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, 2013 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/.
30 #include <ntfs-3g/endians.h>
31 #include <ntfs-3g/types.h>
33 #include "buffer_io.h"
35 #include "lookup_table.h"
37 #include "wimlib_internal.h"
39 #include <ntfs-3g/layout.h>
40 #include <ntfs-3g/acls.h>
41 #include <ntfs-3g/attrib.h>
42 #include <ntfs-3g/misc.h>
43 #include <ntfs-3g/reparse.h>
44 #include <ntfs-3g/security.h> /* ntfs-3g/security.h before ntfs-3g/xattrs.h */
45 #include <ntfs-3g/xattrs.h>
46 #include <ntfs-3g/volume.h>
55 static inline ntfschar *
56 attr_record_name(ATTR_RECORD *ar)
58 return (ntfschar*)((u8*)ar + le16_to_cpu(ar->name_offset));
62 open_ntfs_attr(ntfs_inode *ni, struct ntfs_location *loc)
66 na = ntfs_attr_open(ni,
67 loc->is_reparse_point ? AT_REPARSE_POINT : AT_DATA,
69 loc->stream_name_nchars);
71 ERROR_WITH_ERRNO("Failed to open attribute of \"%"TS"\" in "
72 "NTFS volume", loc->path);
78 read_ntfs_file_prefix(const struct wim_lookup_table_entry *lte,
80 consume_data_callback_t cb,
84 struct ntfs_location *loc = lte->ntfs_loc;
85 ntfs_volume *vol = loc->ntfs_vol;
93 ni = ntfs_pathname_to_inode(vol, NULL, loc->path);
95 ERROR_WITH_ERRNO("Can't find NTFS inode for \"%"TS"\"", loc->path);
96 ret = WIMLIB_ERR_NTFS_3G;
100 na = open_ntfs_attr(ni, loc);
102 ret = WIMLIB_ERR_NTFS_3G;
103 goto out_close_ntfs_inode;
107 out_buf = alloca(WIM_CHUNK_SIZE);
109 out_buf = ctx_or_buf;
110 pos = (loc->is_reparse_point) ? 8 : 0;
111 bytes_remaining = size;
112 while (bytes_remaining) {
113 s64 to_read = min(bytes_remaining, WIM_CHUNK_SIZE);
114 if (ntfs_attr_pread(na, pos, to_read, out_buf) != to_read) {
115 ERROR_WITH_ERRNO("Error reading \"%"TS"\"", loc->path);
116 ret = WIMLIB_ERR_NTFS_3G;
117 goto out_close_ntfs_attr;
120 bytes_remaining -= to_read;
122 ret = cb(out_buf, to_read, ctx_or_buf);
124 goto out_close_ntfs_attr;
132 out_close_ntfs_inode:
133 ntfs_inode_close(ni);
139 read_reparse_tag(ntfs_inode *ni, struct ntfs_location *loc,
140 u32 *reparse_tag_ret)
146 na = open_ntfs_attr(ni, loc);
148 ret = WIMLIB_ERR_NTFS_3G;
152 if (ntfs_attr_pread(na, 0, 8, buf) != 8) {
153 ERROR_WITH_ERRNO("Error reading reparse data");
154 ret = WIMLIB_ERR_NTFS_3G;
155 goto out_close_ntfs_attr;
157 *reparse_tag_ret = le32_to_cpu(*(u32*)buf);
158 DEBUG("ReparseTag = %#x", *reparse_tag_ret);
167 /* Load the streams from a file or reparse point in the NTFS volume into the WIM
170 capture_ntfs_streams(struct wim_inode *inode,
174 struct wim_lookup_table *lookup_table,
178 ntfs_attr_search_ctx *actx;
179 struct ntfs_location *ntfs_loc;
181 struct wim_lookup_table_entry *lte;
183 DEBUG2("Capturing NTFS data streams from `%s'", path);
185 /* Get context to search the streams of the NTFS file. */
186 actx = ntfs_attr_get_search_ctx(ni, NULL);
188 ERROR_WITH_ERRNO("Cannot get NTFS attribute search "
190 return WIMLIB_ERR_NTFS_3G;
193 /* Capture each data stream or reparse data stream. */
194 while (!ntfs_attr_lookup(type, NULL, 0,
195 CASE_SENSITIVE, 0, NULL, 0, actx))
197 u64 data_size = ntfs_get_attribute_value_length(actx->attr);
198 u64 name_length = actx->attr->name_length;
201 if (data_size == 0) {
202 /* Empty stream. No lookup table entry is needed. */
206 ntfs_loc = CALLOC(1, sizeof(*ntfs_loc));
209 ntfs_loc->ntfs_vol = vol;
210 ntfs_loc->path = MALLOC(path_len + 1);
212 goto out_free_ntfs_loc;
213 memcpy(ntfs_loc->path, path, path_len + 1);
215 ntfs_loc->stream_name = MALLOC(name_length * 2);
216 if (!ntfs_loc->stream_name)
217 goto out_free_ntfs_loc;
218 memcpy(ntfs_loc->stream_name,
219 attr_record_name(actx->attr),
220 actx->attr->name_length * 2);
221 ntfs_loc->stream_name_nchars = name_length;
224 lte = new_lookup_table_entry();
226 goto out_free_ntfs_loc;
227 lte->resource_location = RESOURCE_IN_NTFS_VOLUME;
228 lte->ntfs_loc = ntfs_loc;
230 if (type == AT_REPARSE_POINT) {
232 ERROR("Invalid reparse data (only %u bytes)!",
233 (unsigned)data_size);
234 ret = WIMLIB_ERR_NTFS_3G;
237 lte->ntfs_loc->is_reparse_point = true;
238 lte->resource_entry.original_size = data_size - 8;
239 ret = read_reparse_tag(ni, lte->ntfs_loc,
240 &inode->i_reparse_tag);
244 lte->ntfs_loc->is_reparse_point = false;
245 lte->resource_entry.original_size = data_size;
248 if (name_length == 0) {
249 /* Unnamed data stream. Put the reference to it in the
252 ERROR("Found two un-named data streams for `%s'",
254 ret = WIMLIB_ERR_NTFS_3G;
260 /* Named data stream. Put the reference to it in the
261 * alternate data stream entries */
262 struct wim_ads_entry *new_ads_entry;
264 new_ads_entry = inode_add_ads_utf16le(inode,
265 attr_record_name(actx->attr),
269 wimlib_assert(new_ads_entry->stream_name_nbytes == name_length * 2);
270 stream_id = new_ads_entry->stream_id;
271 new_ads_entry->lte = lte;
274 lookup_table_insert_unhashed(lookup_table, lte,
278 if (errno == ENOENT) {
281 ERROR_WITH_ERRNO("Error listing NTFS attributes from `%s'",
283 ret = WIMLIB_ERR_NTFS_3G;
287 free_lookup_table_entry(lte);
290 FREE(ntfs_loc->path);
291 FREE(ntfs_loc->stream_name);
295 ntfs_attr_put_search_ctx(actx);
297 DEBUG2("Successfully captured NTFS streams from `%s'", path);
299 ERROR("Failed to capture NTFS streams from `%s'", path);
303 /* Red-black tree that maps NTFS inode numbers to DOS names */
304 struct dos_name_map {
305 struct rb_root rb_root;
308 struct dos_name_node {
309 struct rb_node rb_node;
315 /* Inserts a new DOS name into the map */
317 insert_dos_name(struct dos_name_map *map, const ntfschar *dos_name,
318 size_t name_nbytes, u64 ntfs_ino)
320 struct dos_name_node *new_node;
322 struct rb_root *root;
323 struct rb_node *rb_parent;
325 DEBUG("DOS name_len = %zu", name_nbytes);
326 new_node = MALLOC(sizeof(struct dos_name_node));
330 /* DOS names are supposed to be 12 characters max (that's 24 bytes,
331 * assuming 2-byte ntfs characters) */
332 wimlib_assert(name_nbytes <= sizeof(new_node->dos_name));
334 /* Initialize the DOS name, DOS name length, and NTFS inode number of
335 * the red-black tree node */
336 memcpy(new_node->dos_name, dos_name, name_nbytes);
337 new_node->name_nbytes = name_nbytes;
338 new_node->ntfs_ino = ntfs_ino;
340 /* Insert the red-black tree node */
341 root = &map->rb_root;
345 struct dos_name_node *this;
347 this = container_of(*p, struct dos_name_node, rb_node);
349 if (new_node->ntfs_ino < this->ntfs_ino)
350 p = &((*p)->rb_left);
351 else if (new_node->ntfs_ino > this->ntfs_ino)
352 p = &((*p)->rb_right);
354 /* This should be impossible since a NTFS inode cannot
355 * have multiple DOS names, and we only should get each
356 * DOS name entry once from the ntfs_readdir() calls. */
357 ERROR("NTFS inode %"PRIu64" has multiple DOS names",
362 rb_link_node(&new_node->rb_node, rb_parent, p);
363 rb_insert_color(&new_node->rb_node, root);
364 DEBUG("Inserted DOS name for inode %"PRIu64, ntfs_ino);
368 /* Returns a structure that contains the DOS name and its length for a NTFS
369 * inode, or NULL if the inode has no DOS name. */
370 static struct dos_name_node *
371 lookup_dos_name(const struct dos_name_map *map, u64 ntfs_ino)
373 struct rb_node *node = map->rb_root.rb_node;
375 struct dos_name_node *this;
376 this = container_of(node, struct dos_name_node, rb_node);
377 if (ntfs_ino < this->ntfs_ino)
378 node = node->rb_left;
379 else if (ntfs_ino > this->ntfs_ino)
380 node = node->rb_right;
388 set_dentry_dos_name(struct wim_dentry *dentry, void *arg)
390 const struct dos_name_map *map = arg;
391 const struct dos_name_node *node;
393 if (dentry->is_win32_name) {
394 node = lookup_dos_name(map, dentry->d_inode->i_ino);
396 dentry->short_name = MALLOC(node->name_nbytes + 2);
397 if (!dentry->short_name)
398 return WIMLIB_ERR_NOMEM;
399 memcpy(dentry->short_name, node->dos_name,
401 dentry->short_name[node->name_nbytes / 2] = 0;
402 dentry->short_name_nbytes = node->name_nbytes;
403 DEBUG("Assigned DOS name to ino %"PRIu64,
404 dentry->d_inode->i_ino);
406 WARNING("NTFS inode %"PRIu64" has Win32 name with no "
407 "corresponding DOS name",
408 dentry->d_inode->i_ino);
415 free_dos_name_tree(struct rb_node *node) {
417 free_dos_name_tree(node->rb_left);
418 free_dos_name_tree(node->rb_right);
419 FREE(container_of(node, struct dos_name_node, rb_node));
424 destroy_dos_name_map(struct dos_name_map *map)
426 free_dos_name_tree(map->rb_root.rb_node);
430 struct wim_dentry *parent;
434 struct wim_lookup_table *lookup_table;
435 struct wim_inode_table *inode_table;
436 struct sd_set *sd_set;
437 struct dos_name_map *dos_name_map;
438 const struct wimlib_capture_config *config;
441 wimlib_progress_func_t progress_func;
445 build_dentry_tree_ntfs_recursive(struct wim_dentry **root_p,
451 struct wim_lookup_table *lookup_table,
452 struct wim_inode_table *inode_table,
453 struct sd_set *sd_set,
454 const struct wimlib_capture_config *config,
455 ntfs_volume *ntfs_vol,
457 wimlib_progress_func_t progress_func);
460 wim_ntfs_capture_filldir(void *dirent, const ntfschar *name,
461 const int name_nchars, const int name_type,
462 const s64 pos, const MFT_REF mref,
463 const unsigned dt_type)
465 struct readdir_ctx *ctx;
466 size_t mbs_name_nbytes;
468 struct wim_dentry *child;
471 size_t name_nbytes = name_nchars * sizeof(ntfschar);
474 if (name_type & FILE_NAME_DOS) {
475 /* If this is the entry for a DOS name, store it for later. */
476 ret = insert_dos_name(ctx->dos_name_map, name,
477 name_nbytes, mref & MFT_REF_MASK_CPU);
479 /* Return now if an error occurred or if this is just a DOS name
480 * and not a Win32+DOS name. */
481 if (ret != 0 || name_type == FILE_NAME_DOS)
484 ret = utf16le_to_tstr(name, name_nbytes,
485 &mbs_name, &mbs_name_nbytes);
489 if (mbs_name[0] == '.' &&
490 (mbs_name[1] == '\0' ||
491 (mbs_name[1] == '.' && mbs_name[2] == '\0'))) {
494 * note: name_type is POSIX for these, so DOS names will not
495 * have been inserted for them. */
497 goto out_free_mbs_name;
500 /* Open the inode for this directory entry and recursively capture the
501 * directory tree rooted at it */
502 ntfs_inode *ni = ntfs_inode_open(ctx->dir_ni->vol, mref);
504 ERROR_WITH_ERRNO("Failed to open NTFS inode");
506 goto out_free_mbs_name;
508 path_len = ctx->path_len;
510 ctx->path[path_len++] = '/';
511 memcpy(ctx->path + path_len, mbs_name, mbs_name_nbytes + 1);
512 path_len += mbs_name_nbytes;
514 ret = build_dentry_tree_ntfs_recursive(&child, ctx->dir_ni,
515 ni, ctx->path, path_len, name_type,
519 ctx->config, ctx->vol,
520 ctx->add_image_flags,
523 dentry_add_child(ctx->parent, child);
524 ntfs_inode_close(ni);
531 /* Recursively build a WIM dentry tree corresponding to a NTFS volume.
532 * At the same time, update the WIM lookup table with lookup table entries for
533 * the NTFS streams, and build an array of security descriptors.
536 build_dentry_tree_ntfs_recursive(struct wim_dentry **root_ret,
542 struct wim_lookup_table *lookup_table,
543 struct wim_inode_table *inode_table,
544 struct sd_set *sd_set,
545 const struct wimlib_capture_config *config,
548 wimlib_progress_func_t progress_func)
552 struct wim_dentry *root;
553 struct wim_inode *inode;
554 ATTR_TYPES stream_type;
556 if (exclude_path(path, path_len, config, false)) {
557 /* Exclude a file or directory tree based on the capture
558 * configuration file */
559 if ((add_image_flags & WIMLIB_ADD_IMAGE_FLAG_EXCLUDE_VERBOSE)
562 union wimlib_progress_info info;
563 info.scan.cur_path = path;
564 info.scan.excluded = true;
565 progress_func(WIMLIB_PROGRESS_MSG_SCAN_DENTRY, &info);
572 /* Get file attributes */
573 struct SECURITY_CONTEXT ctx;
574 memset(&ctx, 0, sizeof(ctx));
576 ret = ntfs_xattr_system_getxattr(&ctx, XATTR_NTFS_ATTRIB,
577 ni, dir_ni, (char *)&attributes,
580 ERROR_WITH_ERRNO("Failed to get NTFS attributes from `%s'",
582 return WIMLIB_ERR_NTFS_3G;
585 if ((add_image_flags & WIMLIB_ADD_IMAGE_FLAG_VERBOSE)
588 union wimlib_progress_info info;
589 info.scan.cur_path = path;
590 info.scan.excluded = false;
591 progress_func(WIMLIB_PROGRESS_MSG_SCAN_DENTRY, &info);
594 /* Create a WIM dentry with an associated inode, which may be shared */
595 ret = inode_table_new_dentry(inode_table,
596 path_basename_with_len(path, path_len),
603 inode = root->d_inode;
605 if (inode->i_nlink > 1) /* Shared inode; nothing more to do */
608 if (name_type & FILE_NAME_WIN32) /* Win32 or Win32+DOS name (rather than POSIX) */
609 root->is_win32_name = 1;
610 inode->i_creation_time = le64_to_cpu(ni->creation_time);
611 inode->i_last_write_time = le64_to_cpu(ni->last_data_change_time);
612 inode->i_last_access_time = le64_to_cpu(ni->last_access_time);
613 inode->i_attributes = le32_to_cpu(attributes);
614 inode->i_resolved = 1;
616 if (attributes & FILE_ATTR_REPARSE_POINT)
617 stream_type = AT_REPARSE_POINT;
619 stream_type = AT_DATA;
621 /* Capture the file's streams; more specifically, this is supposed to:
623 * - Regular files: capture unnamed data stream and any named data
625 * - Directories: capture any named data streams
626 * - Reparse points: capture reparse data only
628 ret = capture_ntfs_streams(inode, ni, path, path_len, lookup_table,
633 if (ni->mrec->flags & MFT_RECORD_IS_DIRECTORY) {
635 /* Recurse to directory children */
637 struct dos_name_map dos_name_map = { .rb_root = {.rb_node = NULL} };
638 struct readdir_ctx ctx = {
642 .path_len = path_len,
643 .lookup_table = lookup_table,
644 .inode_table = inode_table,
646 .dos_name_map = &dos_name_map,
649 .add_image_flags = add_image_flags,
650 .progress_func = progress_func,
652 ret = ntfs_readdir(ni, &pos, &ctx, wim_ntfs_capture_filldir);
654 ERROR_WITH_ERRNO("ntfs_readdir()");
655 ret = WIMLIB_ERR_NTFS_3G;
657 ret = for_dentry_child(root, set_dentry_dos_name,
660 destroy_dos_name_map(&dos_name_map);
665 if (!(add_image_flags & WIMLIB_ADD_IMAGE_FLAG_NO_ACLS)) {
666 /* Get security descriptor */
670 ret = ntfs_xattr_system_getxattr(&ctx, XATTR_NTFS_ACL,
673 if (ret > sizeof(sd)) {
675 ret = ntfs_xattr_system_getxattr(&ctx, XATTR_NTFS_ACL,
676 ni, dir_ni, sd, ret);
679 inode->i_security_id = sd_set_add_sd(sd_set, sd, ret);
680 if (inode->i_security_id == -1) {
681 ERROR("Out of memory");
682 ret = WIMLIB_ERR_NOMEM;
685 DEBUG("Added security ID = %u for `%s'",
686 inode->i_security_id, path);
688 } else if (ret < 0) {
689 ERROR_WITH_ERRNO("Failed to get security information from "
691 ret = WIMLIB_ERR_NTFS_3G;
693 inode->i_security_id = -1;
694 DEBUG("No security ID for `%s'", path);
701 free_dentry_tree(root, lookup_table);
707 do_ntfs_umount(struct _ntfs_volume *vol)
709 DEBUG("Unmounting NTFS volume");
710 if (ntfs_umount(vol, FALSE))
711 return WIMLIB_ERR_NTFS_3G;
717 build_dentry_tree_ntfs(struct wim_dentry **root_p,
719 struct wim_lookup_table *lookup_table,
720 struct wim_inode_table *inode_table,
721 struct sd_set *sd_set,
722 const struct wimlib_capture_config *config,
724 wimlib_progress_func_t progress_func,
731 DEBUG("Mounting NTFS volume `%s' read-only", device);
733 #ifdef HAVE_NTFS_MNT_RDONLY
735 vol = ntfs_mount(device, NTFS_MNT_RDONLY);
737 /* NTFS-3g 2011, 2012 */
738 vol = ntfs_mount(device, MS_RDONLY);
741 ERROR_WITH_ERRNO("Failed to mount NTFS volume `%s' read-only",
743 return WIMLIB_ERR_NTFS_3G;
745 ntfs_open_secure(vol);
747 /* We don't want to capture the special NTFS files such as $Bitmap. Not
748 * to be confused with "hidden" or "system" files which are real files
749 * that we do need to capture. */
750 NVolClearShowSysFiles(vol);
752 DEBUG("Opening root NTFS dentry");
753 root_ni = ntfs_inode_open(vol, FILE_root);
755 ERROR_WITH_ERRNO("Failed to open root inode of NTFS volume "
757 ret = WIMLIB_ERR_NTFS_3G;
761 /* Currently we assume that all the paths fit into this length and there
762 * is no check for overflow. */
763 char *path = MALLOC(32768);
765 ERROR("Could not allocate memory for NTFS pathname");
766 ret = WIMLIB_ERR_NOMEM;
772 ret = build_dentry_tree_ntfs_recursive(root_p, NULL, root_ni, path, 1,
773 FILE_NAME_POSIX, lookup_table,
781 ntfs_inode_close(root_ni);
783 ntfs_index_ctx_put(vol->secure_xsii);
784 ntfs_index_ctx_put(vol->secure_xsdh);
785 ntfs_inode_close(vol->secure_ni);
788 if (do_ntfs_umount(vol)) {
789 ERROR_WITH_ERRNO("Failed to unmount NTFS volume `%s'",
792 ret = WIMLIB_ERR_NTFS_3G;
795 /* We need to leave the NTFS volume mounted so that we can read
796 * the NTFS files again when we are actually writing the WIM */
797 *(ntfs_volume**)extra_arg = vol;