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 Lesser General Public License as published by the Free
15 * Software Foundation; either version 2.1 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 Lesser General Public License for more
23 * You should have received a copy of the GNU Lesser General Public License
24 * along with wimlib; if not, see http://www.gnu.org/licenses/.
28 #include "wimlib_internal.h"
33 #include "lookup_table.h"
35 #include <ntfs-3g/layout.h>
36 #include <ntfs-3g/acls.h>
37 #include <ntfs-3g/attrib.h>
38 #include <ntfs-3g/misc.h>
39 #include <ntfs-3g/reparse.h>
40 #include <ntfs-3g/security.h>
41 #include <ntfs-3g/volume.h>
45 extern int ntfs_inode_get_security(ntfs_inode *ni, u32 selection, char *buf,
46 u32 buflen, u32 *psize);
48 extern int ntfs_inode_get_attributes(ntfs_inode *ni);
50 /* Structure that allows searching the security descriptors by SHA1 message
53 struct wim_security_data *sd;
57 /* Binary tree node of security descriptors, indexed by the @hash field. */
60 u8 hash[SHA1_HASH_SIZE];
62 struct sd_node *right;
65 static void free_sd_tree(struct sd_node *root)
68 free_sd_tree(root->left);
69 free_sd_tree(root->right);
73 /* Frees a security descriptor index set. */
74 static void destroy_sd_set(struct sd_set *sd_set)
76 free_sd_tree(sd_set->root);
79 /* Inserts a a new node into the security descriptor index tree. */
80 static void insert_sd_node(struct sd_node *new, struct sd_node *root)
82 int cmp = hashes_cmp(new->hash, root->hash);
85 insert_sd_node(new, root->left);
90 insert_sd_node(new, root->right);
98 /* Returns the security ID of the security data having a SHA1 message digest of
99 * @hash in the security descriptor index tree rooted at @root.
101 * If not found, return -1. */
102 static int lookup_sd(const u8 hash[SHA1_HASH_SIZE], struct sd_node *root)
107 cmp = hashes_cmp(hash, root->hash);
109 return lookup_sd(hash, root->left);
111 return lookup_sd(hash, root->right);
113 return root->security_id;
117 * Adds a security descriptor to the indexed security descriptor set as well as
118 * the corresponding `struct wim_security_data', and returns the new security
119 * ID; or, if there is an existing security descriptor that is the same, return
120 * the security ID for it. If a new security descriptor cannot be allocated,
123 static int sd_set_add_sd(struct sd_set *sd_set, const char descriptor[],
126 u8 hash[SHA1_HASH_SIZE];
132 struct wim_security_data *sd;
134 sha1_buffer((const u8*)descriptor, size, hash);
136 security_id = lookup_sd(hash, sd_set->root);
137 if (security_id >= 0)
140 new = MALLOC(sizeof(*new));
143 descr_copy = MALLOC(size);
149 memcpy(descr_copy, descriptor, size);
150 new->security_id = sd->num_entries;
153 copy_hash(new->hash, hash);
156 descriptors = REALLOC(sd->descriptors,
157 (sd->num_entries + 1) * sizeof(sd->descriptors[0]));
160 sd->descriptors = descriptors;
161 sizes = REALLOC(sd->sizes,
162 (sd->num_entries + 1) * sizeof(sd->sizes[0]));
166 sd->descriptors[sd->num_entries] = descr_copy;
167 sd->sizes[sd->num_entries] = size;
169 DEBUG("There are now %d security descriptors", sd->num_entries);
170 sd->total_length += size + sizeof(sd->sizes[0]);
173 insert_sd_node(new, sd_set->root);
176 return new->security_id;
185 static inline ntfschar *attr_record_name(ATTR_RECORD *ar)
187 return (ntfschar*)((u8*)ar + le16_to_cpu(ar->name_offset));
190 /* Calculates the SHA1 message digest of a NTFS attribute.
192 * @ni: The NTFS inode containing the attribute.
193 * @ar: The ATTR_RECORD describing the attribute.
194 * @md: If successful, the returned SHA1 message digest.
195 * @reparse_tag_ret: Optional pointer into which the first 4 bytes of the
196 * attribute will be written (to get the reparse
199 * Return 0 on success or nonzero on error.
201 static int ntfs_attr_sha1sum(ntfs_inode *ni, ATTR_RECORD *ar,
202 u8 md[SHA1_HASH_SIZE],
203 u32 *reparse_tag_ret)
211 na = ntfs_attr_open(ni, ar->type, attr_record_name(ar),
214 ERROR_WITH_ERRNO("Failed to open NTFS attribute");
215 return WIMLIB_ERR_NTFS_3G;
218 bytes_remaining = na->data_size;
221 DEBUG2("Calculating SHA1 message digest (%"PRIu64" bytes)",
224 while (bytes_remaining) {
225 s64 to_read = min(bytes_remaining, sizeof(buf));
226 if (ntfs_attr_pread(na, pos, to_read, buf) != to_read) {
227 ERROR_WITH_ERRNO("Error reading NTFS attribute");
228 return WIMLIB_ERR_NTFS_3G;
230 if (bytes_remaining == na->data_size && reparse_tag_ret)
231 *reparse_tag_ret = le32_to_cpu(*(u32*)buf);
232 sha1_update(&ctx, buf, to_read);
234 bytes_remaining -= to_read;
236 sha1_final(md, &ctx);
241 /* Load the streams from a WIM file or reparse point in the NTFS volume into the
242 * WIM lookup table */
243 static int capture_ntfs_streams(struct dentry *dentry, ntfs_inode *ni,
244 char path[], size_t path_len,
245 struct lookup_table *lookup_table,
246 ntfs_volume **ntfs_vol_p,
250 ntfs_attr_search_ctx *actx;
251 u8 attr_hash[SHA1_HASH_SIZE];
252 struct ntfs_location *ntfs_loc = NULL;
254 struct lookup_table_entry *lte;
256 DEBUG2("Capturing NTFS data streams from `%s'", path);
258 /* Get context to search the streams of the NTFS file. */
259 actx = ntfs_attr_get_search_ctx(ni, NULL);
261 ERROR_WITH_ERRNO("Cannot get NTFS attribute search "
263 return WIMLIB_ERR_NTFS_3G;
266 /* Capture each data stream or reparse data stream. */
267 while (!ntfs_attr_lookup(type, NULL, 0,
268 CASE_SENSITIVE, 0, NULL, 0, actx))
270 char *stream_name_utf8;
271 size_t stream_name_utf16_len;
273 u64 data_size = ntfs_get_attribute_value_length(actx->attr);
274 u64 name_length = actx->attr->name_length;
276 if (data_size == 0) {
278 ERROR_WITH_ERRNO("Failed to get size of attribute of "
280 ret = WIMLIB_ERR_NTFS_3G;
283 /* Empty stream. No lookup table entry is needed. */
286 if (type == AT_REPARSE_POINT && data_size < 8) {
287 ERROR("`%s': reparse point buffer too small");
288 ret = WIMLIB_ERR_NTFS_3G;
291 /* Checksum the stream. */
292 ret = ntfs_attr_sha1sum(ni, actx->attr, attr_hash, &reparse_tag);
296 /* Make a lookup table entry for the stream, or use an existing
297 * one if there's already an identical stream. */
298 lte = __lookup_resource(lookup_table, attr_hash);
299 ret = WIMLIB_ERR_NOMEM;
303 ntfs_loc = CALLOC(1, sizeof(*ntfs_loc));
306 ntfs_loc->ntfs_vol_p = ntfs_vol_p;
307 ntfs_loc->path_utf8 = MALLOC(path_len + 1);
308 if (!ntfs_loc->path_utf8)
309 goto out_free_ntfs_loc;
310 memcpy(ntfs_loc->path_utf8, path, path_len + 1);
312 ntfs_loc->stream_name_utf16 = MALLOC(name_length * 2);
313 if (!ntfs_loc->stream_name_utf16)
314 goto out_free_ntfs_loc;
315 memcpy(ntfs_loc->stream_name_utf16,
316 attr_record_name(actx->attr),
317 actx->attr->name_length * 2);
318 ntfs_loc->stream_name_utf16_num_chars = name_length;
321 lte = new_lookup_table_entry();
323 goto out_free_ntfs_loc;
324 lte->ntfs_loc = ntfs_loc;
325 lte->resource_location = RESOURCE_IN_NTFS_VOLUME;
326 if (type == AT_REPARSE_POINT) {
327 dentry->reparse_tag = reparse_tag;
328 ntfs_loc->is_reparse_point = true;
329 lte->resource_entry.original_size = data_size - 8;
330 lte->resource_entry.size = data_size - 8;
332 ntfs_loc->is_reparse_point = false;
333 lte->resource_entry.original_size = data_size;
334 lte->resource_entry.size = data_size;
337 DEBUG("Add resource for `%s' (size = %zu)",
338 dentry->file_name_utf8,
339 lte->resource_entry.original_size);
340 copy_hash(lte->hash, attr_hash);
341 lookup_table_insert(lookup_table, lte);
344 if (name_length == 0) {
345 /* Unnamed data stream. Put the reference to it in the
348 ERROR("Found two un-named data streams for "
350 ret = WIMLIB_ERR_NTFS_3G;
355 /* Named data stream. Put the reference to it in the
356 * alternate data stream entries */
357 struct ads_entry *new_ads_entry;
358 size_t stream_name_utf8_len;
359 stream_name_utf8 = utf16_to_utf8((const char*)attr_record_name(actx->attr),
361 &stream_name_utf8_len);
362 if (!stream_name_utf8)
364 new_ads_entry = dentry_add_ads(dentry, stream_name_utf8);
365 FREE(stream_name_utf8);
369 wimlib_assert(new_ads_entry->stream_name_len == name_length * 2);
371 new_ads_entry->lte = lte;
377 free_lookup_table_entry(lte);
380 FREE(ntfs_loc->path_utf8);
381 FREE(ntfs_loc->stream_name_utf16);
385 ntfs_attr_put_search_ctx(actx);
387 DEBUG2("Successfully captured NTFS streams from `%s'", path);
389 ERROR("Failed to capture NTFS streams from `%s", path);
394 struct dentry *parent;
398 struct lookup_table *lookup_table;
399 struct sd_set *sd_set;
400 const struct capture_config *config;
401 ntfs_volume **ntfs_vol_p;
406 build_dentry_tree_ntfs_recursive(struct dentry **root_p, ntfs_inode *ni,
407 char path[], size_t path_len,
408 struct lookup_table *lookup_table,
409 struct sd_set *sd_set,
410 const struct capture_config *config,
411 ntfs_volume **ntfs_vol_p,
414 static int wim_ntfs_capture_filldir(void *dirent, const ntfschar *name,
415 const int name_len, const int name_type,
416 const s64 pos, const MFT_REF mref,
417 const unsigned dt_type)
419 struct readdir_ctx *ctx;
420 size_t utf8_name_len;
422 struct dentry *child = NULL;
426 if (name_type == FILE_NAME_DOS)
431 utf8_name = utf16_to_utf8((const char*)name, name_len * 2,
436 if (utf8_name[0] == '.' &&
437 (utf8_name[1] == '\0' ||
438 (utf8_name[1] == '.' && utf8_name[2] == '\0'))) {
440 goto out_free_utf8_name;
445 ntfs_inode *ni = ntfs_inode_open(ctx->dir_ni->vol, mref);
447 ERROR_WITH_ERRNO("Failed to open NTFS inode");
450 path_len = ctx->path_len;
452 ctx->path[path_len++] = '/';
453 memcpy(ctx->path + path_len, utf8_name, utf8_name_len + 1);
454 path_len += utf8_name_len;
455 ret = build_dentry_tree_ntfs_recursive(&child, ni, ctx->path, path_len,
456 ctx->lookup_table, ctx->sd_set,
457 ctx->config, ctx->ntfs_vol_p,
461 link_dentry(child, ctx->parent);
463 ntfs_inode_close(ni);
470 /* Recursively build a WIM dentry tree corresponding to a NTFS volume.
471 * At the same time, update the WIM lookup table with lookup table entries for
472 * the NTFS streams, and build an array of security descriptors.
474 static int build_dentry_tree_ntfs_recursive(struct dentry **root_p,
478 struct lookup_table *lookup_table,
479 struct sd_set *sd_set,
480 const struct capture_config *config,
481 ntfs_volume **ntfs_vol_p,
490 mrec_flags = ni->mrec->flags;
491 attributes = ntfs_inode_get_attributes(ni);
493 if (exclude_path(path, config, false)) {
494 if (flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE) {
495 const char *file_type;
496 if (attributes & MFT_RECORD_IS_DIRECTORY)
497 file_type = "directory";
500 printf("Excluding %s `%s' from capture\n",
507 if (flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE)
508 printf("Scanning `%s'\n", path);
510 root = new_dentry(path_basename(path));
512 return WIMLIB_ERR_NOMEM;
515 root->creation_time = le64_to_cpu(ni->creation_time);
516 root->last_write_time = le64_to_cpu(ni->last_data_change_time);
517 root->last_access_time = le64_to_cpu(ni->last_access_time);
518 root->attributes = le32_to_cpu(attributes);
519 root->link_group_id = ni->mft_no;
520 root->resolved = true;
522 if (attributes & FILE_ATTR_REPARSE_POINT) {
523 /* Junction point, symbolic link, or other reparse point */
524 ret = capture_ntfs_streams(root, ni, path, path_len,
525 lookup_table, ntfs_vol_p,
527 } else if (mrec_flags & MFT_RECORD_IS_DIRECTORY) {
529 /* Normal directory */
531 struct readdir_ctx ctx = {
535 .path_len = path_len,
536 .lookup_table = lookup_table,
539 .ntfs_vol_p = ntfs_vol_p,
542 ret = ntfs_readdir(ni, &pos, &ctx, wim_ntfs_capture_filldir);
544 ERROR_WITH_ERRNO("ntfs_readdir()");
545 ret = WIMLIB_ERR_NTFS_3G;
549 ret = capture_ntfs_streams(root, ni, path, path_len,
550 lookup_table, ntfs_vol_p,
556 ret = ntfs_inode_get_security(ni,
557 OWNER_SECURITY_INFORMATION |
558 GROUP_SECURITY_INFORMATION |
559 DACL_SECURITY_INFORMATION |
560 SACL_SECURITY_INFORMATION,
563 ret = ntfs_inode_get_security(ni,
564 OWNER_SECURITY_INFORMATION |
565 GROUP_SECURITY_INFORMATION |
566 DACL_SECURITY_INFORMATION |
567 SACL_SECURITY_INFORMATION,
568 sd, sd_size, &sd_size);
570 ERROR_WITH_ERRNO("Failed to get security information from "
572 ret = WIMLIB_ERR_NTFS_3G;
575 /*print_security_descriptor(sd, sd_size);*/
576 root->security_id = sd_set_add_sd(sd_set, sd, sd_size);
577 if (root->security_id == -1) {
578 ERROR("Out of memory");
579 return WIMLIB_ERR_NOMEM;
581 DEBUG("Added security ID = %u for `%s'",
582 root->security_id, path);
584 root->security_id = -1;
585 DEBUG("No security ID for `%s'", path);
592 static int build_dentry_tree_ntfs(struct dentry **root_p,
594 struct lookup_table *lookup_table,
595 struct wim_security_data *sd,
596 const struct capture_config *config,
603 struct sd_set sd_set = {
607 ntfs_volume **ntfs_vol_p = extra_arg;
609 DEBUG("Mounting NTFS volume `%s' read-only", device);
611 vol = ntfs_mount(device, MS_RDONLY);
613 ERROR_WITH_ERRNO("Failed to mount NTFS volume `%s' read-only",
615 return WIMLIB_ERR_NTFS_3G;
617 ntfs_open_secure(vol);
619 /* We don't want to capture the special NTFS files such as $Bitmap. Not
620 * to be confused with "hidden" or "system" files which are real files
621 * that we do need to capture. */
622 NVolClearShowSysFiles(vol);
624 DEBUG("Opening root NTFS dentry");
625 root_ni = ntfs_inode_open(vol, FILE_root);
627 ERROR_WITH_ERRNO("Failed to open root inode of NTFS volume "
629 ret = WIMLIB_ERR_NTFS_3G;
633 /* Currently we assume that all the UTF-8 paths fit into this length and
634 * there is no check for overflow. */
635 char *path = MALLOC(32768);
637 ERROR("Could not allocate memory for NTFS pathname");
643 ret = build_dentry_tree_ntfs_recursive(root_p, root_ni, path, 1,
644 lookup_table, &sd_set,
645 config, ntfs_vol_p, flags);
648 ntfs_inode_close(root_ni);
649 destroy_sd_set(&sd_set);
653 if (ntfs_umount(vol, FALSE) != 0) {
654 ERROR_WITH_ERRNO("Failed to unmount NTFS volume `%s'",
657 ret = WIMLIB_ERR_NTFS_3G;
660 /* We need to leave the NTFS volume mounted so that we can read
661 * the NTFS files again when we are actually writing the WIM */
669 WIMLIBAPI int wimlib_add_image_from_ntfs_volume(WIMStruct *w,
672 const char *config_str,
676 if (flags & (WIMLIB_ADD_IMAGE_FLAG_DEREFERENCE)) {
677 ERROR("Cannot dereference files when capturing directly from NTFS");
678 return WIMLIB_ERR_INVALID_PARAM;
680 return do_add_image(w, device, name, config_str, config_len, flags,
681 build_dentry_tree_ntfs, &w->ntfs_vol);
684 #else /* WITH_NTFS_3G */
685 WIMLIBAPI int wimlib_add_image_from_ntfs_volume(WIMStruct *w,
688 const char *config_str,
692 ERROR("wimlib was compiled without support for NTFS-3g, so");
693 ERROR("we cannot capture a WIM image directly from a NTFS volume");
694 return WIMLIB_ERR_UNSUPPORTED;
696 #endif /* WITH_NTFS_3G */