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 an image (given by its dentry tree) to the image metadata array of a WIM
43 * file, adds an entry to the lookup table for the image metadata, updates the
44 * image count in the header, and selects the new image.
46 * Does not update the XML data.
48 * On failure, WIMLIB_ERR_NOMEM is returned and no changes are made. Otherwise,
49 * 0 is returned and the image metadata array of @w is modified.
51 * @w: The WIMStruct for the WIM file.
52 * @root_dentry: The root of the directory tree for the image.
53 * @sd: The security data for the image.
55 int add_new_dentry_tree(WIMStruct *w, struct dentry *root_dentry,
56 struct wim_security_data *sd)
58 struct lookup_table_entry *metadata_lte;
59 struct image_metadata *imd;
60 struct image_metadata *new_imd;
62 wimlib_assert(root_dentry != NULL);
64 DEBUG("Reallocating image metadata array for image_count = %u",
65 w->hdr.image_count + 1);
66 imd = CALLOC((w->hdr.image_count + 1), sizeof(struct image_metadata));
69 ERROR("Failed to allocate memory for new image metadata array");
73 memcpy(imd, w->image_metadata,
74 w->hdr.image_count * sizeof(struct image_metadata));
76 metadata_lte = new_lookup_table_entry();
80 metadata_lte->resource_entry.flags = WIM_RESHDR_FLAG_METADATA;
81 random_hash(metadata_lte->hash);
82 lookup_table_insert(w->lookup_table, metadata_lte);
84 new_imd = &imd[w->hdr.image_count];
86 new_imd->root_dentry = root_dentry;
87 new_imd->metadata_lte = metadata_lte;
88 new_imd->security_data = sd;
89 new_imd->modified = 1;
91 FREE(w->image_metadata);
92 w->image_metadata = imd;
98 return WIMLIB_ERR_NOMEM;
104 * Recursively builds a dentry tree from a directory tree on disk, outside the
107 * @root_ret: Place to return a pointer to the root of the dentry tree. Only
108 * modified if successful. NULL if the file or directory was
109 * excluded from capture.
111 * @root_disk_path: The path to the root of the directory tree on disk.
113 * @lookup_table: The lookup table for the WIM file. For each file added to the
114 * dentry tree being built, an entry is added to the lookup table,
115 * unless an identical stream is already in the lookup table.
116 * These lookup table entries that are added point to the path of
119 * @sd: Ignored. (Security data only captured in NTFS mode.)
122 * Configuration for files to be excluded from capture.
124 * @add_flags: Bitwise or of WIMLIB_ADD_IMAGE_FLAG_*
126 * @extra_arg: Ignored. (Only used in NTFS mode.)
128 * @return: 0 on success, nonzero on failure. It is a failure if any of
129 * the files cannot be `stat'ed, or if any of the needed
130 * directories cannot be opened or read. Failure to add the files
131 * to the WIM may still occur later when trying to actually read
132 * the on-disk files during a call to wimlib_write() or
133 * wimlib_overwrite().
135 static int build_dentry_tree(struct dentry **root_ret,
136 const char *root_disk_path,
137 struct lookup_table *lookup_table,
138 struct wim_security_data *sd,
139 const struct capture_config *config,
141 wimlib_progress_func_t progress_func,
144 struct stat root_stbuf;
146 int (*stat_fn)(const char *restrict, struct stat *restrict);
148 const char *filename;
151 if (exclude_path(root_disk_path, config, true)) {
152 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_ROOT) {
153 ERROR("Cannot exclude the root directory from capture");
154 return WIMLIB_ERR_INVALID_CAPTURE_CONFIG;
156 if ((add_image_flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE)
159 union wimlib_progress_info info;
160 info.scan.cur_path = root_disk_path;
161 info.scan.excluded = true;
162 progress_func(WIMLIB_PROGRESS_MSG_SCAN_DENTRY, &info);
168 if ((add_image_flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE)
171 union wimlib_progress_info info;
172 info.scan.cur_path = root_disk_path;
173 info.scan.excluded = false;
174 progress_func(WIMLIB_PROGRESS_MSG_SCAN_DENTRY, &info);
177 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_DEREFERENCE)
182 ret = (*stat_fn)(root_disk_path, &root_stbuf);
184 ERROR_WITH_ERRNO("Failed to stat `%s'", root_disk_path);
185 return WIMLIB_ERR_STAT;
188 if ((add_image_flags & WIMLIB_ADD_IMAGE_FLAG_ROOT) &&
189 !S_ISDIR(root_stbuf.st_mode))
191 /* Do a dereference-stat in case the root is a symbolic link.
192 * This case is allowed, provided that the symbolic link points
194 ret = stat(root_disk_path, &root_stbuf);
196 ERROR_WITH_ERRNO("Failed to stat `%s'", root_disk_path);
197 return WIMLIB_ERR_STAT;
199 if (!S_ISDIR(root_stbuf.st_mode)) {
200 ERROR("`%s' is not a directory", root_disk_path);
201 return WIMLIB_ERR_NOTDIR;
204 if (!S_ISREG(root_stbuf.st_mode) && !S_ISDIR(root_stbuf.st_mode)
205 && !S_ISLNK(root_stbuf.st_mode)) {
206 ERROR("`%s' is not a regular file, directory, or symbolic link.",
208 return WIMLIB_ERR_SPECIAL_FILE;
211 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_ROOT)
214 filename = path_basename(root_disk_path);
216 root = new_dentry_with_timeless_inode(filename);
219 return WIMLIB_ERR_INVALID_UTF8_STRING;
220 else if (errno == ENOMEM)
221 return WIMLIB_ERR_NOMEM;
223 return WIMLIB_ERR_ICONV_NOT_AVAILABLE;
226 inode = root->d_inode;
228 #ifdef HAVE_STAT_NANOSECOND_PRECISION
229 inode->creation_time = timespec_to_wim_timestamp(&root_stbuf.st_mtim);
230 inode->last_write_time = timespec_to_wim_timestamp(&root_stbuf.st_mtim);
231 inode->last_access_time = timespec_to_wim_timestamp(&root_stbuf.st_atim);
233 inode->creation_time = unix_timestamp_to_wim(root_stbuf.st_mtime);
234 inode->last_write_time = unix_timestamp_to_wim(root_stbuf.st_mtime);
235 inode->last_access_time = unix_timestamp_to_wim(root_stbuf.st_atime);
237 if (sizeof(ino_t) >= 8)
238 inode->ino = (u64)root_stbuf.st_ino;
240 inode->ino = (u64)root_stbuf.st_ino |
241 ((u64)root_stbuf.st_dev << ((sizeof(ino_t) * 8) & 63));
243 add_image_flags &= ~WIMLIB_ADD_IMAGE_FLAG_ROOT;
246 if (S_ISREG(root_stbuf.st_mode)) { /* Archiving a regular file */
248 struct lookup_table_entry *lte;
249 u8 hash[SHA1_HASH_SIZE];
251 inode->attributes = FILE_ATTRIBUTE_NORMAL;
253 /* Empty files do not have to have a lookup table entry. */
254 if (root_stbuf.st_size == 0)
257 /* For each regular file, we must check to see if the file is in
258 * the lookup table already; if it is, we increment its refcnt;
259 * otherwise, we create a new lookup table entry and insert it.
262 ret = sha1sum(root_disk_path, hash);
266 lte = __lookup_resource(lookup_table, hash);
269 DEBUG("Add lte reference %u for `%s'", lte->refcnt,
272 char *file_on_disk = STRDUP(root_disk_path);
274 ERROR("Failed to allocate memory for file path");
275 ret = WIMLIB_ERR_NOMEM;
278 lte = new_lookup_table_entry();
281 ret = WIMLIB_ERR_NOMEM;
284 lte->file_on_disk = file_on_disk;
285 lte->resource_location = RESOURCE_IN_FILE_ON_DISK;
286 lte->resource_entry.original_size = root_stbuf.st_size;
287 lte->resource_entry.size = root_stbuf.st_size;
288 copy_hash(lte->hash, hash);
289 lookup_table_insert(lookup_table, lte);
291 root->d_inode->lte = lte;
292 } else if (S_ISDIR(root_stbuf.st_mode)) { /* Archiving a directory */
294 inode->attributes = FILE_ATTRIBUTE_DIRECTORY;
297 struct dirent entry, *result;
298 struct dentry *child;
300 dir = opendir(root_disk_path);
302 ERROR_WITH_ERRNO("Failed to open the directory `%s'",
304 ret = WIMLIB_ERR_OPEN;
308 /* Buffer for names of files in directory. */
309 size_t len = strlen(root_disk_path);
310 char name[len + 1 + FILENAME_MAX + 1];
311 memcpy(name, root_disk_path, len);
314 /* Create a dentry for each entry in the directory on disk, and recurse
315 * to any subdirectories. */
318 ret = readdir_r(dir, &entry, &result);
320 ret = WIMLIB_ERR_READ;
321 ERROR_WITH_ERRNO("Error reading the "
328 if (result->d_name[0] == '.' && (result->d_name[1] == '\0'
329 || (result->d_name[1] == '.' && result->d_name[2] == '\0')))
331 strcpy(name + len + 1, result->d_name);
332 ret = build_dentry_tree(&child, name, lookup_table,
333 NULL, config, add_image_flags,
334 progress_func, NULL);
338 dentry_add_child(root, child);
341 } else { /* Archiving a symbolic link */
342 inode->attributes = FILE_ATTRIBUTE_REPARSE_POINT;
343 inode->reparse_tag = WIM_IO_REPARSE_TAG_SYMLINK;
345 /* The idea here is to call readlink() to get the UNIX target of
346 * the symbolic link, then turn the target into a reparse point
347 * data buffer that contains a relative or absolute symbolic
348 * link (NOT a junction point or *full* path symbolic link with
352 char deref_name_buf[4096];
353 ssize_t deref_name_len;
355 deref_name_len = readlink(root_disk_path, deref_name_buf,
356 sizeof(deref_name_buf) - 1);
357 if (deref_name_len >= 0) {
358 deref_name_buf[deref_name_len] = '\0';
359 DEBUG("Read symlink `%s'", deref_name_buf);
360 ret = inode_set_symlink(root->d_inode, deref_name_buf,
364 * Unfortunately, Windows seems to have the
365 * concept of "file" symbolic links as being
366 * different from "directory" symbolic links...
367 * so FILE_ATTRIBUTE_DIRECTORY needs to be set
368 * on the symbolic link if the *target* of the
369 * symbolic link is a directory.
372 if (stat(root_disk_path, &stbuf) == 0 &&
373 S_ISDIR(stbuf.st_mode))
375 inode->attributes |= FILE_ATTRIBUTE_DIRECTORY;
379 ERROR_WITH_ERRNO("Failed to read target of "
380 "symbolic link `%s'", root_disk_path);
381 ret = WIMLIB_ERR_READLINK;
388 free_dentry_tree(root, lookup_table);
397 COMPRESSION_EXCLUSION_LIST,
401 /* Default capture configuration file when none is specified. */
402 static const char *default_config =
407 "\\System Volume Information\n"
411 "[CompressionExclusionList]\n"
415 "\\WINDOWS\\inf\\*.pnf\n";
417 static void destroy_pattern_list(struct pattern_list *list)
422 static void destroy_capture_config(struct capture_config *config)
424 destroy_pattern_list(&config->exclusion_list);
425 destroy_pattern_list(&config->exclusion_exception);
426 destroy_pattern_list(&config->compression_exclusion_list);
427 destroy_pattern_list(&config->alignment_list);
428 FREE(config->config_str);
429 FREE(config->prefix);
430 memset(config, 0, sizeof(*config));
433 static int pattern_list_add_pattern(struct pattern_list *list,
437 if (list->num_pats >= list->num_allocated_pats) {
438 pats = REALLOC(list->pats,
439 sizeof(list->pats[0]) * (list->num_allocated_pats + 8));
441 return WIMLIB_ERR_NOMEM;
442 list->num_allocated_pats += 8;
445 list->pats[list->num_pats++] = pattern;
449 /* Parses the contents of the image capture configuration file and fills in a
450 * `struct capture_config'. */
451 static int init_capture_config(const char *_config_str, size_t config_len,
452 const char *_prefix, struct capture_config *config)
459 size_t bytes_remaining;
460 enum pattern_type type = NONE;
462 unsigned long line_no = 0;
464 DEBUG("config_len = %zu", config_len);
465 bytes_remaining = config_len;
466 memset(config, 0, sizeof(*config));
467 config_str = MALLOC(config_len);
469 ERROR("Could not duplicate capture config string");
470 return WIMLIB_ERR_NOMEM;
472 prefix = STRDUP(_prefix);
475 return WIMLIB_ERR_NOMEM;
478 memcpy(config_str, _config_str, config_len);
480 config->config_str = config_str;
481 config->prefix = prefix;
482 config->prefix_len = strlen(prefix);
483 while (bytes_remaining) {
486 eol = memchr(p, '\n', bytes_remaining);
488 ERROR("Expected end-of-line in capture config file on "
489 "line %lu", line_no);
490 ret = WIMLIB_ERR_INVALID_CAPTURE_CONFIG;
495 bytes_remaining -= (next_p - p);
499 if (*(eol - 1) == '\r')
503 /* Translate backslash to forward slash */
504 for (char *pp = p; pp != eol; pp++)
508 /* Remove drive letter */
509 if (eol - p > 2 && isalpha(*p) && *(p + 1) == ':')
513 if (strcmp(p, "[ExclusionList]") == 0)
514 type = EXCLUSION_LIST;
515 else if (strcmp(p, "[ExclusionException]") == 0)
516 type = EXCLUSION_EXCEPTION;
517 else if (strcmp(p, "[CompressionExclusionList]") == 0)
518 type = COMPRESSION_EXCLUSION_LIST;
519 else if (strcmp(p, "[AlignmentList]") == 0)
520 type = ALIGNMENT_LIST;
521 else if (p[0] == '[' && strrchr(p, ']')) {
522 ERROR("Unknown capture configuration section `%s'", p);
523 ret = WIMLIB_ERR_INVALID_CAPTURE_CONFIG;
524 } else switch (type) {
526 DEBUG("Adding pattern \"%s\" to exclusion list", p);
527 ret = pattern_list_add_pattern(&config->exclusion_list, p);
529 case EXCLUSION_EXCEPTION:
530 DEBUG("Adding pattern \"%s\" to exclusion exception list", p);
531 ret = pattern_list_add_pattern(&config->exclusion_exception, p);
533 case COMPRESSION_EXCLUSION_LIST:
534 DEBUG("Adding pattern \"%s\" to compression exclusion list", p);
535 ret = pattern_list_add_pattern(&config->compression_exclusion_list, p);
538 DEBUG("Adding pattern \"%s\" to alignment list", p);
539 ret = pattern_list_add_pattern(&config->alignment_list, p);
542 ERROR("Line %lu of capture configuration is not "
543 "in a block (such as [ExclusionList])",
545 ret = WIMLIB_ERR_INVALID_CAPTURE_CONFIG;
553 destroy_capture_config(config);
557 static bool match_pattern(const char *path, const char *path_basename,
558 const struct pattern_list *list)
560 for (size_t i = 0; i < list->num_pats; i++) {
561 const char *pat = list->pats[i];
564 /* Absolute path from root of capture */
567 if (strchr(pat, '/'))
568 /* Relative path from root of capture */
571 /* A file name pattern */
572 string = path_basename;
574 if (fnmatch(pat, string, FNM_PATHNAME
580 DEBUG("`%s' matches the pattern \"%s\"",
588 static void print_pattern_list(const struct pattern_list *list)
590 for (size_t i = 0; i < list->num_pats; i++)
591 printf(" %s\n", list->pats[i]);
594 static void print_capture_config(const struct capture_config *config)
596 if (config->exclusion_list.num_pats) {
597 puts("Files or folders excluded from image capture:");
598 print_pattern_list(&config->exclusion_list);
603 /* Return true if the image capture configuration file indicates we should
604 * exclude the filename @path from capture.
606 * If @exclude_prefix is %true, the part of the path up and including the name
607 * of the directory being captured is not included in the path for matching
608 * purposes. This allows, for example, a pattern like /hiberfil.sys to match a
609 * file /mnt/windows7/hiberfil.sys if we are capturing the /mnt/windows7
612 bool exclude_path(const char *path, const struct capture_config *config,
615 const char *basename = path_basename(path);
616 if (exclude_prefix) {
617 wimlib_assert(strlen(path) >= config->prefix_len);
618 if (memcmp(config->prefix, path, config->prefix_len) == 0
619 && path[config->prefix_len] == '/')
620 path += config->prefix_len;
622 return match_pattern(path, basename, &config->exclusion_list) &&
623 !match_pattern(path, basename, &config->exclusion_exception);
627 WIMLIBAPI int wimlib_add_image(WIMStruct *w, const char *source,
628 const char *name, const char *config_str,
629 size_t config_len, int add_image_flags,
630 wimlib_progress_func_t progress_func)
632 int (*capture_tree)(struct dentry **, const char *,
633 struct lookup_table *,
634 struct wim_security_data *,
635 const struct capture_config *,
636 int, wimlib_progress_func_t, void *);
639 struct dentry *root_dentry = NULL;
640 struct wim_security_data *sd;
641 struct capture_config config;
642 struct image_metadata *imd;
645 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_NTFS) {
647 if (add_image_flags & (WIMLIB_ADD_IMAGE_FLAG_DEREFERENCE)) {
648 ERROR("Cannot dereference files when capturing directly from NTFS");
649 return WIMLIB_ERR_INVALID_PARAM;
651 capture_tree = build_dentry_tree_ntfs;
652 extra_arg = &w->ntfs_vol;
654 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
655 " cannot capture a WIM image directly from a NTFS volume!");
656 return WIMLIB_ERR_UNSUPPORTED;
659 capture_tree = build_dentry_tree;
663 DEBUG("Adding dentry tree from directory or NTFS volume `%s'.", source);
665 if (!name || !*name) {
666 ERROR("Must specify a non-empty string for the image name");
667 return WIMLIB_ERR_INVALID_PARAM;
669 if (!source || !*source) {
670 ERROR("Must specify the name of a directory or NTFS volume");
671 return WIMLIB_ERR_INVALID_PARAM;
674 if (w->hdr.total_parts != 1) {
675 ERROR("Cannot add an image to a split WIM");
676 return WIMLIB_ERR_SPLIT_UNSUPPORTED;
679 if (wimlib_image_name_in_use(w, name)) {
680 ERROR("There is already an image named \"%s\" in `%s'",
682 return WIMLIB_ERR_IMAGE_NAME_COLLISION;
685 DEBUG("Initializing capture configuration");
687 DEBUG("Using default capture configuration");
688 config_str = default_config;
689 config_len = strlen(default_config);
691 ret = init_capture_config(config_str, config_len, source, &config);
694 print_capture_config(&config);
696 DEBUG("Allocating security data");
698 sd = CALLOC(1, sizeof(struct wim_security_data));
700 ret = WIMLIB_ERR_NOMEM;
701 goto out_destroy_config;
703 sd->total_length = 8;
707 union wimlib_progress_info progress;
708 progress.scan.source = source;
709 progress_func(WIMLIB_PROGRESS_MSG_SCAN_BEGIN, &progress);
712 DEBUG("Building dentry tree.");
713 ret = (*capture_tree)(&root_dentry, source, w->lookup_table, sd,
714 &config, add_image_flags | WIMLIB_ADD_IMAGE_FLAG_ROOT,
715 progress_func, extra_arg);
716 destroy_capture_config(&config);
719 ERROR("Failed to build dentry tree for `%s'", source);
720 goto out_free_security_data;
724 union wimlib_progress_info progress;
725 progress.scan.source = source;
726 progress_func(WIMLIB_PROGRESS_MSG_SCAN_END, &progress);
729 DEBUG("Calculating full paths of dentries.");
730 ret = for_dentry_in_tree(root_dentry, calculate_dentry_full_path, NULL);
732 goto out_free_dentry_tree;
734 ret = add_new_dentry_tree(w, root_dentry, sd);
736 goto out_free_dentry_tree;
738 imd = &w->image_metadata[w->hdr.image_count - 1];
740 ret = dentry_tree_fix_inodes(root_dentry, &imd->inode_list);
742 goto out_destroy_imd;
744 DEBUG("Assigning hard link group IDs");
745 assign_inode_numbers(&imd->inode_list);
747 ret = xml_add_image(w, name);
749 goto out_destroy_imd;
751 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_BOOT)
752 wimlib_set_boot_idx(w, w->hdr.image_count);
755 destroy_image_metadata(&w->image_metadata[w->hdr.image_count - 1],
757 w->hdr.image_count--;
759 out_free_dentry_tree:
760 free_dentry_tree(root_dentry, w->lookup_table);
761 out_free_security_data:
762 free_security_data(sd);
764 destroy_capture_config(&config);