/*
* ntfs-3g_capture.c
*
- * Capture a WIM image directly from a NTFS volume using libntfs-3g. We capture
+ * Capture a WIM image directly from an NTFS volume using libntfs-3g. We capture
* everything we can, including security data and alternate data streams.
*/
/*
- * Copyright (C) 2012, 2013 Eric Biggers
+ * Copyright (C) 2012, 2013, 2014 Eric Biggers
*
* This file is part of wimlib, a library for working with WIM files.
*
}
-/* Load the streams from a file or reparse point in the NTFS volume into the WIM
- * lookup table */
+/* Load the streams from a file or reparse point in the NTFS volume */
static int
capture_ntfs_streams(struct wim_inode *inode,
ntfs_inode *ni,
char *path,
size_t path_len,
- struct wim_lookup_table *lookup_table,
+ struct list_head *unhashed_streams,
ntfs_volume *vol,
ATTR_TYPES type)
{
/* Capture each data stream or reparse data stream. */
while (!ntfs_attr_lookup(type, NULL, 0,
- WIMLIB_CASE_SENSITIVE, 0, NULL, 0, actx))
+ CASE_SENSITIVE, 0, NULL, 0, actx))
{
u64 data_size = ntfs_get_attribute_value_length(actx->attr);
u64 name_length = actx->attr->name_length;
* dentry's inode. */
if (inode->i_lte) {
if (lte) {
- ERROR("Found two un-named data streams for \"%s\" "
- "(sizes = %"PRIu64", %"PRIu64")",
- path, inode->i_lte->size,
- lte->size);
- ret = WIMLIB_ERR_NTFS_3G;
- goto out_free_lte;
+ if (!(inode->i_attributes &
+ FILE_ATTRIBUTE_REPARSE_POINT))
+ {
+ WARNING("Found two un-named "
+ "data streams for \"%s\" "
+ "(sizes = %"PRIu64", "
+ "%"PRIu64")",
+ path,
+ inode->i_lte->size,
+ lte->size);
+ }
+ free_lookup_table_entry(lte);
+ continue;
}
} else {
stream_id = 0;
new_ads_entry->lte = lte;
}
if (lte) {
- lookup_table_insert_unhashed(lookup_table, lte,
- inode, stream_id);
+ add_unhashed_stream(lte, inode,
+ stream_id, unhashed_streams);
}
}
if (errno == ENOENT) {
return ret;
}
-/* Red-black tree that maps NTFS inode numbers to DOS names */
+/* Binary tree that maps NTFS inode numbers to DOS names */
struct dos_name_map {
- struct rb_root rb_root;
+ struct avl_tree_node *root;
};
struct dos_name_node {
- struct rb_node rb_node;
+ struct avl_tree_node index_node;
char dos_name[24];
int name_nbytes;
le64 ntfs_ino;
};
+#define DOS_NAME_NODE(avl_node) \
+ avl_tree_entry(avl_node, struct dos_name_node, index_node)
+
+static int
+_avl_cmp_by_ntfs_ino(const struct avl_tree_node *n1,
+ const struct avl_tree_node *n2)
+{
+ return cmp_u64(DOS_NAME_NODE(n1)->ntfs_ino,
+ DOS_NAME_NODE(n2)->ntfs_ino);
+}
+
/* Inserts a new DOS name into the map */
static int
insert_dos_name(struct dos_name_map *map, const ntfschar *dos_name,
size_t name_nbytes, le64 ntfs_ino)
{
struct dos_name_node *new_node;
- struct rb_node **p;
- struct rb_root *root;
- struct rb_node *rb_parent;
DEBUG("DOS name_len = %zu", name_nbytes);
new_node = MALLOC(sizeof(struct dos_name_node));
wimlib_assert(name_nbytes <= sizeof(new_node->dos_name));
/* Initialize the DOS name, DOS name length, and NTFS inode number of
- * the red-black tree node */
+ * the search tree node */
memcpy(new_node->dos_name, dos_name, name_nbytes);
new_node->name_nbytes = name_nbytes;
new_node->ntfs_ino = ntfs_ino;
- /* Insert the red-black tree node */
- root = &map->rb_root;
- p = &root->rb_node;
- rb_parent = NULL;
- while (*p) {
- struct dos_name_node *this;
-
- this = container_of(*p, struct dos_name_node, rb_node);
- rb_parent = *p;
- if (new_node->ntfs_ino < this->ntfs_ino)
- p = &((*p)->rb_left);
- else if (new_node->ntfs_ino > this->ntfs_ino)
- p = &((*p)->rb_right);
- else {
- /* This should be impossible since a NTFS inode cannot
- * have multiple DOS names, and we only should get each
- * DOS name entry once from the ntfs_readdir() calls. */
- ERROR("NTFS inode %"PRIu64" has multiple DOS names",
- le64_to_cpu(ntfs_ino));
- return -1;
- }
+ /* Insert the search tree node */
+ if (avl_tree_insert(&map->root, &new_node->index_node,
+ _avl_cmp_by_ntfs_ino))
+ {
+ /* This should be impossible since an NTFS inode cannot
+ * have multiple DOS names, and we only should get each
+ * DOS name entry once from the ntfs_readdir() calls. */
+ ERROR("NTFS inode %"PRIu64" has multiple DOS names",
+ le64_to_cpu(ntfs_ino));
+ FREE(new_node);
+ return -1;
}
- rb_link_node(&new_node->rb_node, rb_parent, p);
- rb_insert_color(&new_node->rb_node, root);
DEBUG("Inserted DOS name for inode %"PRIu64, le64_to_cpu(ntfs_ino));
return 0;
}
-/* Returns a structure that contains the DOS name and its length for a NTFS
+/* Returns a structure that contains the DOS name and its length for an NTFS
* inode, or NULL if the inode has no DOS name. */
static struct dos_name_node *
lookup_dos_name(const struct dos_name_map *map, u64 ntfs_ino)
{
- struct rb_node *node = map->rb_root.rb_node;
- while (node) {
- struct dos_name_node *this;
- this = container_of(node, struct dos_name_node, rb_node);
- if (ntfs_ino < this->ntfs_ino)
- node = node->rb_left;
- else if (ntfs_ino > this->ntfs_ino)
- node = node->rb_right;
- else
- return this;
- }
- return NULL;
+ struct dos_name_node dummy;
+ struct avl_tree_node *res;
+
+ dummy.ntfs_ino = cpu_to_le64(ntfs_ino);
+
+ res = avl_tree_lookup_node(map->root, &dummy.index_node,
+ _avl_cmp_by_ntfs_ino);
+ if (!res)
+ return NULL;
+ return DOS_NAME_NODE(res);
}
static int
-set_dentry_dos_name(struct wim_dentry *dentry, void *arg)
+set_dentry_dos_name(struct wim_dentry *dentry, const struct dos_name_map *map)
{
- const struct dos_name_map *map = arg;
const struct dos_name_node *node;
if (dentry->is_win32_name) {
node = lookup_dos_name(map, dentry->d_inode->i_ino);
if (node) {
- dentry->short_name = MALLOC(node->name_nbytes + 2);
+ dentry->short_name = utf16le_dupz((const utf16lechar *)node->dos_name,
+ node->name_nbytes);
if (!dentry->short_name)
return WIMLIB_ERR_NOMEM;
- memcpy(dentry->short_name, node->dos_name,
- node->name_nbytes);
- dentry->short_name[node->name_nbytes / 2] = 0;
dentry->short_name_nbytes = node->name_nbytes;
DEBUG("Assigned DOS name to ino %"PRIu64,
dentry->d_inode->i_ino);
}
static void
-free_dos_name_tree(struct rb_node *node) {
+free_dos_name_tree(struct avl_tree_node *node) {
if (node) {
- free_dos_name_tree(node->rb_left);
- free_dos_name_tree(node->rb_right);
- FREE(container_of(node, struct dos_name_node, rb_node));
+ free_dos_name_tree(node->left);
+ free_dos_name_tree(node->right);
+ FREE(DOS_NAME_NODE(node));
}
}
static void
destroy_dos_name_map(struct dos_name_map *map)
{
- free_dos_name_tree(map->rb_root.rb_node);
+ free_dos_name_tree(map->root);
}
struct readdir_ctx {
return ret;
}
-/* Recursively build a WIM dentry tree corresponding to a NTFS volume.
+/* Recursively build a WIM dentry tree corresponding to an NTFS volume.
* At the same time, update the WIM lookup table with lookup table entries for
* the NTFS streams, and build an array of security descriptors.
*/
int ret;
struct wim_dentry *root = NULL;
struct wim_inode *inode = NULL;
- ATTR_TYPES stream_type;
- if (exclude_path(path, path_len, params->config, false)) {
+ if (should_exclude_path(path, path_len, params->config)) {
/* Exclude a file or directory tree based on the capture
* configuration file. */
ret = 0;
}
/* Create a WIM dentry with an associated inode, which may be shared */
- ret = inode_table_new_dentry(¶ms->inode_table,
+ ret = inode_table_new_dentry(params->inode_table,
path_basename_with_len(path, path_len),
ni->mft_no, 0, false, &root);
if (ret)
inode->i_attributes = le32_to_cpu(attributes);
inode->i_resolved = 1;
- if (attributes & FILE_ATTR_REPARSE_POINT)
- stream_type = AT_REPARSE_POINT;
- else
- stream_type = AT_DATA;
+ /* Capture streams. */
+
+ if (attributes & FILE_ATTR_REPARSE_POINT) {
+ /* Capture reparse data stream. */
+ ret = capture_ntfs_streams(inode, ni, path, path_len,
+ params->unhashed_streams,
+ vol, AT_REPARSE_POINT);
+ if (ret)
+ goto out;
+ }
- /* Capture the file's streams; more specifically, this is supposed to:
+ /* Capture data streams.
*
- * - Regular files: capture unnamed data stream and any named data
- * streams
- * - Directories: capture any named data streams
- * - Reparse points: capture reparse data only
- */
+ * Directories should not have an unnamed data stream, but they may have
+ * named data streams.
+ *
+ * Reparse points may have an unnamed data stream (which will be ignored
+ * in favor of the reparse data stream), and they also may have named
+ * data streams.
+ *
+ * Regular files can have an unnamed data stream as well as named data
+ * streams. */
ret = capture_ntfs_streams(inode, ni, path, path_len,
- params->lookup_table, vol, stream_type);
+ params->unhashed_streams, vol, AT_DATA);
if (ret)
goto out;
/* Recurse to directory children */
s64 pos = 0;
- struct dos_name_map dos_name_map = { .rb_root = {.rb_node = NULL} };
+ struct dos_name_map dos_name_map = { .root = NULL };
struct readdir_ctx ctx = {
.parent = root,
.path = path,
ERROR_WITH_ERRNO("Error reading directory \"%s\"", path);
ret = WIMLIB_ERR_NTFS_3G;
} else {
- ret = for_dentry_child(root, set_dentry_dos_name,
- &dos_name_map);
+ struct wim_dentry *child;
+
+ ret = 0;
+ for_dentry_child(child, root) {
+ ret = set_dentry_dos_name(child, &dos_name_map);
+ if (ret)
+ break;
+ }
}
destroy_dos_name_map(&dos_name_map);
if (ret)
ret = ntfs_get_ntfs_acl(&sec_ctx, ni, sd, ret);
}
if (ret > 0) {
- inode->i_security_id = sd_set_add_sd(¶ms->sd_set,
+ inode->i_security_id = sd_set_add_sd(params->sd_set,
sd, ret);
if (inode->i_security_id == -1) {
ERROR("Out of memory");