLIB_ERR_SPLIT_UNSUPPORTED;
670 if (wimlib_image_name_in_use(w, name)) {
671 ERROR("There is already an image named \"%s\" in `%s'",
673 return WIMLIB_ERR_IMAGE_NAME_COLLISION;
676 DEBUG("Initializing capture configuration");
678 DEBUG("Using default capture configuration");
679 config_str = default_config;
680 config_len = strlen(default_config);
682 ret = init_capture_config(config_str, config_len, source, &config);
685 print_capture_config(&config);
687 DEBUG("Allocating security data");
689 sd = CALLOC(1, sizeof(struct wim_security_data));
691 ret = WIMLIB_ERR_NOMEM;
692 goto out_destroy_config;
694 sd->total_length = 8;
698 union wimlib_progress_info progress;
699 progress.scan.source = source;
700 progress_func(WIMLIB_PROGRESS_MSG_SCAN_BEGIN, &progress);
703 DEBUG("Building dentry tree.");
704 ret = (*capture_tree)(&root_dentry, source, w->lookup_table, sd,
705 &config, add_image_flags | WIMLIB_ADD_IMAGE_FLAG_ROOT,
706 progress_func, extra_arg);
707 destroy_capture_config(&config);
710 ERROR("Failed to build dentry tree for `%s'", source);
711 goto out_free_security_data;
715 union wimlib_progress_info progress;
716 progress.scan.source = source;
717 progress_func(WIMLIB_PROGRESS_MSG_SCAN_END, &progress);
720 DEBUG("Calculating full paths of dentries.");
721 ret = for_dentry_in_tree(root_dentry, calculate_dentry_full_path, NULL);
723 goto out_free_dentry_tree;
725 ret = add_new_dentry_tree(w, root_dentry, sd);
727 goto out_free_dentry_tree;
729 imd = &w->image_metadata[w->hdr.image_count - 1];
731 ret = dentry_tree_fix_inodes(root_dentry, &imd->inode_list);
733 goto out_destroy_imd;
735 DEBUG("Assigning hard link group IDs");
736 assign_inode_numbers(&imd->inode_list);
738 ret = xml_add_image(w, name);
740 goto out_destroy_imd;
742 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_BOOT)
743 wimlib_set_boot_idx(w, w->hdr.image_count);
746 destroy_image_metadata(&w->image_metadata[w->hdr.image_count - 1],
748 w->hdr.image_count--;
750 out_free_dentry_tree:
751 free_dentry_tree(root_dentry, w->lookup_table);
752 out_free_security_data:
753 free_security_data(sd);
755 destroy_capture_config(&config);