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/.
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>
46 #ifndef WITH_NEW_NTFS_3G
47 extern int ntfs_get_inode_security(ntfs_inode *ni, u32 selection, char *buf,
48 u32 buflen, u32 *psize);
50 extern u32 ntfs_get_inode_attributes(ntfs_inode *ni);
53 /* Structure that allows searching the security descriptors by SHA1 message
56 struct wim_security_data *sd;
60 /* Binary tree node of security descriptors, indexed by the @hash field. */
63 u8 hash[SHA1_HASH_SIZE];
65 struct sd_node *right;
68 static void free_sd_tree(struct sd_node *root)
71 free_sd_tree(root->left);
72 free_sd_tree(root->right);
76 /* Frees a security descriptor index set. */
77 static void destroy_sd_set(struct sd_set *sd_set)
79 free_sd_tree(sd_set->root);
82 /* Inserts a a new node into the security descriptor index tree. */
83 static void insert_sd_node(struct sd_node *new, struct sd_node *root)
85 int cmp = hashes_cmp(new->hash, root->hash);
88 insert_sd_node(new, root->left);
93 insert_sd_node(new, root->right);
101 /* Returns the security ID of the security data having a SHA1 message digest of
102 * @hash in the security descriptor index tree rooted at @root.
104 * If not found, return -1. */
105 static int lookup_sd(const u8 hash[SHA1_HASH_SIZE], struct sd_node *root)
110 cmp = hashes_cmp(hash, root->hash);
112 return lookup_sd(hash, root->left);
114 return lookup_sd(hash, root->right);
116 return root->security_id;
120 * Adds a security descriptor to the indexed security descriptor set as well as
121 * the corresponding `struct wim_security_data', and returns the new security
122 * ID; or, if there is an existing security descriptor that is the same, return
123 * the security ID for it. If a new security descriptor cannot be allocated,
126 static int sd_set_add_sd(struct sd_set *sd_set, const char descriptor[],
129 u8 hash[SHA1_HASH_SIZE];
135 struct wim_security_data *sd;
137 sha1_buffer((const u8*)descriptor, size, hash);
139 security_id = lookup_sd(hash, sd_set->root);
140 if (security_id >= 0)
143 new = MALLOC(sizeof(*new));
146 descr_copy = MALLOC(size);
152 memcpy(descr_copy, descriptor, size);
153 new->security_id = sd->num_entries;
156 copy_hash(new->hash, hash);
159 descriptors = REALLOC(sd->descriptors,
160 (sd->num_entries + 1) * sizeof(sd->descriptors[0]));
163 sd->descriptors = descriptors;
164 sizes = REALLOC(sd->sizes,
165 (sd->num_entries + 1) * sizeof(sd->sizes[0]));
169 sd->descriptors[sd->num_entries] = descr_copy;
170 sd->sizes[sd->num_entries] = size;
172 DEBUG("There are now %d security descriptors", sd->num_entries);
173 sd->total_length += size + sizeof(sd->sizes[0]);
176 insert_sd_node(new, sd_set->root);
179 return new->security_id;
188 static inline ntfschar *attr_record_name(ATTR_RECORD *ar)
190 return (ntfschar*)((u8*)ar + le16_to_cpu(ar->name_offset));
193 /* Calculates the SHA1 message digest of a NTFS attribute.
195 * @ni: The NTFS inode containing the attribute.
196 * @ar: The ATTR_RECORD describing the attribute.
197 * @md: If successful, the returned SHA1 message digest.
198 * @reparse_tag_ret: Optional pointer into which the first 4 bytes of the
199 * attribute will be written (to get the reparse
202 * Return 0 on success or nonzero on error.
204 static int ntfs_attr_sha1sum(ntfs_inode *ni, ATTR_RECORD *ar,
205 u8 md[SHA1_HASH_SIZE],
206 u32 *reparse_tag_ret)
214 na = ntfs_attr_open(ni, ar->type, attr_record_name(ar),
217 ERROR_WITH_ERRNO("Failed to open NTFS attribute");
218 return WIMLIB_ERR_NTFS_3G;
221 bytes_remaining = na->data_size;
224 DEBUG2("Calculating SHA1 message digest (%"PRIu64" bytes)",
227 while (bytes_remaining) {
228 s64 to_read = min(bytes_remaining, sizeof(buf));
229 if (ntfs_attr_pread(na, pos, to_read, buf) != to_read) {
230 ERROR_WITH_ERRNO("Error reading NTFS attribute");
231 return WIMLIB_ERR_NTFS_3G;
233 if (bytes_remaining == na->data_size && reparse_tag_ret)
234 *reparse_tag_ret = le32_to_cpu(*(u32*)buf);
235 sha1_update(&ctx, buf, to_read);
237 bytes_remaining -= to_read;
239 sha1_final(md, &ctx);
244 /* Load the streams from a WIM file or reparse point in the NTFS volume into the
245 * WIM lookup table */
246 static int capture_ntfs_streams(struct dentry *dentry, ntfs_inode *ni,
247 char path[], size_t path_len,
248 struct lookup_table *lookup_table,
249 ntfs_volume **ntfs_vol_p,
253 ntfs_attr_search_ctx *actx;
254 u8 attr_hash[SHA1_HASH_SIZE];
255 struct ntfs_location *ntfs_loc = NULL;
257 struct lookup_table_entry *lte;
259 DEBUG2("Capturing NTFS data streams from `%s'", path);
261 /* Get context to search the streams of the NTFS file. */
262 actx = ntfs_attr_get_search_ctx(ni, NULL);
264 ERROR_WITH_ERRNO("Cannot get NTFS attribute search "
266 return WIMLIB_ERR_NTFS_3G;
269 /* Capture each data stream or reparse data stream. */
270 while (!ntfs_attr_lookup(type, NULL, 0,
271 CASE_SENSITIVE, 0, NULL, 0, actx))
273 char *stream_name_utf8;
274 size_t stream_name_utf16_len;
276 u64 data_size = ntfs_get_attribute_value_length(actx->attr);
277 u64 name_length = actx->attr->name_length;
279 if (data_size == 0) {
281 ERROR_WITH_ERRNO("Failed to get size of attribute of "
283 ret = WIMLIB_ERR_NTFS_3G;
286 /* Empty stream. No lookup table entry is needed. */
289 if (type == AT_REPARSE_POINT && data_size < 8) {
290 ERROR("`%s': reparse point buffer too small");
291 ret = WIMLIB_ERR_NTFS_3G;
294 /* Checksum the stream. */
295 ret = ntfs_attr_sha1sum(ni, actx->attr, attr_hash, &reparse_tag);
299 /* Make a lookup table entry for the stream, or use an existing
300 * one if there's already an identical stream. */
301 lte = __lookup_resource(lookup_table, attr_hash);
302 ret = WIMLIB_ERR_NOMEM;
306 ntfs_loc = CALLOC(1, sizeof(*ntfs_loc));
309 ntfs_loc->ntfs_vol_p = ntfs_vol_p;
310 ntfs_loc->path_utf8 = MALLOC(path_len + 1);
311 if (!ntfs_loc->path_utf8)
312 goto out_free_ntfs_loc;
313 memcpy(ntfs_loc->path_utf8, path, path_len + 1);
315 ntfs_loc->stream_name_utf16 = MALLOC(name_length * 2);
316 if (!ntfs_loc->stream_name_utf16)
317 goto out_free_ntfs_loc;
318 memcpy(ntfs_loc->stream_name_utf16,
319 attr_record_name(actx->attr),
320 actx->attr->name_length * 2);
321 ntfs_loc->stream_name_utf16_num_chars = name_length;
324 lte = new_lookup_table_entry();
326 goto out_free_ntfs_loc;
327 lte->ntfs_loc = ntfs_loc;
328 lte->resource_location = RESOURCE_IN_NTFS_VOLUME;
329 if (type == AT_REPARSE_POINT) {
330 dentry->reparse_tag = reparse_tag;
331 ntfs_loc->is_reparse_point = true;
332 lte->resource_entry.original_size = data_size - 8;
333 lte->resource_entry.size = data_size - 8;
335 ntfs_loc->is_reparse_point = false;
336 lte->resource_entry.original_size = data_size;
337 lte->resource_entry.size = data_size;
340 DEBUG("Add resource for `%s' (size = %zu)",
341 dentry->file_name_utf8,
342 lte->resource_entry.original_size);
343 copy_hash(lte->hash, attr_hash);
344 lookup_table_insert(lookup_table, lte);
347 if (name_length == 0) {
348 /* Unnamed data stream. Put the reference to it in the
351 ERROR("Found two un-named data streams for "
353 ret = WIMLIB_ERR_NTFS_3G;
358 /* Named data stream. Put the reference to it in the
359 * alternate data stream entries */
360 struct ads_entry *new_ads_entry;
361 size_t stream_name_utf8_len;
362 stream_name_utf8 = utf16_to_utf8((const char*)attr_record_name(actx->attr),
364 &stream_name_utf8_len);
365 if (!stream_name_utf8)
367 new_ads_entry = dentry_add_ads(dentry, stream_name_utf8);
368 FREE(stream_name_utf8);
372 wimlib_assert(new_ads_entry->stream_name_len == name_length * 2);
374 new_ads_entry->lte = lte;
380 free_lookup_table_entry(lte);
383 FREE(ntfs_loc->path_utf8);
384 FREE(ntfs_loc->stream_name_utf16);
388 ntfs_attr_put_search_ctx(actx);
390 DEBUG2("Successfully captured NTFS streams from `%s'", path);
392 ERROR("Failed to capture NTFS streams from `%s", path);
397 struct dentry *parent;
401 struct lookup_table *lookup_table;
402 struct sd_set *sd_set;
403 const struct capture_config *config;
404 ntfs_volume **ntfs_vol_p;
409 build_dentry_tree_ntfs_recursive(struct dentry **root_p, ntfs_inode *ni,
410 char path[], size_t path_len,
411 struct lookup_table *lookup_table,
412 struct sd_set *sd_set,
413 const struct capture_config *config,
414 ntfs_volume **ntfs_vol_p,
417 static int wim_ntfs_capture_filldir(void *dirent, const ntfschar *name,
418 const int name_len, const int name_type,
419 const s64 pos, const MFT_REF mref,
420 const unsigned dt_type)
422 struct readdir_ctx *ctx;
423 size_t utf8_name_len;
425 struct dentry *child = NULL;
429 if (name_type == FILE_NAME_DOS)
434 utf8_name = utf16_to_utf8((const char*)name, name_len * 2,
439 if (utf8_name[0] == '.' &&
440 (utf8_name[1] == '\0' ||
441 (utf8_name[1] == '.' && utf8_name[2] == '\0'))) {
443 goto out_free_utf8_name;
448 ntfs_inode *ni = ntfs_inode_open(ctx->dir_ni->vol, mref);
450 ERROR_WITH_ERRNO("Failed to open NTFS inode");
453 path_len = ctx->path_len;
455 ctx->path[path_len++] = '/';
456 memcpy(ctx->path + path_len, utf8_name, utf8_name_len + 1);
457 path_len += utf8_name_len;
458 ret = build_dentry_tree_ntfs_recursive(&child, ni, ctx->path, path_len,
459 ctx->lookup_table, ctx->sd_set,
460 ctx->config, ctx->ntfs_vol_p,
464 link_dentry(child, ctx->parent);
466 ntfs_inode_close(ni);
473 /* Recursively build a WIM dentry tree corresponding to a NTFS volume.
474 * At the same time, update the WIM lookup table with lookup table entries for
475 * the NTFS streams, and build an array of security descriptors.
477 static int build_dentry_tree_ntfs_recursive(struct dentry **root_p,
481 struct lookup_table *lookup_table,
482 struct sd_set *sd_set,
483 const struct capture_config *config,
484 ntfs_volume **ntfs_vol_p,
493 mrec_flags = ni->mrec->flags;
494 attributes = ntfs_get_inode_attributes(ni);
496 if (exclude_path(path, config, false)) {
497 if (flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE) {
498 const char *file_type;
499 if (attributes & MFT_RECORD_IS_DIRECTORY)
500 file_type = "directory";
503 printf("Excluding %s `%s' from capture\n",
510 if (flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE)
511 printf("Scanning `%s'\n", path);
513 root = new_dentry(path_basename(path));
515 return WIMLIB_ERR_NOMEM;
518 root->creation_time = le64_to_cpu(ni->creation_time);
519 root->last_write_time = le64_to_cpu(ni->last_data_change_time);
520 root->last_access_time = le64_to_cpu(ni->last_access_time);
521 root->attributes = le32_to_cpu(attributes);
522 root->link_group_id = ni->mft_no;
523 root->resolved = true;
525 if (attributes & FILE_ATTR_REPARSE_POINT) {
526 /* Junction point, symbolic link, or other reparse point */
527 ret = capture_ntfs_streams(root, ni, path, path_len,
528 lookup_table, ntfs_vol_p,
530 } else if (mrec_flags & MFT_RECORD_IS_DIRECTORY) {
532 /* Normal directory */
534 struct readdir_ctx ctx = {
538 .path_len = path_len,
539 .lookup_table = lookup_table,
542 .ntfs_vol_p = ntfs_vol_p,
545 ret = ntfs_readdir(ni, &pos, &ctx, wim_ntfs_capture_filldir);
547 ERROR_WITH_ERRNO("ntfs_readdir()");
548 ret = WIMLIB_ERR_NTFS_3G;
552 ret = capture_ntfs_streams(root, ni, path, path_len,
553 lookup_table, ntfs_vol_p,
559 ret = ntfs_get_inode_security(ni,
560 OWNER_SECURITY_INFORMATION |
561 GROUP_SECURITY_INFORMATION |
562 DACL_SECURITY_INFORMATION |
563 SACL_SECURITY_INFORMATION,
566 ret = ntfs_get_inode_security(ni,
567 OWNER_SECURITY_INFORMATION |
568 GROUP_SECURITY_INFORMATION |
569 DACL_SECURITY_INFORMATION |
570 SACL_SECURITY_INFORMATION,
571 sd, sd_size, &sd_size);
573 ERROR_WITH_ERRNO("Failed to get security information from "
575 ret = WIMLIB_ERR_NTFS_3G;
578 /*print_security_descriptor(sd, sd_size);*/
579 root->security_id = sd_set_add_sd(sd_set, sd, sd_size);
580 if (root->security_id == -1) {
581 ERROR("Out of memory");
582 return WIMLIB_ERR_NOMEM;
584 DEBUG("Added security ID = %u for `%s'",
585 root->security_id, path);
587 root->security_id = -1;
588 DEBUG("No security ID for `%s'", path);
595 static int build_dentry_tree_ntfs(struct dentry **root_p,
597 struct lookup_table *lookup_table,
598 struct wim_security_data *sd,
599 const struct capture_config *config,
606 struct sd_set sd_set = {
610 ntfs_volume **ntfs_vol_p = extra_arg;
612 DEBUG("Mounting NTFS volume `%s' read-only", device);
614 vol = ntfs_mount(device, MS_RDONLY);
616 ERROR_WITH_ERRNO("Failed to mount NTFS volume `%s' read-only",
618 return WIMLIB_ERR_NTFS_3G;
620 ntfs_open_secure(vol);
622 /* We don't want to capture the special NTFS files such as $Bitmap. Not
623 * to be confused with "hidden" or "system" files which are real files
624 * that we do need to capture. */
625 NVolClearShowSysFiles(vol);
627 DEBUG("Opening root NTFS dentry");
628 root_ni = ntfs_inode_open(vol, FILE_root);
630 ERROR_WITH_ERRNO("Failed to open root inode of NTFS volume "
632 ret = WIMLIB_ERR_NTFS_3G;
636 /* Currently we assume that all the UTF-8 paths fit into this length and
637 * there is no check for overflow. */
638 char *path = MALLOC(32768);
640 ERROR("Could not allocate memory for NTFS pathname");
646 ret = build_dentry_tree_ntfs_recursive(root_p, root_ni, path, 1,
647 lookup_table, &sd_set,
648 config, ntfs_vol_p, flags);
651 ntfs_inode_close(root_ni);
652 destroy_sd_set(&sd_set);
656 if (ntfs_umount(vol, FALSE) != 0) {
657 ERROR_WITH_ERRNO("Failed to unmount NTFS volume `%s'",
660 ret = WIMLIB_ERR_NTFS_3G;
663 /* We need to leave the NTFS volume mounted so that we can read
664 * the NTFS files again when we are actually writing the WIM */
672 WIMLIBAPI int wimlib_add_image_from_ntfs_volume(WIMStruct *w,
675 const char *config_str,
679 if (flags & (WIMLIB_ADD_IMAGE_FLAG_DEREFERENCE)) {
680 ERROR("Cannot dereference files when capturing directly from NTFS");
681 return WIMLIB_ERR_INVALID_PARAM;
683 return do_add_image(w, device, name, config_str, config_len, flags,
684 build_dentry_tree_ntfs, &w->ntfs_vol);
687 #else /* WITH_NTFS_3G */
688 WIMLIBAPI int wimlib_add_image_from_ntfs_volume(WIMStruct *w,
691 const char *config_str,
695 ERROR("wimlib was compiled without support for NTFS-3g, so");
696 ERROR("we cannot capture a WIM image directly from a NTFS volume");
697 return WIMLIB_ERR_UNSUPPORTED;
699 #endif /* WITH_NTFS_3G */