4 * Capture a WIM image from a NTFS volume. We capture everything we can,
5 * including security data and alternate data streams.
9 * Copyright (C) 2012 Eric Biggers
11 * This file is part of wimlib, a library for working with WIM files.
13 * wimlib is free software; you can redistribute it and/or modify it under the
14 * terms of the GNU General Public License as published by the Free
15 * Software Foundation; either version 3 of the License, or (at your option)
18 * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
19 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
20 * A PARTICULAR PURPOSE. See the GNU General Public License for more
23 * You should have received a copy of the GNU General Public License
24 * along with wimlib; if not, see http://www.gnu.org/licenses/.
31 #include <ntfs-3g/endians.h>
32 #include <ntfs-3g/types.h>
35 #include "wimlib_internal.h"
40 #include "lookup_table.h"
42 #include <ntfs-3g/layout.h>
43 #include <ntfs-3g/acls.h>
44 #include <ntfs-3g/attrib.h>
45 #include <ntfs-3g/misc.h>
46 #include <ntfs-3g/reparse.h>
47 #include <ntfs-3g/security.h> /* security.h before xattrs.h */
48 #include <ntfs-3g/xattrs.h>
49 #include <ntfs-3g/volume.h>
54 /* Structure that allows searching the security descriptors by SHA1 message
57 struct wim_security_data *sd;
61 /* Binary tree node of security descriptors, indexed by the @hash field. */
64 u8 hash[SHA1_HASH_SIZE];
66 struct sd_node *right;
69 static void free_sd_tree(struct sd_node *root)
72 free_sd_tree(root->left);
73 free_sd_tree(root->right);
77 /* Frees a security descriptor index set. */
78 static void destroy_sd_set(struct sd_set *sd_set)
80 free_sd_tree(sd_set->root);
83 /* Inserts a a new node into the security descriptor index tree. */
84 static void insert_sd_node(struct sd_node *new, struct sd_node *root)
86 int cmp = hashes_cmp(new->hash, root->hash);
89 insert_sd_node(new, root->left);
94 insert_sd_node(new, root->right);
102 /* Returns the security ID of the security data having a SHA1 message digest of
103 * @hash in the security descriptor index tree rooted at @root.
105 * If not found, return -1. */
106 static int lookup_sd(const u8 hash[SHA1_HASH_SIZE], struct sd_node *root)
111 cmp = hashes_cmp(hash, root->hash);
113 return lookup_sd(hash, root->left);
115 return lookup_sd(hash, root->right);
117 return root->security_id;
121 * Adds a security descriptor to the indexed security descriptor set as well as
122 * the corresponding `struct wim_security_data', and returns the new security
123 * ID; or, if there is an existing security descriptor that is the same, return
124 * the security ID for it. If a new security descriptor cannot be allocated,
127 static int sd_set_add_sd(struct sd_set *sd_set, const char descriptor[],
130 u8 hash[SHA1_HASH_SIZE];
136 struct wim_security_data *sd;
138 sha1_buffer((const u8*)descriptor, size, hash);
140 security_id = lookup_sd(hash, sd_set->root);
141 if (security_id >= 0)
144 new = MALLOC(sizeof(*new));
147 descr_copy = MALLOC(size);
153 memcpy(descr_copy, descriptor, size);
154 new->security_id = sd->num_entries;
157 copy_hash(new->hash, hash);
160 descriptors = REALLOC(sd->descriptors,
161 (sd->num_entries + 1) * sizeof(sd->descriptors[0]));
164 sd->descriptors = descriptors;
165 sizes = REALLOC(sd->sizes,
166 (sd->num_entries + 1) * sizeof(sd->sizes[0]));
170 sd->descriptors[sd->num_entries] = descr_copy;
171 sd->sizes[sd->num_entries] = size;
173 DEBUG("There are now %d security descriptors", sd->num_entries);
174 sd->total_length += size + sizeof(sd->sizes[0]);
177 insert_sd_node(new, sd_set->root);
180 return new->security_id;
189 static inline ntfschar *attr_record_name(ATTR_RECORD *ar)
191 return (ntfschar*)((u8*)ar + le16_to_cpu(ar->name_offset));
194 /* Calculates the SHA1 message digest of a NTFS attribute.
196 * @ni: The NTFS inode containing the attribute.
197 * @ar: The ATTR_RECORD describing the attribute.
198 * @md: If successful, the returned SHA1 message digest.
199 * @reparse_tag_ret: Optional pointer into which the first 4 bytes of the
200 * attribute will be written (to get the reparse
203 * Return 0 on success or nonzero on error.
205 static int ntfs_attr_sha1sum(ntfs_inode *ni, ATTR_RECORD *ar,
206 u8 md[SHA1_HASH_SIZE],
207 bool is_reparse_point,
208 u32 *reparse_tag_ret)
216 na = ntfs_attr_open(ni, ar->type, attr_record_name(ar),
219 ERROR_WITH_ERRNO("Failed to open NTFS attribute");
220 return WIMLIB_ERR_NTFS_3G;
223 bytes_remaining = na->data_size;
225 if (is_reparse_point) {
226 if (ntfs_attr_pread(na, 0, 8, buf) != 8)
228 *reparse_tag_ret = le32_to_cpu(*(u32*)buf);
230 bytes_remaining -= 8;
234 while (bytes_remaining) {
235 s64 to_read = min(bytes_remaining, sizeof(buf));
236 if (ntfs_attr_pread(na, pos, to_read, buf) != to_read)
238 sha1_update(&ctx, buf, to_read);
240 bytes_remaining -= to_read;
242 sha1_final(md, &ctx);
246 ERROR_WITH_ERRNO("Error reading NTFS attribute");
247 return WIMLIB_ERR_NTFS_3G;
250 /* Load the streams from a file or reparse point in the NTFS volume into the WIM
252 static int capture_ntfs_streams(struct dentry *dentry, ntfs_inode *ni,
253 char path[], size_t path_len,
254 struct lookup_table *lookup_table,
255 ntfs_volume **ntfs_vol_p,
258 ntfs_attr_search_ctx *actx;
259 u8 attr_hash[SHA1_HASH_SIZE];
260 struct ntfs_location *ntfs_loc = NULL;
262 struct lookup_table_entry *lte;
264 DEBUG2("Capturing NTFS data streams from `%s'", path);
266 /* Get context to search the streams of the NTFS file. */
267 actx = ntfs_attr_get_search_ctx(ni, NULL);
269 ERROR_WITH_ERRNO("Cannot get NTFS attribute search "
271 return WIMLIB_ERR_NTFS_3G;
274 /* Capture each data stream or reparse data stream. */
275 while (!ntfs_attr_lookup(type, NULL, 0,
276 CASE_SENSITIVE, 0, NULL, 0, actx))
278 char *stream_name_utf8;
280 u64 data_size = ntfs_get_attribute_value_length(actx->attr);
281 u64 name_length = actx->attr->name_length;
283 if (data_size == 0) {
285 ERROR_WITH_ERRNO("Failed to get size of attribute of "
287 ret = WIMLIB_ERR_NTFS_3G;
290 /* Empty stream. No lookup table entry is needed. */
293 if (type == AT_REPARSE_POINT && data_size < 8) {
294 ERROR("`%s': reparse point buffer too small",
296 ret = WIMLIB_ERR_NTFS_3G;
299 /* Checksum the stream. */
300 ret = ntfs_attr_sha1sum(ni, actx->attr, attr_hash,
301 type == AT_REPARSE_POINT, &reparse_tag);
305 /* Make a lookup table entry for the stream, or use an existing
306 * one if there's already an identical stream. */
307 lte = __lookup_resource(lookup_table, attr_hash);
308 ret = WIMLIB_ERR_NOMEM;
312 ntfs_loc = CALLOC(1, sizeof(*ntfs_loc));
315 ntfs_loc->ntfs_vol_p = ntfs_vol_p;
316 ntfs_loc->path_utf8 = MALLOC(path_len + 1);
317 if (!ntfs_loc->path_utf8)
318 goto out_free_ntfs_loc;
319 memcpy(ntfs_loc->path_utf8, path, path_len + 1);
321 ntfs_loc->stream_name_utf16 = MALLOC(name_length * 2);
322 if (!ntfs_loc->stream_name_utf16)
323 goto out_free_ntfs_loc;
324 memcpy(ntfs_loc->stream_name_utf16,
325 attr_record_name(actx->attr),
326 actx->attr->name_length * 2);
327 ntfs_loc->stream_name_utf16_num_chars = name_length;
330 lte = new_lookup_table_entry();
332 goto out_free_ntfs_loc;
333 lte->ntfs_loc = ntfs_loc;
334 lte->resource_location = RESOURCE_IN_NTFS_VOLUME;
335 if (type == AT_REPARSE_POINT) {
336 dentry->d_inode->reparse_tag = reparse_tag;
337 ntfs_loc->is_reparse_point = true;
338 lte->resource_entry.original_size = data_size - 8;
339 lte->resource_entry.size = data_size - 8;
341 ntfs_loc->is_reparse_point = false;
342 lte->resource_entry.original_size = data_size;
343 lte->resource_entry.size = data_size;
346 DEBUG("Add resource for `%s' (size = %zu)",
347 dentry->file_name_utf8,
348 lte->resource_entry.original_size);
349 copy_hash(lte->hash, attr_hash);
350 lookup_table_insert(lookup_table, lte);
353 if (name_length == 0) {
354 /* Unnamed data stream. Put the reference to it in the
356 if (dentry->d_inode->lte) {
357 ERROR("Found two un-named data streams for "
359 ret = WIMLIB_ERR_NTFS_3G;
362 dentry->d_inode->lte = lte;
364 /* Named data stream. Put the reference to it in the
365 * alternate data stream entries */
366 struct ads_entry *new_ads_entry;
367 size_t stream_name_utf8_len;
368 stream_name_utf8 = utf16_to_utf8((const char*)attr_record_name(actx->attr),
370 &stream_name_utf8_len);
371 if (!stream_name_utf8)
373 new_ads_entry = inode_add_ads(dentry->d_inode, stream_name_utf8);
374 FREE(stream_name_utf8);
378 wimlib_assert(new_ads_entry->stream_name_len == name_length * 2);
380 new_ads_entry->lte = lte;
386 free_lookup_table_entry(lte);
389 FREE(ntfs_loc->path_utf8);
390 FREE(ntfs_loc->stream_name_utf16);
394 ntfs_attr_put_search_ctx(actx);
396 DEBUG2("Successfully captured NTFS streams from `%s'", path);
398 ERROR("Failed to capture NTFS streams from `%s", path);
403 struct dentry *parent;
407 struct lookup_table *lookup_table;
408 struct sd_set *sd_set;
409 const struct capture_config *config;
410 ntfs_volume **ntfs_vol_p;
415 build_dentry_tree_ntfs_recursive(struct dentry **root_p, ntfs_inode *dir_ni,
416 ntfs_inode *ni, char path[], size_t path_len,
418 struct lookup_table *lookup_table,
419 struct sd_set *sd_set,
420 const struct capture_config *config,
421 ntfs_volume **ntfs_vol_p,
424 static int wim_ntfs_capture_filldir(void *dirent, const ntfschar *name,
425 const int name_len, const int name_type,
426 const s64 pos, const MFT_REF mref,
427 const unsigned dt_type)
429 struct readdir_ctx *ctx;
430 size_t utf8_name_len;
432 struct dentry *child = NULL;
436 if (name_type == FILE_NAME_DOS)
441 utf8_name = utf16_to_utf8((const char*)name, name_len * 2,
446 if (utf8_name[0] == '.' &&
447 (utf8_name[1] == '\0' ||
448 (utf8_name[1] == '.' && utf8_name[2] == '\0'))) {
450 goto out_free_utf8_name;
455 ntfs_inode *ni = ntfs_inode_open(ctx->dir_ni->vol, mref);
457 ERROR_WITH_ERRNO("Failed to open NTFS inode");
458 goto out_free_utf8_name;
460 path_len = ctx->path_len;
462 ctx->path[path_len++] = '/';
463 memcpy(ctx->path + path_len, utf8_name, utf8_name_len + 1);
464 path_len += utf8_name_len;
465 ret = build_dentry_tree_ntfs_recursive(&child, ctx->dir_ni,
466 ni, ctx->path, path_len, name_type,
467 ctx->lookup_table, ctx->sd_set,
468 ctx->config, ctx->ntfs_vol_p,
472 dentry_add_child(ctx->parent, child);
474 ntfs_inode_close(ni);
481 static int change_dentry_short_name(struct dentry *dentry,
482 const char short_name_utf8[],
483 int short_name_utf8_len)
485 size_t short_name_utf16_len;
486 char *short_name_utf16;
487 short_name_utf16 = utf8_to_utf16(short_name_utf8, short_name_utf8_len,
488 &short_name_utf16_len);
489 if (!short_name_utf16) {
490 ERROR_WITH_ERRNO("Failed to convert short name to UTF-16");
491 return WIMLIB_ERR_NOMEM;
493 dentry->short_name = short_name_utf16;
494 dentry->short_name_len = short_name_utf16_len;
498 /* Recursively build a WIM dentry tree corresponding to a NTFS volume.
499 * At the same time, update the WIM lookup table with lookup table entries for
500 * the NTFS streams, and build an array of security descriptors.
502 static int build_dentry_tree_ntfs_recursive(struct dentry **root_p,
508 struct lookup_table *lookup_table,
509 struct sd_set *sd_set,
510 const struct capture_config *config,
511 ntfs_volume **ntfs_vol_p,
517 char dos_name_utf8[64];
520 mrec_flags = ni->mrec->flags;
521 struct SECURITY_CONTEXT ctx;
522 memset(&ctx, 0, sizeof(ctx));
524 ret = ntfs_xattr_system_getxattr(&ctx, XATTR_NTFS_ATTRIB,
525 ni, dir_ni, (char *)&attributes,
528 ERROR_WITH_ERRNO("Failed to get NTFS attributes from `%s'",
530 return WIMLIB_ERR_NTFS_3G;
533 if (exclude_path(path, config, false)) {
534 if (flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE) {
535 const char *file_type;
536 if (attributes & MFT_RECORD_IS_DIRECTORY)
537 file_type = "directory";
540 printf("Excluding %s `%s' from capture\n",
547 if (flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE)
548 printf("Scanning `%s'\n", path);
550 root = new_dentry_with_timeless_inode(path_basename(path));
552 return WIMLIB_ERR_NOMEM;
555 if (dir_ni && (name_type == FILE_NAME_WIN32_AND_DOS
556 || name_type == FILE_NAME_WIN32))
558 ret = ntfs_get_ntfs_dos_name(ni, dir_ni, dos_name_utf8,
559 sizeof(dos_name_utf8));
561 DEBUG("Changing short name of `%s'", path);
562 ret = change_dentry_short_name(root, dos_name_utf8,
568 if (errno != ENODATA) {
569 ERROR_WITH_ERRNO("Error getting DOS name "
571 return WIMLIB_ERR_NTFS_3G;
577 root->d_inode->creation_time = le64_to_cpu(ni->creation_time);
578 root->d_inode->last_write_time = le64_to_cpu(ni->last_data_change_time);
579 root->d_inode->last_access_time = le64_to_cpu(ni->last_access_time);
580 root->d_inode->attributes = le32_to_cpu(attributes);
581 root->d_inode->ino = ni->mft_no;
582 root->d_inode->resolved = true;
584 if (attributes & FILE_ATTR_REPARSE_POINT) {
585 /* Junction point, symbolic link, or other reparse point */
586 ret = capture_ntfs_streams(root, ni, path, path_len,
587 lookup_table, ntfs_vol_p,
589 } else if (mrec_flags & MFT_RECORD_IS_DIRECTORY) {
591 /* Normal directory */
593 struct readdir_ctx ctx = {
597 .path_len = path_len,
598 .lookup_table = lookup_table,
601 .ntfs_vol_p = ntfs_vol_p,
604 ret = ntfs_readdir(ni, &pos, &ctx, wim_ntfs_capture_filldir);
606 ERROR_WITH_ERRNO("ntfs_readdir()");
607 ret = WIMLIB_ERR_NTFS_3G;
611 ret = capture_ntfs_streams(root, ni, path, path_len,
612 lookup_table, ntfs_vol_p,
621 ret = ntfs_xattr_system_getxattr(&ctx, XATTR_NTFS_ACL,
624 if (ret > sizeof(sd)) {
626 ret = ntfs_xattr_system_getxattr(&ctx, XATTR_NTFS_ACL,
627 ni, dir_ni, sd, ret);
630 root->d_inode->security_id = sd_set_add_sd(sd_set, sd, ret);
631 if (root->d_inode->security_id == -1) {
632 ERROR("Out of memory");
633 return WIMLIB_ERR_NOMEM;
635 DEBUG("Added security ID = %u for `%s'",
636 root->d_inode->security_id, path);
638 } else if (ret < 0) {
639 ERROR_WITH_ERRNO("Failed to get security information from "
641 ret = WIMLIB_ERR_NTFS_3G;
643 root->d_inode->security_id = -1;
644 DEBUG("No security ID for `%s'", path);
649 static int build_dentry_tree_ntfs(struct dentry **root_p,
651 struct lookup_table *lookup_table,
652 struct wim_security_data *sd,
653 const struct capture_config *config,
660 struct sd_set sd_set = {
664 ntfs_volume **ntfs_vol_p = extra_arg;
666 DEBUG("Mounting NTFS volume `%s' read-only", device);
668 vol = ntfs_mount(device, MS_RDONLY);
670 ERROR_WITH_ERRNO("Failed to mount NTFS volume `%s' read-only",
672 return WIMLIB_ERR_NTFS_3G;
674 ntfs_open_secure(vol);
676 /* We don't want to capture the special NTFS files such as $Bitmap. Not
677 * to be confused with "hidden" or "system" files which are real files
678 * that we do need to capture. */
679 NVolClearShowSysFiles(vol);
681 DEBUG("Opening root NTFS dentry");
682 root_ni = ntfs_inode_open(vol, FILE_root);
684 ERROR_WITH_ERRNO("Failed to open root inode of NTFS volume "
686 ret = WIMLIB_ERR_NTFS_3G;
690 /* Currently we assume that all the UTF-8 paths fit into this length and
691 * there is no check for overflow. */
692 char *path = MALLOC(32768);
694 ERROR("Could not allocate memory for NTFS pathname");
700 ret = build_dentry_tree_ntfs_recursive(root_p, NULL, root_ni, path, 1,
701 FILE_NAME_POSIX, lookup_table,
702 &sd_set, config, ntfs_vol_p,
706 ntfs_inode_close(root_ni);
707 destroy_sd_set(&sd_set);
711 if (ntfs_umount(vol, FALSE) != 0) {
712 ERROR_WITH_ERRNO("Failed to unmount NTFS volume `%s'",
715 ret = WIMLIB_ERR_NTFS_3G;
718 /* We need to leave the NTFS volume mounted so that we can read
719 * the NTFS files again when we are actually writing the WIM */
727 WIMLIBAPI int wimlib_add_image_from_ntfs_volume(WIMStruct *w,
730 const char *config_str,
734 if (flags & (WIMLIB_ADD_IMAGE_FLAG_DEREFERENCE)) {
735 ERROR("Cannot dereference files when capturing directly from NTFS");
736 return WIMLIB_ERR_INVALID_PARAM;
738 return do_add_image(w, device, name, config_str, config_len, flags,
739 build_dentry_tree_ntfs, &w->ntfs_vol);
742 #else /* WITH_NTFS_3G */
743 WIMLIBAPI int wimlib_add_image_from_ntfs_volume(WIMStruct *w,
746 const char *config_str,
750 ERROR("wimlib was compiled without support for NTFS-3g, so");
751 ERROR("we cannot capture a WIM image directly from a NTFS volume");
752 return WIMLIB_ERR_UNSUPPORTED;
754 #endif /* WITH_NTFS_3G */