6 * Copyright (C) 2012, 2013 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/.
26 #if defined(__CYGWIN__) || defined(__WIN32__)
33 # include "security.h"
36 # include <sys/stat.h>
37 # include "timestamp.h"
40 #include "wimlib_internal.h"
42 #include "lookup_table.h"
55 #define WIMLIB_ADD_IMAGE_FLAG_ROOT 0x80000000
56 #define WIMLIB_ADD_IMAGE_FLAG_SOURCE 0x40000000
59 * Adds the dentry tree and security data for a new image to the image metadata
60 * array of the WIMStruct.
62 int add_new_dentry_tree(WIMStruct *w, struct wim_dentry *root_dentry,
63 struct wim_security_data *sd)
65 struct wim_lookup_table_entry *metadata_lte;
66 struct wim_image_metadata *imd;
67 struct wim_image_metadata *new_imd;
69 wimlib_assert(root_dentry != NULL);
71 DEBUG("Reallocating image metadata array for image_count = %u",
72 w->hdr.image_count + 1);
73 imd = CALLOC((w->hdr.image_count + 1), sizeof(struct wim_image_metadata));
76 ERROR("Failed to allocate memory for new image metadata array");
80 memcpy(imd, w->image_metadata,
81 w->hdr.image_count * sizeof(struct wim_image_metadata));
83 metadata_lte = new_lookup_table_entry();
87 metadata_lte->resource_entry.flags = WIM_RESHDR_FLAG_METADATA;
88 random_hash(metadata_lte->hash);
89 lookup_table_insert(w->lookup_table, metadata_lte);
91 new_imd = &imd[w->hdr.image_count];
93 new_imd->root_dentry = root_dentry;
94 new_imd->metadata_lte = metadata_lte;
95 new_imd->security_data = sd;
96 new_imd->modified = 1;
98 FREE(w->image_metadata);
99 w->image_metadata = imd;
100 w->hdr.image_count++;
105 return WIMLIB_ERR_NOMEM;
109 #if defined(__CYGWIN__) || defined(__WIN32__)
111 static u64 FILETIME_to_u64(const FILETIME *ft)
113 return ((u64)ft->dwHighDateTime << 32) | (u64)ft->dwLowDateTime;
117 static int build_dentry_tree(struct wim_dentry **root_ret,
118 const char *root_disk_path,
119 struct wim_lookup_table *lookup_table,
120 struct wim_security_data *sd,
121 const struct capture_config *config,
123 wimlib_progress_func_t progress_func,
126 static int win32_get_short_name(struct wim_dentry *dentry,
127 const wchar_t *path_utf16)
129 WIN32_FIND_DATAW dat;
130 if (FindFirstFileW(path_utf16, &dat) &&
131 dat.cAlternateFileName[0] != L'\0')
133 size_t short_name_len = wcslen(dat.cAlternateFileName) * 2;
134 size_t n = short_name_len + sizeof(wchar_t);
135 dentry->short_name = MALLOC(n);
136 if (!dentry->short_name)
137 return WIMLIB_ERR_NOMEM;
138 memcpy(dentry->short_name, dat.cAlternateFileName, n);
139 dentry->short_name_len = short_name_len;
144 static int win32_get_security_descriptor(struct wim_dentry *dentry,
145 struct sd_set *sd_set,
146 const wchar_t *path_utf16)
148 SECURITY_INFORMATION requestedInformation;
153 requestedInformation = DACL_SECURITY_INFORMATION |
154 SACL_SECURITY_INFORMATION |
155 OWNER_SECURITY_INFORMATION |
156 GROUP_SECURITY_INFORMATION;
157 /* Request length of security descriptor */
158 status = GetFileSecurityW(path_utf16, requestedInformation,
159 NULL, 0, &lenNeeded);
160 err = GetLastError();
161 if (!status && err == ERROR_INSUFFICIENT_BUFFER) {
162 DWORD len = lenNeeded;
164 if (GetFileSecurityW(path_utf16, requestedInformation,
165 buf, len, &lenNeeded))
167 int security_id = sd_set_add_sd(sd_set, buf, len);
169 return WIMLIB_ERR_NOMEM;
171 dentry->d_inode->i_security_id = security_id;
175 err = GetLastError();
178 ERROR("Win32 API: Failed to read security descriptor of \"%ls\"",
181 return WIMLIB_ERR_READ;
184 /* Reads the directory entries of directory using a Win32 API and recursively
185 * calls build_dentry_tree() on them. */
186 static int win32_recurse_directory(struct wim_dentry *root,
187 const char *root_disk_path,
188 struct wim_lookup_table *lookup_table,
189 struct wim_security_data *sd,
190 const struct capture_config *config,
192 wimlib_progress_func_t progress_func,
193 struct sd_set *sd_set,
194 const wchar_t *path_utf16,
195 size_t path_utf16_nchars)
197 WIN32_FIND_DATAW dat;
203 /* Begin reading the directory by calling FindFirstFileW.
204 * Unlike UNIX opendir(), FindFirstFileW has file globbing built
205 * into it. But this isn't what we actually want, so just add a
206 * dummy glob to get all entries. */
207 wchar_t pattern_buf[path_utf16_nchars + 3];
208 memcpy(pattern_buf, path_utf16,
209 path_utf16_nchars * sizeof(wchar_t));
210 pattern_buf[path_utf16_nchars] = L'/';
211 pattern_buf[path_utf16_nchars + 1] = L'*';
212 pattern_buf[path_utf16_nchars + 2] = L'\0';
213 hFind = FindFirstFileW(pattern_buf, &dat);
215 if (hFind == INVALID_HANDLE_VALUE) {
216 err = GetLastError();
217 if (err == ERROR_FILE_NOT_FOUND) {
220 ERROR("Win32 API: Failed to read directory \"%s\"",
223 return WIMLIB_ERR_READ;
228 /* Skip . and .. entries */
229 if (!(dat.cFileName[0] == L'.' &&
230 (dat.cFileName[1] == L'\0' ||
231 (dat.cFileName[1] == L'.' && dat.cFileName[2] == L'\0'))))
233 struct wim_dentry *child;
236 size_t utf8_name_nbytes;
237 ret = utf16_to_utf8((const char*)dat.cFileName,
238 wcslen(dat.cFileName) * sizeof(wchar_t),
244 char name[strlen(root_disk_path) + 1 + utf8_name_nbytes + 1];
245 sprintf(name, "%s/%s", root_disk_path, utf8_name);
247 ret = build_dentry_tree(&child, name, lookup_table,
248 sd, config, add_image_flags,
249 progress_func, sd_set);
253 dentry_add_child(root, child);
255 } while (FindNextFileW(hFind, &dat));
256 err = GetLastError();
257 if (err != ERROR_NO_MORE_FILES) {
258 ERROR("Win32 API: Failed to read directory \"%s\"", root_disk_path);
261 ret = WIMLIB_ERR_READ;
268 /* Load a reparse point into a WIM inode. It is just stored in memory.
270 * @hFile: Open handle to a reparse point, with permission to read the reparse
273 * @inode: WIM inode for the reparse point.
275 * @lookup_table: Stream lookup table for the WIM; an entry will be added to it
276 * for the reparse point unless an entry already exists for
277 * the exact same data stream.
279 * @path: External path to the parse point (UTF-8). Used for error messages
282 * Returns 0 on success; nonzero on failure. */
283 static int win32_capture_reparse_point(HANDLE hFile,
284 struct wim_inode *inode,
285 struct wim_lookup_table *lookup_table,
288 /* "Reparse point data, including the tag and optional GUID,
289 * cannot exceed 16 kilobytes." - MSDN */
290 char reparse_point_buf[16 * 1024];
293 if (!DeviceIoControl(hFile, FSCTL_GET_REPARSE_POINT,
294 NULL, 0, reparse_point_buf,
295 sizeof(reparse_point_buf), &bytesReturned, NULL))
297 DWORD err = GetLastError();
298 ERROR("Win32 API: Failed to get reparse data of \"%s\"", path);
300 return WIMLIB_ERR_READ;
302 if (bytesReturned < 8) {
303 ERROR("Reparse data on \"%s\" is invalid", path);
304 return WIMLIB_ERR_READ;
306 inode->i_reparse_tag = *(u32*)reparse_point_buf;
307 return inode_add_ads_with_data(inode, "",
308 (const u8*)reparse_point_buf + 8,
309 bytesReturned - 8, lookup_table);
312 /* Calculate the SHA1 message digest of a Win32 data stream, which may be either
313 * an unnamed or named data stream.
315 * @path: Path to the file, with the stream noted at the end for named
316 * streams. UTF-16LE encoding.
318 * @hash: On success, the SHA1 message digest of the stream is written to
321 * Returns 0 on success; nonzero on failure.
323 static int win32_sha1sum(const wchar_t *path, u8 hash[SHA1_HASH_SIZE])
331 hFile = win32_open_file_readonly(path);
332 if (hFile == INVALID_HANDLE_VALUE)
333 return WIMLIB_ERR_OPEN;
337 if (!ReadFile(hFile, buf, sizeof(buf), &bytesRead, NULL)) {
338 ret = WIMLIB_ERR_READ;
339 goto out_close_handle;
343 sha1_update(&ctx, buf, bytesRead);
346 sha1_final(hash, &ctx);
352 /* Scans an unnamed or named stream of a Win32 file (not a reparse point
353 * stream); calculates its SHA1 message digest and either creates a `struct
354 * wim_lookup_table_entry' in memory for it, or uses an existing 'struct
355 * wim_lookup_table_entry' for an identical stream.
357 * @path_utf16: Path to the file (UTF-16LE).
359 * @path_utf16_nchars: Number of 2-byte characters in @path_utf16.
361 * @inode: WIM inode to save the stream into.
363 * @lookup_table: Stream lookup table for the WIM.
365 * @dat: A `WIN32_FIND_STREAM_DATA' structure that specifies the
368 * Returns 0 on success; nonzero on failure.
370 static int win32_capture_stream(const wchar_t *path_utf16,
371 size_t path_utf16_nchars,
372 struct wim_inode *inode,
373 struct wim_lookup_table *lookup_table,
374 WIN32_FIND_STREAM_DATA *dat)
376 struct wim_ads_entry *ads_entry;
377 u8 hash[SHA1_HASH_SIZE];
378 struct wim_lookup_table_entry *lte;
381 bool is_named_stream;
386 /* The stream name should be returned as :NAME:TYPE */
387 p = dat->cStreamName;
389 goto out_invalid_stream_name;
391 colon = wcschr(p, L':');
393 goto out_invalid_stream_name;
395 if (wcscmp(colon + 1, L"$DATA")) {
396 /* Not a DATA stream */
401 is_named_stream = (p != colon);
402 if (is_named_stream) {
403 /* Allocate an ADS entry for the named stream. */
404 char *utf8_stream_name;
405 size_t utf8_stream_name_len;
406 ret = utf16_to_utf8((const char *)p,
407 (colon - p) * sizeof(wchar_t),
409 &utf8_stream_name_len);
412 ads_entry = inode_add_ads(inode, utf8_stream_name);
413 FREE(utf8_stream_name);
415 ret = WIMLIB_ERR_NOMEM;
420 /* Create a UTF-16 string @spath that gives the filename, then a colon,
421 * then the stream name. Or, if it's an unnamed stream, just the
422 * filename. It is MALLOC()'ed so that it can be saved in the
423 * wim_lookup_table_entry if needed. */
425 spath_nchars = path_utf16_nchars;
427 spath_nchars += colon - p + 1;
429 spath = MALLOC((spath_nchars + 1) * sizeof(wchar_t));
430 memcpy(spath, path_utf16, path_utf16_nchars * sizeof(wchar_t));
431 if (is_named_stream) {
432 spath[path_utf16_nchars] = L':';
433 memcpy(&spath[path_utf16_nchars + 1], p, (colon - p) * sizeof(wchar_t));
435 spath[spath_nchars] = L'\0';
437 ret = win32_sha1sum(spath, hash);
439 err = GetLastError();
440 ERROR("Win32 API: Failed to read \"%ls\" to calculate SHA1sum",
446 lte = __lookup_resource(lookup_table, hash);
448 /* Use existing wim_lookup_table_entry that has the same SHA1
452 /* Make a new wim_lookup_table_entry */
453 lte = new_lookup_table_entry();
455 ret = WIMLIB_ERR_NOMEM;
458 lte->file_on_disk = (char*)spath;
460 lte->resource_location = RESOURCE_WIN32;
461 lte->resource_entry.original_size = (uint64_t)dat->StreamSize.QuadPart;
462 lte->resource_entry.size = (uint64_t)dat->StreamSize.QuadPart;
463 copy_hash(lte->hash, hash);
464 lookup_table_insert(lookup_table, lte);
467 ads_entry->lte = lte;
474 out_invalid_stream_name:
475 ERROR("Invalid stream name: \"%ls:%ls\"", path_utf16, dat->cStreamName);
476 ret = WIMLIB_ERR_READ;
480 /* Scans a Win32 file for unnamed and named data streams (not reparse point
483 * @path_utf16: Path to the file (UTF-16LE).
485 * @path_utf16_nchars: Number of 2-byte characters in @path_utf16.
487 * @inode: WIM inode to save the stream into.
489 * @lookup_table: Stream lookup table for the WIM.
491 * Returns 0 on success; nonzero on failure.
493 static int win32_capture_streams(const wchar_t *path_utf16,
494 size_t path_utf16_nchars,
495 struct wim_inode *inode,
496 struct wim_lookup_table *lookup_table)
498 WIN32_FIND_STREAM_DATA dat;
503 hFind = FindFirstStreamW(path_utf16, FindStreamInfoStandard, &dat, 0);
504 if (hFind == INVALID_HANDLE_VALUE) {
505 err = GetLastError();
507 /* Seems legal for this to return ERROR_HANDLE_EOF on reparse
508 * points and directories */
509 if ((inode->i_attributes &
510 (FILE_ATTRIBUTE_REPARSE_POINT | FILE_ATTRIBUTE_DIRECTORY))
511 && err == ERROR_HANDLE_EOF)
515 ERROR("Win32 API: Failed to look up data streams of \"%ls\"",
518 return WIMLIB_ERR_READ;
522 ret = win32_capture_stream(path_utf16,
528 } while (FindNextStreamW(hFind, &dat));
529 err = GetLastError();
530 if (err != ERROR_HANDLE_EOF) {
531 ERROR("Win32 API: Error reading data streams from \"%ls\"", path_utf16);
533 ret = WIMLIB_ERR_READ;
542 * build_dentry_tree():
543 * Recursively builds a tree of WIM dentries from an on-disk directory
546 * @root_ret: Place to return a pointer to the root of the dentry tree. Only
547 * modified if successful. Set to NULL if the file or directory was
548 * excluded from capture.
550 * @root_disk_path: The path to the root of the directory tree on disk (UTF-8).
552 * @lookup_table: The lookup table for the WIM file. For each file added to the
553 * dentry tree being built, an entry is added to the lookup table,
554 * unless an identical stream is already in the lookup table.
555 * These lookup table entries that are added point to the path of
558 * @sd: Ignored. (Security data only captured in NTFS mode.)
561 * Configuration for files to be excluded from capture.
563 * @add_flags: Bitwise or of WIMLIB_ADD_IMAGE_FLAG_*
565 * @extra_arg: Ignored in UNIX builds; used to pass sd_set pointer in Windows
568 * @return: 0 on success, nonzero on failure. It is a failure if any of
569 * the files cannot be `stat'ed, or if any of the needed
570 * directories cannot be opened or read. Failure to add the files
571 * to the WIM may still occur later when trying to actually read
572 * the on-disk files during a call to wimlib_write() or
573 * wimlib_overwrite().
575 static int build_dentry_tree(struct wim_dentry **root_ret,
576 const char *root_disk_path,
577 struct wim_lookup_table *lookup_table,
578 struct wim_security_data *sd,
579 const struct capture_config *config,
581 wimlib_progress_func_t progress_func,
584 struct wim_dentry *root = NULL;
586 struct wim_inode *inode;
588 if (exclude_path(root_disk_path, config, true)) {
589 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_ROOT) {
590 ERROR("Cannot exclude the root directory from capture");
591 ret = WIMLIB_ERR_INVALID_CAPTURE_CONFIG;
594 if ((add_image_flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE)
597 union wimlib_progress_info info;
598 info.scan.cur_path = root_disk_path;
599 info.scan.excluded = true;
600 progress_func(WIMLIB_PROGRESS_MSG_SCAN_DENTRY, &info);
605 if ((add_image_flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE)
608 union wimlib_progress_info info;
609 info.scan.cur_path = root_disk_path;
610 info.scan.excluded = false;
611 progress_func(WIMLIB_PROGRESS_MSG_SCAN_DENTRY, &info);
614 #if !defined(__CYGWIN__) && !defined(__WIN32__)
615 /* UNIX version of capturing a directory tree */
616 struct stat root_stbuf;
617 int (*stat_fn)(const char *restrict, struct stat *restrict);
618 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_DEREFERENCE)
623 ret = (*stat_fn)(root_disk_path, &root_stbuf);
625 ERROR_WITH_ERRNO("Failed to stat `%s'", root_disk_path);
629 if ((add_image_flags & WIMLIB_ADD_IMAGE_FLAG_ROOT) &&
630 !S_ISDIR(root_stbuf.st_mode))
632 /* Do a dereference-stat in case the root is a symbolic link.
633 * This case is allowed, provided that the symbolic link points
635 ret = stat(root_disk_path, &root_stbuf);
637 ERROR_WITH_ERRNO("Failed to stat `%s'", root_disk_path);
638 ret = WIMLIB_ERR_STAT;
641 if (!S_ISDIR(root_stbuf.st_mode)) {
642 ERROR("`%s' is not a directory", root_disk_path);
643 ret = WIMLIB_ERR_NOTDIR;
647 if (!S_ISREG(root_stbuf.st_mode) && !S_ISDIR(root_stbuf.st_mode)
648 && !S_ISLNK(root_stbuf.st_mode)) {
649 ERROR("`%s' is not a regular file, directory, or symbolic link.",
651 ret = WIMLIB_ERR_SPECIAL_FILE;
655 root = new_dentry_with_timeless_inode(path_basename(root_disk_path));
658 ret = WIMLIB_ERR_INVALID_UTF8_STRING;
659 else if (errno == ENOMEM)
660 ret = WIMLIB_ERR_NOMEM;
662 ret = WIMLIB_ERR_ICONV_NOT_AVAILABLE;
666 inode = root->d_inode;
668 #ifdef HAVE_STAT_NANOSECOND_PRECISION
669 inode->i_creation_time = timespec_to_wim_timestamp(&root_stbuf.st_mtim);
670 inode->i_last_write_time = timespec_to_wim_timestamp(&root_stbuf.st_mtim);
671 inode->i_last_access_time = timespec_to_wim_timestamp(&root_stbuf.st_atim);
673 inode->i_creation_time = unix_timestamp_to_wim(root_stbuf.st_mtime);
674 inode->i_last_write_time = unix_timestamp_to_wim(root_stbuf.st_mtime);
675 inode->i_last_access_time = unix_timestamp_to_wim(root_stbuf.st_atime);
677 if (sizeof(ino_t) >= 8)
678 inode->i_ino = (u64)root_stbuf.st_ino;
680 inode->i_ino = (u64)root_stbuf.st_ino |
681 ((u64)root_stbuf.st_dev << ((sizeof(ino_t) * 8) & 63));
682 inode->i_resolved = 1;
683 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_UNIX_DATA) {
684 ret = inode_set_unix_data(inode, root_stbuf.st_uid,
688 UNIX_DATA_ALL | UNIX_DATA_CREATE);
692 add_image_flags &= ~(WIMLIB_ADD_IMAGE_FLAG_ROOT | WIMLIB_ADD_IMAGE_FLAG_SOURCE);
693 if (S_ISREG(root_stbuf.st_mode)) { /* Archiving a regular file */
695 struct wim_lookup_table_entry *lte;
696 u8 hash[SHA1_HASH_SIZE];
698 inode->i_attributes = FILE_ATTRIBUTE_NORMAL;
700 /* Empty files do not have to have a lookup table entry. */
701 if (root_stbuf.st_size == 0)
704 /* For each regular file, we must check to see if the file is in
705 * the lookup table already; if it is, we increment its refcnt;
706 * otherwise, we create a new lookup table entry and insert it.
709 ret = sha1sum(root_disk_path, hash);
713 lte = __lookup_resource(lookup_table, hash);
716 DEBUG("Add lte reference %u for `%s'", lte->refcnt,
719 char *file_on_disk = STRDUP(root_disk_path);
721 ERROR("Failed to allocate memory for file path");
722 ret = WIMLIB_ERR_NOMEM;
725 lte = new_lookup_table_entry();
728 ret = WIMLIB_ERR_NOMEM;
731 lte->file_on_disk = file_on_disk;
732 lte->resource_location = RESOURCE_IN_FILE_ON_DISK;
733 lte->resource_entry.original_size = root_stbuf.st_size;
734 lte->resource_entry.size = root_stbuf.st_size;
735 copy_hash(lte->hash, hash);
736 lookup_table_insert(lookup_table, lte);
738 root->d_inode->i_lte = lte;
739 } else if (S_ISDIR(root_stbuf.st_mode)) { /* Archiving a directory */
741 inode->i_attributes = FILE_ATTRIBUTE_DIRECTORY;
744 struct dirent entry, *result;
745 struct wim_dentry *child;
747 dir = opendir(root_disk_path);
749 ERROR_WITH_ERRNO("Failed to open the directory `%s'",
751 ret = WIMLIB_ERR_OPEN;
755 /* Buffer for names of files in directory. */
756 size_t len = strlen(root_disk_path);
757 char name[len + 1 + FILENAME_MAX + 1];
758 memcpy(name, root_disk_path, len);
761 /* Create a dentry for each entry in the directory on disk, and recurse
762 * to any subdirectories. */
765 ret = readdir_r(dir, &entry, &result);
767 ret = WIMLIB_ERR_READ;
768 ERROR_WITH_ERRNO("Error reading the "
775 if (result->d_name[0] == '.' && (result->d_name[1] == '\0'
776 || (result->d_name[1] == '.' && result->d_name[2] == '\0')))
778 strcpy(name + len + 1, result->d_name);
779 ret = build_dentry_tree(&child, name, lookup_table,
780 NULL, config, add_image_flags,
781 progress_func, NULL);
785 dentry_add_child(root, child);
788 } else { /* Archiving a symbolic link */
789 inode->i_attributes = FILE_ATTRIBUTE_REPARSE_POINT;
790 inode->i_reparse_tag = WIM_IO_REPARSE_TAG_SYMLINK;
792 /* The idea here is to call readlink() to get the UNIX target of
793 * the symbolic link, then turn the target into a reparse point
794 * data buffer that contains a relative or absolute symbolic
795 * link (NOT a junction point or *full* path symbolic link with
799 char deref_name_buf[4096];
800 ssize_t deref_name_len;
802 deref_name_len = readlink(root_disk_path, deref_name_buf,
803 sizeof(deref_name_buf) - 1);
804 if (deref_name_len >= 0) {
805 deref_name_buf[deref_name_len] = '\0';
806 DEBUG("Read symlink `%s'", deref_name_buf);
807 ret = inode_set_symlink(root->d_inode, deref_name_buf,
811 * Unfortunately, Windows seems to have the
812 * concept of "file" symbolic links as being
813 * different from "directory" symbolic links...
814 * so FILE_ATTRIBUTE_DIRECTORY needs to be set
815 * on the symbolic link if the *target* of the
816 * symbolic link is a directory.
819 if (stat(root_disk_path, &stbuf) == 0 &&
820 S_ISDIR(stbuf.st_mode))
822 inode->i_attributes |= FILE_ATTRIBUTE_DIRECTORY;
826 ERROR_WITH_ERRNO("Failed to read target of "
827 "symbolic link `%s'", root_disk_path);
828 ret = WIMLIB_ERR_READLINK;
832 /* Win32 version of capturing a directory tree */
835 size_t path_utf16_nchars;
836 struct sd_set *sd_set;
839 if (extra_arg == NULL) {
840 sd_set = alloca(sizeof(struct sd_set));
841 sd_set->rb_root.rb_node = NULL,
847 ret = utf8_to_utf16(root_disk_path, strlen(root_disk_path),
848 (char**)&path_utf16, &path_utf16_nchars);
851 path_utf16_nchars /= sizeof(wchar_t);
853 HANDLE hFile = win32_open_file_readonly(path_utf16);
854 if (hFile == INVALID_HANDLE_VALUE) {
855 err = GetLastError();
856 ERROR("Win32 API: Failed to open \"%s\"", root_disk_path);
858 ret = WIMLIB_ERR_OPEN;
859 goto out_destroy_sd_set;
862 BY_HANDLE_FILE_INFORMATION file_info;
863 if (!GetFileInformationByHandle(hFile, &file_info)) {
864 err = GetLastError();
865 ERROR("Win32 API: Failed to get file information for \"%s\"",
868 ret = WIMLIB_ERR_STAT;
869 goto out_close_handle;
872 /* Create a WIM dentry */
873 root = new_dentry_with_timeless_inode(path_basename(root_disk_path));
876 ret = WIMLIB_ERR_INVALID_UTF8_STRING;
877 else if (errno == ENOMEM)
878 ret = WIMLIB_ERR_NOMEM;
880 ret = WIMLIB_ERR_ICONV_NOT_AVAILABLE;
881 goto out_close_handle;
884 /* Start preparing the associated WIM inode */
885 inode = root->d_inode;
887 inode->i_attributes = file_info.dwFileAttributes;
888 inode->i_creation_time = FILETIME_to_u64(&file_info.ftCreationTime);
889 inode->i_last_write_time = FILETIME_to_u64(&file_info.ftLastWriteTime);
890 inode->i_last_access_time = FILETIME_to_u64(&file_info.ftLastAccessTime);
891 inode->i_ino = ((u64)file_info.nFileIndexHigh << 32) |
892 (u64)file_info.nFileIndexLow;
894 inode->i_resolved = 1;
895 add_image_flags &= ~(WIMLIB_ADD_IMAGE_FLAG_ROOT | WIMLIB_ADD_IMAGE_FLAG_SOURCE);
897 /* Get DOS name and security descriptor (if any). */
898 ret = win32_get_short_name(root, path_utf16);
900 goto out_close_handle;
901 ret = win32_get_security_descriptor(root, sd_set, path_utf16);
903 goto out_close_handle;
905 if (inode_is_directory(inode)) {
906 /* Directory (not a reparse point) --- recurse to children */
908 /* But first... directories may have alternate data streams that
909 * need to be captured. */
910 ret = win32_capture_streams(path_utf16,
915 goto out_close_handle;
916 ret = win32_recurse_directory(root,
926 } else if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
927 /* Reparse point: save the reparse tag and data */
928 ret = win32_capture_reparse_point(hFile,
933 /* Not a directory, not a reparse point; capture the default
934 * file contents and any alternate data streams. */
935 ret = win32_capture_streams(path_utf16,
943 if (extra_arg == NULL)
944 destroy_sd_set(sd_set);
948 /* The below lines of code are common to both UNIX and Win32 builds. It
949 * simply returns the captured directory tree if the capture was
950 * successful, or frees it if the capture was unsuccessful. */
955 free_dentry_tree(root, lookup_table);
963 COMPRESSION_EXCLUSION_LIST,
967 #define COMPAT_DEFAULT_CONFIG
969 /* Default capture configuration file when none is specified. */
970 static const char *default_config =
971 #ifdef COMPAT_DEFAULT_CONFIG /* XXX: This policy is being moved to library
972 users. The next ABI-incompatible library
973 version will default to the empty string here. */
978 "\\System Volume Information\n"
982 "[CompressionExclusionList]\n"
986 "\\WINDOWS\\inf\\*.pnf\n";
991 static void destroy_pattern_list(struct pattern_list *list)
996 static void destroy_capture_config(struct capture_config *config)
998 destroy_pattern_list(&config->exclusion_list);
999 destroy_pattern_list(&config->exclusion_exception);
1000 destroy_pattern_list(&config->compression_exclusion_list);
1001 destroy_pattern_list(&config->alignment_list);
1002 FREE(config->config_str);
1003 FREE(config->prefix);
1004 memset(config, 0, sizeof(*config));
1007 static int pattern_list_add_pattern(struct pattern_list *list,
1008 const char *pattern)
1011 if (list->num_pats >= list->num_allocated_pats) {
1012 pats = REALLOC(list->pats,
1013 sizeof(list->pats[0]) * (list->num_allocated_pats + 8));
1015 return WIMLIB_ERR_NOMEM;
1016 list->num_allocated_pats += 8;
1019 list->pats[list->num_pats++] = pattern;
1023 /* Parses the contents of the image capture configuration file and fills in a
1024 * `struct capture_config'. */
1025 static int init_capture_config(struct capture_config *config,
1026 const char *_config_str, size_t config_len)
1032 size_t bytes_remaining;
1033 enum pattern_type type = NONE;
1035 unsigned long line_no = 0;
1037 DEBUG("config_len = %zu", config_len);
1038 bytes_remaining = config_len;
1039 memset(config, 0, sizeof(*config));
1040 config_str = MALLOC(config_len);
1042 ERROR("Could not duplicate capture config string");
1043 return WIMLIB_ERR_NOMEM;
1046 memcpy(config_str, _config_str, config_len);
1047 next_p = config_str;
1048 config->config_str = config_str;
1049 while (bytes_remaining) {
1052 eol = memchr(p, '\n', bytes_remaining);
1054 ERROR("Expected end-of-line in capture config file on "
1055 "line %lu", line_no);
1056 ret = WIMLIB_ERR_INVALID_CAPTURE_CONFIG;
1061 bytes_remaining -= (next_p - p);
1065 if (*(eol - 1) == '\r')
1069 /* Translate backslash to forward slash */
1070 for (char *pp = p; pp != eol; pp++)
1074 /* Remove drive letter */
1075 if (eol - p > 2 && isalpha(*p) && *(p + 1) == ':')
1079 if (strcmp(p, "[ExclusionList]") == 0)
1080 type = EXCLUSION_LIST;
1081 else if (strcmp(p, "[ExclusionException]") == 0)
1082 type = EXCLUSION_EXCEPTION;
1083 else if (strcmp(p, "[CompressionExclusionList]") == 0)
1084 type = COMPRESSION_EXCLUSION_LIST;
1085 else if (strcmp(p, "[AlignmentList]") == 0)
1086 type = ALIGNMENT_LIST;
1087 else if (p[0] == '[' && strrchr(p, ']')) {
1088 ERROR("Unknown capture configuration section `%s'", p);
1089 ret = WIMLIB_ERR_INVALID_CAPTURE_CONFIG;
1090 } else switch (type) {
1091 case EXCLUSION_LIST:
1092 DEBUG("Adding pattern \"%s\" to exclusion list", p);
1093 ret = pattern_list_add_pattern(&config->exclusion_list, p);
1095 case EXCLUSION_EXCEPTION:
1096 DEBUG("Adding pattern \"%s\" to exclusion exception list", p);
1097 ret = pattern_list_add_pattern(&config->exclusion_exception, p);
1099 case COMPRESSION_EXCLUSION_LIST:
1100 DEBUG("Adding pattern \"%s\" to compression exclusion list", p);
1101 ret = pattern_list_add_pattern(&config->compression_exclusion_list, p);
1103 case ALIGNMENT_LIST:
1104 DEBUG("Adding pattern \"%s\" to alignment list", p);
1105 ret = pattern_list_add_pattern(&config->alignment_list, p);
1108 ERROR("Line %lu of capture configuration is not "
1109 "in a block (such as [ExclusionList])",
1111 ret = WIMLIB_ERR_INVALID_CAPTURE_CONFIG;
1119 destroy_capture_config(config);
1123 static int capture_config_set_prefix(struct capture_config *config,
1124 const char *_prefix)
1126 char *prefix = STRDUP(_prefix);
1129 return WIMLIB_ERR_NOMEM;
1130 FREE(config->prefix);
1131 config->prefix = prefix;
1132 config->prefix_len = strlen(prefix);
1136 static bool match_pattern(const char *path, const char *path_basename,
1137 const struct pattern_list *list)
1139 for (size_t i = 0; i < list->num_pats; i++) {
1140 const char *pat = list->pats[i];
1143 /* Absolute path from root of capture */
1146 if (strchr(pat, '/'))
1147 /* Relative path from root of capture */
1150 /* A file name pattern */
1151 string = path_basename;
1153 if (fnmatch(pat, string, FNM_PATHNAME
1159 DEBUG("`%s' matches the pattern \"%s\"",
1167 /* Return true if the image capture configuration file indicates we should
1168 * exclude the filename @path from capture.
1170 * If @exclude_prefix is %true, the part of the path up and including the name
1171 * of the directory being captured is not included in the path for matching
1172 * purposes. This allows, for example, a pattern like /hiberfil.sys to match a
1173 * file /mnt/windows7/hiberfil.sys if we are capturing the /mnt/windows7
1176 bool exclude_path(const char *path, const struct capture_config *config,
1177 bool exclude_prefix)
1179 const char *basename = path_basename(path);
1180 if (exclude_prefix) {
1181 wimlib_assert(strlen(path) >= config->prefix_len);
1182 if (memcmp(config->prefix, path, config->prefix_len) == 0
1183 && path[config->prefix_len] == '/')
1184 path += config->prefix_len;
1186 return match_pattern(path, basename, &config->exclusion_list) &&
1187 !match_pattern(path, basename, &config->exclusion_exception);
1191 /* Strip leading and trailing forward slashes from a string. Modifies it in
1192 * place and returns the stripped string. */
1193 static const char *canonicalize_target_path(char *target_path)
1196 if (target_path == NULL)
1199 if (*target_path == '\0')
1201 else if (*target_path == '/')
1207 p = target_path + strlen(target_path) - 1;
1213 #if defined(__CYGWIN__) || defined(__WIN32__)
1214 static void zap_backslashes(char *s)
1224 /* Strip leading and trailing slashes from the target paths */
1225 static void canonicalize_targets(struct wimlib_capture_source *sources,
1228 while (num_sources--) {
1229 DEBUG("Canonicalizing { source: \"%s\", target=\"%s\"}",
1230 sources->fs_source_path,
1231 sources->wim_target_path);
1232 #if defined(__CYGWIN__) || defined(__WIN32__)
1233 /* The Windows API can handle forward slashes. Just get rid of
1234 * backslashes to avoid confusing other parts of the library
1236 zap_backslashes(sources->fs_source_path);
1237 if (sources->wim_target_path)
1238 zap_backslashes(sources->wim_target_path);
1240 sources->wim_target_path =
1241 (char*)canonicalize_target_path(sources->wim_target_path);
1242 DEBUG("Canonical target: \"%s\"", sources->wim_target_path);
1247 static int capture_source_cmp(const void *p1, const void *p2)
1249 const struct wimlib_capture_source *s1 = p1, *s2 = p2;
1250 return strcmp(s1->wim_target_path, s2->wim_target_path);
1253 /* Sorts the capture sources lexicographically by target path. This occurs
1254 * after leading and trailing forward slashes are stripped.
1256 * One purpose of this is to make sure that target paths that are inside other
1257 * target paths are added after the containing target paths. */
1258 static void sort_sources(struct wimlib_capture_source *sources,
1261 qsort(sources, num_sources, sizeof(sources[0]), capture_source_cmp);
1264 static int check_sorted_sources(struct wimlib_capture_source *sources,
1265 size_t num_sources, int add_image_flags)
1267 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_NTFS) {
1268 if (num_sources != 1) {
1269 ERROR("Must specify exactly 1 capture source "
1270 "(the NTFS volume) in NTFS mode!");
1271 return WIMLIB_ERR_INVALID_PARAM;
1273 if (sources[0].wim_target_path[0] != '\0') {
1274 ERROR("In NTFS capture mode the target path inside "
1275 "the image must be the root directory!");
1276 return WIMLIB_ERR_INVALID_PARAM;
1278 } else if (num_sources != 0) {
1279 /* This code is disabled because the current code
1280 * unconditionally attempts to do overlays. So, duplicate
1281 * target paths are OK. */
1283 if (num_sources > 1 && sources[0].wim_target_path[0] == '\0') {
1284 ERROR("Cannot specify root target when using multiple "
1285 "capture sources!");
1286 return WIMLIB_ERR_INVALID_PARAM;
1288 for (size_t i = 0; i < num_sources - 1; i++) {
1289 size_t len = strlen(sources[i].wim_target_path);
1291 const char *target1 = sources[i].wim_target_path;
1293 const char *target2 = sources[j].wim_target_path;
1294 DEBUG("target1=%s, target2=%s",
1296 if (strncmp(target1, target2, len) ||
1299 if (target2[len] == '/') {
1300 ERROR("Invalid target `%s': is a prefix of `%s'",
1302 return WIMLIB_ERR_INVALID_PARAM;
1304 if (target2[len] == '\0') {
1305 ERROR("Invalid target `%s': is a duplicate of `%s'",
1307 return WIMLIB_ERR_INVALID_PARAM;
1309 } while (++j != num_sources);
1317 /* Creates a new directory to place in the WIM image. This is to create parent
1318 * directories that are not part of any target as needed. */
1319 static struct wim_dentry *
1320 new_filler_directory(const char *name)
1322 struct wim_dentry *dentry;
1323 DEBUG("Creating filler directory \"%s\"", name);
1324 dentry = new_dentry_with_inode(name);
1326 /* Set the inode number to 0 for now. The final inode number
1327 * will be assigned later by assign_inode_numbers(). */
1328 dentry->d_inode->i_ino = 0;
1329 dentry->d_inode->i_resolved = 1;
1330 dentry->d_inode->i_attributes = FILE_ATTRIBUTE_DIRECTORY;
1335 /* Transfers the children of @branch to @target. It is an error if @target is
1336 * not a directory or if both @branch and @target contain a child dentry with
1338 static int do_overlay(struct wim_dentry *target, struct wim_dentry *branch)
1340 struct rb_root *rb_root;
1342 if (!dentry_is_directory(target)) {
1343 ERROR("Cannot overlay directory `%s' over non-directory",
1344 branch->file_name_utf8);
1345 return WIMLIB_ERR_INVALID_OVERLAY;
1348 rb_root = &branch->d_inode->i_children;
1349 while (rb_root->rb_node) { /* While @branch has children... */
1350 struct wim_dentry *child = rbnode_dentry(rb_root->rb_node);
1351 /* Move @child to the directory @target */
1352 unlink_dentry(child);
1353 if (!dentry_add_child(target, child)) {
1354 /* Revert the change to avoid leaking the directory tree
1355 * rooted at @child */
1356 dentry_add_child(branch, child);
1357 ERROR("Overlay error: file `%s' already exists "
1358 "as a child of `%s'",
1359 child->file_name_utf8, target->file_name_utf8);
1360 return WIMLIB_ERR_INVALID_OVERLAY;
1367 /* Attach or overlay a branch onto the WIM image.
1370 * Pointer to the root of the WIM image, or pointer to NULL if it has not
1375 * Path in the WIM image to add the branch, with leading and trailing
1378 static int attach_branch(struct wim_dentry **root_p,
1379 struct wim_dentry *branch,
1383 struct wim_dentry *dentry, *parent, *target;
1385 if (*target_path == '\0') {
1386 /* Target: root directory */
1388 /* Overlay on existing root */
1389 return do_overlay(*root_p, branch);
1397 /* Adding a non-root branch. Create root if it hasn't been created
1400 *root_p = new_filler_directory("");
1402 return WIMLIB_ERR_NOMEM;
1405 /* Walk the path to the branch, creating filler directories as needed.
1408 while ((slash = strchr(target_path, '/'))) {
1410 dentry = get_dentry_child_with_name(parent, target_path);
1412 dentry = new_filler_directory(target_path);
1414 return WIMLIB_ERR_NOMEM;
1415 dentry_add_child(parent, dentry);
1418 target_path = slash;
1419 /* Skip over slashes. Note: this cannot overrun the length of
1420 * the string because the last character cannot be a slash, as
1421 * trailing slashes were tripped. */
1424 } while (*target_path == '/');
1427 /* If the target path already existed, overlay the branch onto it.
1428 * Otherwise, set the branch as the target path. */
1429 target = get_dentry_child_with_name(parent, branch->file_name_utf8);
1431 return do_overlay(target, branch);
1433 dentry_add_child(parent, branch);
1438 WIMLIBAPI int wimlib_add_image_multisource(WIMStruct *w,
1439 struct wimlib_capture_source *sources,
1442 const char *config_str,
1444 int add_image_flags,
1445 wimlib_progress_func_t progress_func)
1447 int (*capture_tree)(struct wim_dentry **, const char *,
1448 struct wim_lookup_table *,
1449 struct wim_security_data *,
1450 const struct capture_config *,
1451 int, wimlib_progress_func_t, void *);
1453 struct wim_dentry *root_dentry;
1454 struct wim_dentry *branch;
1455 struct wim_security_data *sd;
1456 struct capture_config config;
1457 struct wim_image_metadata *imd;
1460 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_NTFS) {
1462 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_DEREFERENCE) {
1463 ERROR("Cannot dereference files when capturing directly from NTFS");
1464 return WIMLIB_ERR_INVALID_PARAM;
1466 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_UNIX_DATA) {
1467 ERROR("Capturing UNIX owner and mode not supported "
1468 "when capturing directly from NTFS");
1469 return WIMLIB_ERR_INVALID_PARAM;
1471 capture_tree = build_dentry_tree_ntfs;
1472 extra_arg = &w->ntfs_vol;
1474 ERROR("wimlib was compiled without support for NTFS-3g, so\n"
1475 " cannot capture a WIM image directly from a NTFS volume!");
1476 return WIMLIB_ERR_UNSUPPORTED;
1479 capture_tree = build_dentry_tree;
1483 #if defined(__CYGWIN__) || defined(__WIN32__)
1484 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_UNIX_DATA) {
1485 ERROR("Capturing UNIX-specific data is not supported on Windows");
1486 return WIMLIB_ERR_INVALID_PARAM;
1488 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_DEREFERENCE) {
1489 ERROR("Dereferencing symbolic links is not supported on Windows");
1490 return WIMLIB_ERR_INVALID_PARAM;
1494 if (!name || !*name) {
1495 ERROR("Must specify a non-empty string for the image name");
1496 return WIMLIB_ERR_INVALID_PARAM;
1499 if (w->hdr.total_parts != 1) {
1500 ERROR("Cannot add an image to a split WIM");
1501 return WIMLIB_ERR_SPLIT_UNSUPPORTED;
1504 if (wimlib_image_name_in_use(w, name)) {
1505 ERROR("There is already an image named \"%s\" in `%s'",
1507 return WIMLIB_ERR_IMAGE_NAME_COLLISION;
1511 DEBUG("Using default capture configuration");
1512 config_str = default_config;
1513 config_len = strlen(default_config);
1515 ret = init_capture_config(&config, config_str, config_len);
1519 DEBUG("Allocating security data");
1520 sd = CALLOC(1, sizeof(struct wim_security_data));
1522 ret = WIMLIB_ERR_NOMEM;
1523 goto out_destroy_capture_config;
1525 sd->total_length = 8;
1528 DEBUG("Using %zu capture sources", num_sources);
1529 canonicalize_targets(sources, num_sources);
1530 sort_sources(sources, num_sources);
1531 ret = check_sorted_sources(sources, num_sources, add_image_flags);
1533 ret = WIMLIB_ERR_INVALID_PARAM;
1534 goto out_free_security_data;
1537 DEBUG("Building dentry tree.");
1538 if (num_sources == 0) {
1539 root_dentry = new_filler_directory("");
1541 ret = WIMLIB_ERR_NOMEM;
1542 goto out_free_security_data;
1551 union wimlib_progress_info progress;
1553 DEBUG("Building dentry tree for source %zu of %zu "
1554 "(\"%s\" => \"%s\")", i + 1, num_sources,
1555 sources[i].fs_source_path,
1556 sources[i].wim_target_path);
1557 if (progress_func) {
1558 memset(&progress, 0, sizeof(progress));
1559 progress.scan.source = sources[i].fs_source_path;
1560 progress.scan.wim_target_path = sources[i].wim_target_path;
1561 progress_func(WIMLIB_PROGRESS_MSG_SCAN_BEGIN, &progress);
1563 ret = capture_config_set_prefix(&config,
1564 sources[i].fs_source_path);
1566 goto out_free_dentry_tree;
1567 flags = add_image_flags | WIMLIB_ADD_IMAGE_FLAG_SOURCE;
1568 if (!*sources[i].wim_target_path)
1569 flags |= WIMLIB_ADD_IMAGE_FLAG_ROOT;
1570 ret = (*capture_tree)(&branch, sources[i].fs_source_path,
1571 w->lookup_table, sd,
1574 progress_func, extra_arg);
1576 ERROR("Failed to build dentry tree for `%s'",
1577 sources[i].fs_source_path);
1578 goto out_free_dentry_tree;
1581 /* Use the target name, not the source name, for
1582 * the root of each branch from a capture
1583 * source. (This will also set the root dentry
1584 * of the entire image to be unnamed.) */
1585 ret = set_dentry_name(branch,
1586 path_basename(sources[i].wim_target_path));
1588 goto out_free_branch;
1590 ret = attach_branch(&root_dentry, branch,
1591 sources[i].wim_target_path);
1593 goto out_free_branch;
1596 progress_func(WIMLIB_PROGRESS_MSG_SCAN_END, &progress);
1597 } while (++i != num_sources);
1600 DEBUG("Calculating full paths of dentries.");
1601 ret = for_dentry_in_tree(root_dentry, calculate_dentry_full_path, NULL);
1603 goto out_free_dentry_tree;
1605 ret = add_new_dentry_tree(w, root_dentry, sd);
1607 goto out_free_dentry_tree;
1609 imd = &w->image_metadata[w->hdr.image_count - 1];
1611 ret = dentry_tree_fix_inodes(root_dentry, &imd->inode_list);
1613 goto out_destroy_imd;
1615 DEBUG("Assigning hard link group IDs");
1616 assign_inode_numbers(&imd->inode_list);
1618 ret = xml_add_image(w, name);
1620 goto out_destroy_imd;
1622 if (add_image_flags & WIMLIB_ADD_IMAGE_FLAG_BOOT)
1623 wimlib_set_boot_idx(w, w->hdr.image_count);
1627 destroy_image_metadata(&w->image_metadata[w->hdr.image_count - 1],
1629 w->hdr.image_count--;
1632 free_dentry_tree(branch, w->lookup_table);
1633 out_free_dentry_tree:
1634 free_dentry_tree(root_dentry, w->lookup_table);
1635 out_free_security_data:
1636 free_security_data(sd);
1637 out_destroy_capture_config:
1638 destroy_capture_config(&config);
1643 WIMLIBAPI int wimlib_add_image(WIMStruct *w, const char *source,
1644 const char *name, const char *config_str,
1645 size_t config_len, int add_image_flags,
1646 wimlib_progress_func_t progress_func)
1648 if (!source || !*source)
1649 return WIMLIB_ERR_INVALID_PARAM;
1651 char *fs_source_path = STRDUP(source);
1653 struct wimlib_capture_source capture_src = {
1654 .fs_source_path = fs_source_path,
1655 .wim_target_path = NULL,
1658 ret = wimlib_add_image_multisource(w, &capture_src, 1, name,
1659 config_str, config_len,
1660 add_image_flags, progress_func);
1661 FREE(fs_source_path);