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 /* Frees a security descriptor index tree. */
67 static void free_sd_set(struct sd_node *root)
70 free_sd_set(root->left);
71 free_sd_set(root->right);
76 /* Inserts a a new node into the security descriptor index tree. */
77 static void insert_sd_node(struct sd_node *new, struct sd_node *root)
79 int cmp = hashes_cmp(new->hash, root->hash);
82 insert_sd_node(new, root->left);
87 insert_sd_node(new, root->right);
95 /* Returns the security ID of the security data having a SHA1 message digest of
96 * @hash in the security descriptor index tree rooted at @root.
98 * If not found, return -1. */
99 static int lookup_sd(const u8 hash[SHA1_HASH_SIZE], struct sd_node *root)
104 cmp = hashes_cmp(hash, root->hash);
106 return lookup_sd(hash, root->left);
108 return lookup_sd(hash, root->right);
110 return root->security_id;
114 * Adds a security descriptor to the indexed security descriptor set as well as
115 * the corresponding `struct wim_security_data', and returns the new security
116 * ID; or, if there is an existing security descriptor that is the same, return
117 * the security ID for it. If a new security descriptor cannot be allocated,
120 static int sd_set_add_sd(struct sd_set *sd_set, const u8 *descriptor,
123 u8 hash[SHA1_HASH_SIZE];
129 struct wim_security_data *sd;
131 sha1_buffer(descriptor, size, hash);
132 security_id = lookup_sd(hash, sd_set->root);
133 if (security_id >= 0)
136 new = MALLOC(sizeof(*new));
139 descr_copy = MALLOC(size);
145 memcpy(descr_copy, descriptor, size);
146 new->security_id = sd->num_entries;
149 copy_hash(new->hash, hash);
152 descriptors = REALLOC(sd->descriptors,
153 (sd->num_entries + 1) * sizeof(sd->descriptors[0]));
156 sd->descriptors = descriptors;
157 sizes = REALLOC(sd->sizes,
158 (sd->num_entries + 1) * sizeof(sd->sizes[0]));
162 sd->descriptors[sd->num_entries] = descr_copy;
163 sd->sizes[sd->num_entries] = size;
165 sd->total_length += size + sizeof(sd->sizes[0]);
168 insert_sd_node(sd_set->root, new);
171 return new->security_id;
180 static inline ntfschar *attr_record_name(ATTR_RECORD *ar)
182 return (ntfschar*)((u8*)ar + le16_to_cpu(ar->name_offset));
185 /* Calculates the SHA1 message digest of a NTFS attribute.
187 * @ni: The NTFS inode containing the attribute.
188 * @ar: The ATTR_RECORD describing the attribute.
189 * @md: If successful, the returned SHA1 message digest.
191 * Return 0 on success or nonzero on error.
193 static int ntfs_attr_sha1sum(ntfs_inode *ni, ATTR_RECORD *ar,
194 u8 md[SHA1_HASH_SIZE])
202 na = ntfs_attr_open(ni, ar->type, attr_record_name(ar),
205 ERROR_WITH_ERRNO("Failed to open NTFS attribute");
206 return WIMLIB_ERR_NTFS_3G;
209 bytes_remaining = na->data_size;
212 DEBUG("Calculating SHA1 message digest (%"PRIu64" bytes)",
215 while (bytes_remaining) {
216 s64 to_read = min(bytes_remaining, sizeof(buf));
217 if (ntfs_attr_pread(na, pos, to_read, buf) != to_read) {
218 ERROR_WITH_ERRNO("Error reading NTFS attribute");
219 return WIMLIB_ERR_NTFS_3G;
221 sha1_update(&ctx, buf, to_read);
223 bytes_remaining -= to_read;
225 sha1_final(md, &ctx);
230 /* Load the streams from a WIM file or reparse point in the NTFS volume into the
231 * WIM lookup table */
232 static int capture_ntfs_streams(struct dentry *dentry, ntfs_inode *ni,
233 char path[], size_t path_len,
234 struct lookup_table *lookup_table,
235 ntfs_volume **ntfs_vol_p,
239 ntfs_attr_search_ctx *actx;
240 u8 attr_hash[SHA1_HASH_SIZE];
241 struct ntfs_location *ntfs_loc;
242 struct lookup_table_entry *lte;
245 DEBUG("Capturing NTFS data streams from `%s'", path);
247 /* Get context to search the streams of the NTFS file. */
248 actx = ntfs_attr_get_search_ctx(ni, NULL);
250 ERROR_WITH_ERRNO("Cannot get attribute search "
252 return WIMLIB_ERR_NTFS_3G;
255 /* Capture each data stream or reparse data stream. */
256 while (!ntfs_attr_lookup(type, NULL, 0,
257 CASE_SENSITIVE, 0, NULL, 0, actx))
259 char *stream_name_utf8;
260 size_t stream_name_utf16_len;
262 /* Checksum the stream. */
263 ret = ntfs_attr_sha1sum(ni, actx->attr, attr_hash);
267 /* Make a lookup table entry for the stream, or use an existing
268 * one if there's already an identical stream. */
269 lte = __lookup_resource(lookup_table, attr_hash);
270 ret = WIMLIB_ERR_NOMEM;
274 struct ntfs_location *ntfs_loc;
276 ntfs_loc = CALLOC(1, sizeof(*ntfs_loc));
279 ntfs_loc->ntfs_vol_p = ntfs_vol_p;
280 ntfs_loc->path_utf8 = MALLOC(path_len + 1);
281 if (!ntfs_loc->path_utf8)
282 goto out_free_ntfs_loc;
283 memcpy(ntfs_loc->path_utf8, path, path_len + 1);
284 ntfs_loc->stream_name_utf16 = MALLOC(actx->attr->name_length * 2);
285 if (!ntfs_loc->stream_name_utf16)
286 goto out_free_ntfs_loc;
287 memcpy(ntfs_loc->stream_name_utf16,
288 attr_record_name(actx->attr),
289 actx->attr->name_length * 2);
291 ntfs_loc->stream_name_utf16_num_chars = actx->attr->name_length;
292 ntfs_loc->is_reparse_point = (type == AT_REPARSE_POINT);
293 lte = new_lookup_table_entry();
295 goto out_free_ntfs_loc;
296 lte->ntfs_loc = ntfs_loc;
297 lte->resource_location = RESOURCE_IN_NTFS_VOLUME;
298 lte->resource_entry.original_size = actx->attr->data_size;
299 lte->resource_entry.size = actx->attr->data_size;
300 copy_hash(lte->hash, attr_hash);
301 lookup_table_insert(lookup_table, lte);
303 if (actx->attr->name_length == 0) {
305 ERROR("Found two un-named data streams for "
307 ret = WIMLIB_ERR_NTFS_3G;
312 struct ads_entry *new_ads_entry;
313 stream_name_utf8 = utf16_to_utf8((const u8*)attr_record_name(actx->attr),
314 actx->attr->name_length,
315 &stream_name_utf16_len);
316 if (!stream_name_utf8)
318 new_ads_entry = dentry_add_ads(dentry, stream_name_utf8);
319 FREE(stream_name_utf8);
323 new_ads_entry->lte = lte;
329 free_lookup_table_entry(lte);
332 FREE(ntfs_loc->path_utf8);
333 FREE(ntfs_loc->stream_name_utf16);
337 ntfs_attr_put_search_ctx(actx);
339 DEBUG("Successfully captured NTFS streams from `%s'", path);
341 DEBUG("Failed to capture NTFS streams from `%s", path);
346 struct dentry *parent;
350 struct lookup_table *lookup_table;
351 struct sd_set *sd_set;
352 const struct capture_config *config;
353 ntfs_volume **ntfs_vol_p;
356 static int __build_dentry_tree_ntfs(struct dentry **root_p, ntfs_inode *ni,
357 char path[], size_t path_len,
358 struct lookup_table *lookup_table,
359 struct sd_set *sd_set,
360 const struct capture_config *config,
361 ntfs_volume **ntfs_vol_p);
364 static int wim_ntfs_capture_filldir(void *dirent, const ntfschar *name,
365 const int name_len, const int name_type,
366 const s64 pos, const MFT_REF mref,
367 const unsigned dt_type)
369 struct readdir_ctx *ctx;
370 size_t utf8_name_len;
372 struct dentry *child;
378 utf8_name = utf16_to_utf8((const u8*)name, name_len * 2,
383 if (utf8_name[0] == '.' &&
384 (utf8_name[1] == '\0' ||
385 (utf8_name[1] == '.' && utf8_name[2] == '\0'))) {
386 DEBUG("Skipping dentry `%s'", utf8_name);
388 goto out_free_utf8_name;
391 DEBUG("Opening inode for `%s'", utf8_name);
395 ntfs_inode *ni = ntfs_inode_open(ctx->dir_ni->vol, mref);
397 ERROR_WITH_ERRNO("Failed to open NTFS inode");
400 path_len = ctx->path_len;
402 ctx->path[path_len++] = '/';
403 memcpy(ctx->path + path_len, utf8_name, utf8_name_len + 1);
404 path_len += utf8_name_len;
405 ret = __build_dentry_tree_ntfs(&child, ni, ctx->path, path_len,
406 ctx->lookup_table, ctx->sd_set,
407 ctx->config, ctx->ntfs_vol_p);
408 DEBUG("Linking dentry `%s' with parent `%s'",
409 child->file_name_utf8, ctx->parent->file_name_utf8);
411 link_dentry(child, ctx->parent);
412 DEBUG("Return %d", ret);
414 ntfs_inode_close(ni);
421 /* Recursively build a WIM dentry tree corresponding to a NTFS volume.
422 * At the same time, update the WIM lookup table with lookup table entries for
423 * the NTFS streams, and build an array of security descriptors.
425 static int __build_dentry_tree_ntfs(struct dentry **root_p, ntfs_inode *ni,
426 char path[], size_t path_len,
427 struct lookup_table *lookup_table,
428 struct sd_set *sd_set,
429 const struct capture_config *config,
430 ntfs_volume **ntfs_vol_p)
438 if (exclude_path(path, config, false)) {
439 DEBUG("Excluding `%s' from capture", path);
443 DEBUG("Starting recursive capture at path = `%s'", path);
444 mrec_flags = ni->mrec->flags;
445 attributes = ntfs_inode_get_attributes(ni);
447 root = new_dentry(path_basename(path));
449 return WIMLIB_ERR_NOMEM;
451 root->creation_time = le64_to_cpu(ni->creation_time);
452 root->last_write_time = le64_to_cpu(ni->last_data_change_time);
453 root->last_access_time = le64_to_cpu(ni->last_access_time);
454 root->security_id = le32_to_cpu(ni->security_id);
455 root->attributes = le32_to_cpu(attributes);
456 root->hard_link = ni->mft_no;
457 root->resolved = true;
459 if (attributes & FILE_ATTR_REPARSE_POINT) {
460 DEBUG("Reparse point `%s'", path);
461 /* Junction point, symbolic link, or other reparse point */
462 ret = capture_ntfs_streams(root, ni, path, path_len,
463 lookup_table, ntfs_vol_p,
465 } else if (mrec_flags & MFT_RECORD_IS_DIRECTORY) {
466 DEBUG("Directory `%s'", path);
468 /* Normal directory */
470 struct readdir_ctx ctx = {
474 .path_len = path_len,
475 .lookup_table = lookup_table,
478 .ntfs_vol_p = ntfs_vol_p,
480 ret = ntfs_readdir(ni, &pos, &ctx, wim_ntfs_capture_filldir);
482 ret = WIMLIB_ERR_NTFS_3G;
484 DEBUG("Normal file `%s'", path);
486 ret = capture_ntfs_streams(root, ni, path, path_len,
487 lookup_table, ntfs_vol_p,
493 DEBUG("Getting security information from `%s'", path);
494 ret = ntfs_inode_get_security(ni,
495 OWNER_SECURITY_INFORMATION |
496 GROUP_SECURITY_INFORMATION |
497 DACL_SECURITY_INFORMATION |
498 SACL_SECURITY_INFORMATION,
501 ret = ntfs_inode_get_security(ni,
502 OWNER_SECURITY_INFORMATION |
503 GROUP_SECURITY_INFORMATION |
504 DACL_SECURITY_INFORMATION |
505 SACL_SECURITY_INFORMATION,
506 sd, sd_size, &sd_size);
508 ERROR_WITH_ERRNO("Failed to get security information from "
510 ret = WIMLIB_ERR_NTFS_3G;
513 /*print_security_descriptor(sd, sd_size);*/
514 root->security_id = sd_set_add_sd(sd_set, sd, sd_size);
515 DEBUG("Added security ID = %u for `%s'",
516 root->security_id, path);
518 root->security_id = -1;
519 DEBUG("No security ID for `%s'", path);
527 static int build_dentry_tree_ntfs(struct dentry **root_p,
529 struct lookup_table *lookup_table,
530 struct wim_security_data *sd,
531 const struct capture_config *config,
541 ntfs_volume **ntfs_vol_p = extra_arg;
543 DEBUG("Mounting NTFS volume `%s' read-only", device);
545 vol = ntfs_mount(device, MS_RDONLY);
547 ERROR_WITH_ERRNO("Failed to mount NTFS volume `%s' read-only",
549 return WIMLIB_ERR_NTFS_3G;
552 NVolClearShowSysFiles(vol);
554 DEBUG("Opening root NTFS dentry");
555 root_ni = ntfs_inode_open(vol, FILE_root);
557 ERROR_WITH_ERRNO("Failed to open root inode of NTFS volume "
559 ret = WIMLIB_ERR_NTFS_3G;
565 ret = __build_dentry_tree_ntfs(root_p, root_ni, path, 1,
566 lookup_table, &tree, config,
568 ntfs_inode_close(root_ni);
571 if (ntfs_umount(vol, FALSE) != 0) {
572 ERROR_WITH_ERRNO("Failed to unmount NTFS volume `%s'", device);
574 ret = WIMLIB_ERR_NTFS_3G;
581 WIMLIBAPI int wimlib_add_image_from_ntfs_volume(WIMStruct *w,
584 const char *config_str,
588 if (flags & (WIMLIB_ADD_IMAGE_FLAG_DEREFERENCE)) {
589 ERROR("Cannot dereference files when capturing directly from NTFS");
590 return WIMLIB_ERR_INVALID_PARAM;
592 return do_add_image(w, device, name, config_str, config_len, flags,
593 build_dentry_tree_ntfs, &w->ntfs_vol);
596 #else /* WITH_NTFS_3G */
597 WIMLIBAPI int wimlib_add_image_from_ntfs_volume(WIMStruct *w,
600 const char *description,
601 const char *flags_element,
603 const char *config_str,
606 ERROR("wimlib was compiled without support for NTFS-3g, so");
607 ERROR("we cannot capture a WIM image directly from a NTFS volume");
608 return WIMLIB_ERR_UNSUPPORTED;
610 #endif /* WITH_NTFS_3G */