4 * Capture a WIM image from a NTFS volume. We capture everything we can,
5 * including security data and alternate data streams. There should be no loss
10 * Copyright (C) 2012 Eric Biggers
12 * This file is part of wimlib, a library for working with WIM files.
14 * wimlib is free software; you can redistribute it and/or modify it under the
15 * terms of the GNU Lesser General Public License as published by the Free
16 * Software Foundation; either version 2.1 of the License, or (at your option)
19 * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
20 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
21 * A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
24 * You should have received a copy of the GNU Lesser General Public License
25 * along with wimlib; if not, see http://www.gnu.org/licenses/.
29 #include "wimlib_internal.h"
34 #include "lookup_table.h"
36 #include <ntfs-3g/layout.h>
37 #include <ntfs-3g/acls.h>
38 #include <ntfs-3g/attrib.h>
39 #include <ntfs-3g/misc.h>
40 #include <ntfs-3g/reparse.h>
41 #include <ntfs-3g/security.h>
42 #include <ntfs-3g/volume.h>
46 extern int ntfs_inode_get_security(ntfs_inode *ni, u32 selection, char *buf,
47 u32 buflen, u32 *psize);
49 extern int ntfs_inode_get_attributes(ntfs_inode *ni);
51 /* Structure that allows searching the security descriptors by SHA1 message
54 struct wim_security_data *sd;
58 /* Binary tree node of security descriptors, indexed by the @hash field. */
61 u8 hash[SHA1_HASH_SIZE];
63 struct sd_node *right;
66 static void free_sd_tree(struct sd_node *root)
69 free_sd_tree(root->left);
70 free_sd_tree(root->right);
74 /* Frees a security descriptor index set. */
75 static void destroy_sd_set(struct sd_set *sd_set)
77 free_sd_tree(sd_set->root);
80 /* Inserts a a new node into the security descriptor index tree. */
81 static void insert_sd_node(struct sd_node *new, struct sd_node *root)
83 int cmp = hashes_cmp(new->hash, root->hash);
86 insert_sd_node(new, root->left);
91 insert_sd_node(new, root->right);
99 /* Returns the security ID of the security data having a SHA1 message digest of
100 * @hash in the security descriptor index tree rooted at @root.
102 * If not found, return -1. */
103 static int lookup_sd(const u8 hash[SHA1_HASH_SIZE], struct sd_node *root)
108 cmp = hashes_cmp(hash, root->hash);
110 return lookup_sd(hash, root->left);
112 return lookup_sd(hash, root->right);
114 return root->security_id;
118 * Adds a security descriptor to the indexed security descriptor set as well as
119 * the corresponding `struct wim_security_data', and returns the new security
120 * ID; or, if there is an existing security descriptor that is the same, return
121 * the security ID for it. If a new security descriptor cannot be allocated,
124 static int sd_set_add_sd(struct sd_set *sd_set, const char descriptor[],
127 u8 hash[SHA1_HASH_SIZE];
133 struct wim_security_data *sd;
135 sha1_buffer((const u8*)descriptor, size, hash);
137 security_id = lookup_sd(hash, sd_set->root);
138 if (security_id >= 0)
141 new = MALLOC(sizeof(*new));
144 descr_copy = MALLOC(size);
150 memcpy(descr_copy, descriptor, size);
151 new->security_id = sd->num_entries;
154 copy_hash(new->hash, hash);
157 descriptors = REALLOC(sd->descriptors,
158 (sd->num_entries + 1) * sizeof(sd->descriptors[0]));
161 sd->descriptors = descriptors;
162 sizes = REALLOC(sd->sizes,
163 (sd->num_entries + 1) * sizeof(sd->sizes[0]));
167 sd->descriptors[sd->num_entries] = descr_copy;
168 sd->sizes[sd->num_entries] = size;
170 DEBUG("There are now %d security descriptors", sd->num_entries);
171 sd->total_length += size + sizeof(sd->sizes[0]);
174 insert_sd_node(new, sd_set->root);
177 return new->security_id;
186 static inline ntfschar *attr_record_name(ATTR_RECORD *ar)
188 return (ntfschar*)((u8*)ar + le16_to_cpu(ar->name_offset));
191 /* Calculates the SHA1 message digest of a NTFS attribute.
193 * @ni: The NTFS inode containing the attribute.
194 * @ar: The ATTR_RECORD describing the attribute.
195 * @md: If successful, the returned SHA1 message digest.
196 * @reparse_tag_ret: Optional pointer into which the first 4 bytes of the
197 * attribute will be written (to get the reparse
200 * Return 0 on success or nonzero on error.
202 static int ntfs_attr_sha1sum(ntfs_inode *ni, ATTR_RECORD *ar,
203 u8 md[SHA1_HASH_SIZE],
204 u32 *reparse_tag_ret)
212 na = ntfs_attr_open(ni, ar->type, attr_record_name(ar),
215 ERROR_WITH_ERRNO("Failed to open NTFS attribute");
216 return WIMLIB_ERR_NTFS_3G;
219 bytes_remaining = na->data_size;
222 DEBUG2("Calculating SHA1 message digest (%"PRIu64" bytes)",
225 while (bytes_remaining) {
226 s64 to_read = min(bytes_remaining, sizeof(buf));
227 if (ntfs_attr_pread(na, pos, to_read, buf) != to_read) {
228 ERROR_WITH_ERRNO("Error reading NTFS attribute");
229 return WIMLIB_ERR_NTFS_3G;
231 if (bytes_remaining == na->data_size && reparse_tag_ret)
232 *reparse_tag_ret = le32_to_cpu(*(u32*)buf);
233 sha1_update(&ctx, buf, to_read);
235 bytes_remaining -= to_read;
237 sha1_final(md, &ctx);
242 /* Load the streams from a WIM file or reparse point in the NTFS volume into the
243 * WIM lookup table */
244 static int capture_ntfs_streams(struct dentry *dentry, ntfs_inode *ni,
245 char path[], size_t path_len,
246 struct lookup_table *lookup_table,
247 ntfs_volume **ntfs_vol_p,
251 ntfs_attr_search_ctx *actx;
252 u8 attr_hash[SHA1_HASH_SIZE];
253 struct ntfs_location *ntfs_loc = NULL;
255 struct lookup_table_entry *lte;
257 DEBUG2("Capturing NTFS data streams from `%s'", path);
259 /* Get context to search the streams of the NTFS file. */
260 actx = ntfs_attr_get_search_ctx(ni, NULL);
262 ERROR_WITH_ERRNO("Cannot get NTFS attribute search "
264 return WIMLIB_ERR_NTFS_3G;
267 /* Capture each data stream or reparse data stream. */
268 while (!ntfs_attr_lookup(type, NULL, 0,
269 CASE_SENSITIVE, 0, NULL, 0, actx))
271 char *stream_name_utf8;
272 size_t stream_name_utf16_len;
274 u64 data_size = ntfs_get_attribute_value_length(actx->attr);
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);
311 ntfs_loc->stream_name_utf16 = MALLOC(actx->attr->name_length * 2);
312 if (!ntfs_loc->stream_name_utf16)
313 goto out_free_ntfs_loc;
314 memcpy(ntfs_loc->stream_name_utf16,
315 attr_record_name(actx->attr),
316 actx->attr->name_length * 2);
318 ntfs_loc->stream_name_utf16_num_chars = actx->attr->name_length;
319 lte = new_lookup_table_entry();
321 goto out_free_ntfs_loc;
322 lte->ntfs_loc = ntfs_loc;
323 lte->resource_location = RESOURCE_IN_NTFS_VOLUME;
324 if (type == AT_REPARSE_POINT) {
325 dentry->reparse_tag = reparse_tag;
326 ntfs_loc->is_reparse_point = true;
327 lte->resource_entry.original_size = data_size - 8;
328 lte->resource_entry.size = data_size - 8;
330 ntfs_loc->is_reparse_point = false;
331 lte->resource_entry.original_size = data_size;
332 lte->resource_entry.size = data_size;
335 DEBUG("Add resource for `%s' (size = %zu)",
336 dentry->file_name_utf8,
337 lte->resource_entry.original_size);
338 copy_hash(lte->hash, attr_hash);
339 lookup_table_insert(lookup_table, lte);
342 if (actx->attr->name_length == 0) {
343 /* Unnamed data stream. Put the reference to it in the
346 ERROR("Found two un-named data streams for "
348 ret = WIMLIB_ERR_NTFS_3G;
353 /* Named data stream. Put the reference to it in the
354 * alternate data stream entries */
355 struct ads_entry *new_ads_entry;
356 size_t stream_name_utf8_len;
357 stream_name_utf8 = utf16_to_utf8((const char*)attr_record_name(actx->attr),
358 actx->attr->name_length,
359 &stream_name_utf8_len);
360 if (!stream_name_utf8)
362 new_ads_entry = dentry_add_ads(dentry, stream_name_utf8);
363 FREE(stream_name_utf8);
367 new_ads_entry->lte = lte;
373 free_lookup_table_entry(lte);
376 FREE(ntfs_loc->path_utf8);
377 FREE(ntfs_loc->stream_name_utf16);
381 ntfs_attr_put_search_ctx(actx);
383 DEBUG2("Successfully captured NTFS streams from `%s'", path);
385 ERROR("Failed to capture NTFS streams from `%s", path);
390 struct dentry *parent;
394 struct lookup_table *lookup_table;
395 struct sd_set *sd_set;
396 const struct capture_config *config;
397 ntfs_volume **ntfs_vol_p;
402 build_dentry_tree_ntfs_recursive(struct dentry **root_p, ntfs_inode *ni,
403 char path[], size_t path_len,
404 struct lookup_table *lookup_table,
405 struct sd_set *sd_set,
406 const struct capture_config *config,
407 ntfs_volume **ntfs_vol_p,
410 static int wim_ntfs_capture_filldir(void *dirent, const ntfschar *name,
411 const int name_len, const int name_type,
412 const s64 pos, const MFT_REF mref,
413 const unsigned dt_type)
415 struct readdir_ctx *ctx;
416 size_t utf8_name_len;
418 struct dentry *child = NULL;
422 if (name_type == FILE_NAME_DOS)
427 utf8_name = utf16_to_utf8((const char*)name, name_len * 2,
432 if (utf8_name[0] == '.' &&
433 (utf8_name[1] == '\0' ||
434 (utf8_name[1] == '.' && utf8_name[2] == '\0'))) {
436 goto out_free_utf8_name;
441 ntfs_inode *ni = ntfs_inode_open(ctx->dir_ni->vol, mref);
443 ERROR_WITH_ERRNO("Failed to open NTFS inode");
446 path_len = ctx->path_len;
448 ctx->path[path_len++] = '/';
449 memcpy(ctx->path + path_len, utf8_name, utf8_name_len + 1);
450 path_len += utf8_name_len;
451 ret = build_dentry_tree_ntfs_recursive(&child, ni, ctx->path, path_len,
452 ctx->lookup_table, ctx->sd_set,
453 ctx->config, ctx->ntfs_vol_p,
457 link_dentry(child, ctx->parent);
459 ntfs_inode_close(ni);
466 /* Recursively build a WIM dentry tree corresponding to a NTFS volume.
467 * At the same time, update the WIM lookup table with lookup table entries for
468 * the NTFS streams, and build an array of security descriptors.
470 static int build_dentry_tree_ntfs_recursive(struct dentry **root_p,
474 struct lookup_table *lookup_table,
475 struct sd_set *sd_set,
476 const struct capture_config *config,
477 ntfs_volume **ntfs_vol_p,
486 mrec_flags = ni->mrec->flags;
487 attributes = ntfs_inode_get_attributes(ni);
489 if (exclude_path(path, config, false)) {
490 if (flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE) {
491 const char *file_type;
492 if (attributes & MFT_RECORD_IS_DIRECTORY)
493 file_type = "directory";
496 printf("Excluding %s `%s' from capture\n",
503 if (flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE)
504 printf("Scanning `%s'\n", path);
506 root = new_dentry(path_basename(path));
508 return WIMLIB_ERR_NOMEM;
511 root->creation_time = le64_to_cpu(ni->creation_time);
512 root->last_write_time = le64_to_cpu(ni->last_data_change_time);
513 root->last_access_time = le64_to_cpu(ni->last_access_time);
514 root->attributes = le32_to_cpu(attributes);
515 root->link_group_id = ni->mft_no;
516 root->resolved = true;
518 if (attributes & FILE_ATTR_REPARSE_POINT) {
519 /* Junction point, symbolic link, or other reparse point */
520 ret = capture_ntfs_streams(root, ni, path, path_len,
521 lookup_table, ntfs_vol_p,
523 } else if (mrec_flags & MFT_RECORD_IS_DIRECTORY) {
525 /* Normal directory */
527 struct readdir_ctx ctx = {
531 .path_len = path_len,
532 .lookup_table = lookup_table,
535 .ntfs_vol_p = ntfs_vol_p,
538 ret = ntfs_readdir(ni, &pos, &ctx, wim_ntfs_capture_filldir);
540 ERROR_WITH_ERRNO("ntfs_readdir()");
541 ret = WIMLIB_ERR_NTFS_3G;
545 ret = capture_ntfs_streams(root, ni, path, path_len,
546 lookup_table, ntfs_vol_p,
552 ret = ntfs_inode_get_security(ni,
553 OWNER_SECURITY_INFORMATION |
554 GROUP_SECURITY_INFORMATION |
555 DACL_SECURITY_INFORMATION |
556 SACL_SECURITY_INFORMATION,
559 ret = ntfs_inode_get_security(ni,
560 OWNER_SECURITY_INFORMATION |
561 GROUP_SECURITY_INFORMATION |
562 DACL_SECURITY_INFORMATION |
563 SACL_SECURITY_INFORMATION,
564 sd, sd_size, &sd_size);
566 ERROR_WITH_ERRNO("Failed to get security information from "
568 ret = WIMLIB_ERR_NTFS_3G;
571 /*print_security_descriptor(sd, sd_size);*/
572 root->security_id = sd_set_add_sd(sd_set, sd, sd_size);
573 if (root->security_id == -1) {
574 ERROR("Out of memory");
575 return WIMLIB_ERR_NOMEM;
577 DEBUG("Added security ID = %u for `%s'",
578 root->security_id, path);
580 root->security_id = -1;
581 DEBUG("No security ID for `%s'", path);
588 static int build_dentry_tree_ntfs(struct dentry **root_p,
590 struct lookup_table *lookup_table,
591 struct wim_security_data *sd,
592 const struct capture_config *config,
599 struct sd_set sd_set = {
603 ntfs_volume **ntfs_vol_p = extra_arg;
605 DEBUG("Mounting NTFS volume `%s' read-only", device);
607 vol = ntfs_mount(device, MS_RDONLY);
609 ERROR_WITH_ERRNO("Failed to mount NTFS volume `%s' read-only",
611 return WIMLIB_ERR_NTFS_3G;
613 ntfs_open_secure(vol);
615 /* We don't want to capture the special NTFS files such as $Bitmap. Not
616 * to be confused with "hidden" or "system" files which are real files
617 * that we do need to capture. */
618 NVolClearShowSysFiles(vol);
620 DEBUG("Opening root NTFS dentry");
621 root_ni = ntfs_inode_open(vol, FILE_root);
623 ERROR_WITH_ERRNO("Failed to open root inode of NTFS volume "
625 ret = WIMLIB_ERR_NTFS_3G;
629 /* Currently we assume that all the UTF-8 paths fit into this length and
630 * there is no check for overflow. */
631 char *path = MALLOC(32768);
633 ERROR("Could not allocate memory for NTFS pathname");
639 ret = build_dentry_tree_ntfs_recursive(root_p, root_ni, path, 1,
640 lookup_table, &sd_set,
641 config, ntfs_vol_p, flags);
644 ntfs_inode_close(root_ni);
645 destroy_sd_set(&sd_set);
649 if (ntfs_umount(vol, FALSE) != 0) {
650 ERROR_WITH_ERRNO("Failed to unmount NTFS volume `%s'",
653 ret = WIMLIB_ERR_NTFS_3G;
656 /* We need to leave the NTFS volume mounted so that we can read
657 * the NTFS files again when we are actually writing the WIM */
665 WIMLIBAPI int wimlib_add_image_from_ntfs_volume(WIMStruct *w,
668 const char *config_str,
672 if (flags & (WIMLIB_ADD_IMAGE_FLAG_DEREFERENCE)) {
673 ERROR("Cannot dereference files when capturing directly from NTFS");
674 return WIMLIB_ERR_INVALID_PARAM;
676 return do_add_image(w, device, name, config_str, config_len, flags,
677 build_dentry_tree_ntfs, &w->ntfs_vol);
680 #else /* WITH_NTFS_3G */
681 WIMLIBAPI int wimlib_add_image_from_ntfs_volume(WIMStruct *w,
684 const char *config_str,
688 ERROR("wimlib was compiled without support for NTFS-3g, so");
689 ERROR("we cannot capture a WIM image directly from a NTFS volume");
690 return WIMLIB_ERR_UNSUPPORTED;
692 #endif /* WITH_NTFS_3G */