#include "wimlib/lookup_table.h"
#include "wimlib/metadata.h"
#include "wimlib/resource.h"
+#include "wimlib/security.h"
#include "wimlib/sha1.h"
#include "wimlib/timestamp.h"
#define WIM_ADS_ENTRY_DISK_SIZE 38
-/* WIM directory entry (on-disk format) */
+/* On-disk format of a WIM dentry (directory entry), located in the metadata
+ * resource for a WIM image. */
struct wim_dentry_on_disk {
+
+ /* Length of this directory entry in bytes, not including any alternate
+ * data stream entries. Should be a multiple of 8 so that the following
+ * dentry or alternate data stream entry is aligned on an 8-byte
+ * boundary. (If not, wimlib will round it up.) It must be at least as
+ * long as the fixed-length fields of the dentry (WIM_DENTRY_DISK_SIZE),
+ * plus the lengths of the file name and/or short name if present.
+ *
+ * It is also possible for this field to be 0. This situation, which is
+ * undocumented, indicates the end of a list of sibling nodes in a
+ * directory. It also means the real length is 8, because the dentry
+ * included only the length field, but that takes up 8 bytes. */
le64 length;
+
+ /* Attributes of the file or directory. This is a bitwise OR of the
+ * FILE_ATTRIBUTE_* constants and should correspond to the value
+ * retrieved by GetFileAttributes() on Windows. */
le32 attributes;
+
+ /* A value that specifies the security descriptor for this file or
+ * directory. If -1, the file or directory has no security descriptor.
+ * Otherwise, it is a 0-based index into the WIM image's table of
+ * security descriptors (see: `struct wim_security_data') */
sle32 security_id;
+
+ /* Offset, in bytes, from the start of the uncompressed metadata
+ * resource of this directory's child directory entries, or 0 if this
+ * directory entry does not correspond to a directory or otherwise does
+ * not have any children. */
le64 subdir_offset;
+
+ /* Reserved fields */
le64 unused_1;
le64 unused_2;
+
+
+ /* Creation time, last access time, and last write time, in
+ * 100-nanosecond intervals since 12:00 a.m UTC January 1, 1601. They
+ * should correspond to the times gotten by calling GetFileTime() on
+ * Windows. */
le64 creation_time;
le64 last_access_time;
le64 last_write_time;
+
+ /* Vaguely, the SHA-1 message digest ("hash") of the file's contents.
+ * More specifically, this is for the "unnamed data stream" rather than
+ * any "alternate data streams". This hash value is used to look up the
+ * corresponding entry in the WIM's stream lookup table to actually find
+ * the file contents within the WIM.
+ *
+ * If the file has no unnamed data stream (e.g. is a directory), then
+ * this field will be all zeroes. If the unnamed data stream is empty
+ * (i.e. an "empty file"), then this field is also expected to be all
+ * zeroes. (It will be if wimlib created the WIM image, at least;
+ * otherwise it can't be ruled out that the SHA-1 message digest of 0
+ * bytes of data is given explicitly.)
+ *
+ * If the file has reparse data, then this field will instead specify
+ * the SHA-1 message digest of the reparse data. If it is somehow
+ * possible for a file to have both an unnamed data stream and reparse
+ * data, then this is not handled by wimlib.
+ *
+ * As a further special case, if this field is all zeroes but there is
+ * an alternate data stream entry with no name and a nonzero SHA-1
+ * message digest field, then that hash must be used instead of this
+ * one. (wimlib does not use this quirk on WIM images it creates.)
+ */
u8 unnamed_stream_hash[SHA1_HASH_SIZE];
+
+ /* The format of the following data is not yet completely known and they
+ * do not correspond to Microsoft's documentation.
+ *
+ * If this directory entry is for a reparse point (has
+ * FILE_ATTRIBUTE_REPARSE_POINT set in the attributes field), then the
+ * version of the following fields containing the reparse tag is valid.
+ * Furthermore, the field notated as not_rpfixed, as far as I can tell,
+ * is supposed to be set to 1 if reparse point fixups (a.k.a. fixing the
+ * targets of absolute symbolic links) were *not* done, and otherwise 0.
+ *
+ * If this directory entry is not for a reparse point, then the version
+ * of the following fields containing the hard_link_group_id is valid.
+ * All MS says about this field is that "If this file is part of a hard
+ * link set, all the directory entries in the set will share the same
+ * value in this field.". However, more specifically I have observed
+ * the following:
+ * - If the file is part of a hard link set of size 1, then the
+ * hard_link_group_id should be set to either 0, which is treated
+ * specially as indicating "not hardlinked", or any unique value.
+ * - The specific nonzero values used to identity hard link sets do
+ * not matter, as long as they are unique.
+ * - However, due to bugs in Microsoft's software, it is actually NOT
+ * guaranteed that directory entries that share the same hard link
+ * group ID are actually hard linked to each either. We have to
+ * handle this by using special code to use distinguishing features
+ * (which is possible because some information about the underlying
+ * inode is repeated in each dentry) to split up these fake hard link
+ * groups into what they actually are supposed to be.
+ */
union {
struct {
le32 rp_unknown_1;
le64 hard_link_group_id;
} _packed_attribute nonreparse;
};
+
+ /* Number of alternate data stream entries that directly follow this
+ * dentry on-disk. */
le16 num_alternate_data_streams;
+
+ /* Length of this file's UTF-16LE encoded short name (8.3 DOS-compatible
+ * name), if present, in bytes, excluding the null terminator. If this
+ * file has no short name, then this field should be 0. */
le16 short_name_nbytes;
+
+ /* Length of this file's UTF-16LE encoded "long" name, excluding the
+ * null terminator. If this file has no short name, then this field
+ * should be 0. It's expected that only the root dentry has this field
+ * set to 0. */
le16 file_name_nbytes;
- /* Follewed by variable length file name, if file_name_nbytes != 0 */
- utf16lechar file_name[];
+ /* Followed by variable length file name, in UTF16-LE, if
+ * file_name_nbytes != 0. Includes null terminator. */
+ /*utf16lechar file_name[];*/
- /* Followed by variable length short name, if short_name_nbytes != 0 */
+ /* Followed by variable length short name, in UTF16-LE, if
+ * short_name_nbytes != 0. Includes null terminator. */
/*utf16lechar short_name[];*/
} _packed_attribute;
inode->i_next_stream_id = 1;
inode->i_not_rpfixed = 1;
INIT_LIST_HEAD(&inode->i_list);
- #ifdef WITH_FUSE
- if (pthread_mutex_init(&inode->i_mutex, NULL) != 0) {
- ERROR_WITH_ERRNO("Error initializing mutex");
- FREE(inode);
- return NULL;
- }
- #endif
INIT_LIST_HEAD(&inode->i_dentry);
}
return inode;
destroy_ads_entry(&inode->i_ads_entries[i]);
FREE(inode->i_ads_entries);
}
- #ifdef WITH_FUSE
- wimlib_assert(inode->i_num_opened_fds == 0);
- FREE(inode->i_fds);
- pthread_mutex_destroy(&inode->i_mutex);
- #endif
/* HACK: This may instead delete the inode from i_list, but the
* hlist_del() behaves the same as list_del(). */
if (!hlist_unhashed(&inode->i_hlist))
hlist_del(&inode->i_hlist);
- FREE(inode->i_extracted_file);
FREE(inode);
}
}
* Returns zero on success; nonzero on failure.
*/
int
-read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len,
- struct wim_dentry *dentry)
+read_dentry_tree(const u8 * restrict metadata_resource,
+ u64 metadata_resource_len,
+ struct wim_dentry * restrict dentry)
{
u64 cur_offset = dentry->subdir_offset;
struct wim_dentry *child;
* entries. */
cur_offset += dentry_total_length(child);
+ if (unlikely(!dentry_has_long_name(child))) {
+ WARNING("Ignoring unnamed dentry in "
+ "directory \"%"TS"\"",
+ dentry_full_path(dentry));
+ free_dentry(child);
+ continue;
+ }
+
duplicate = dentry_add_child(dentry, child);
if (unlikely(duplicate)) {
const tchar *child_type, *duplicate_type;
child_type, dentry_full_path(duplicate),
duplicate_type);
free_dentry(child);
- } else {
- inode_add_dentry(child, child->d_inode);
- /* If there are children of this child, call this
- * procedure recursively. */
- if (child->subdir_offset != 0) {
- if (likely(dentry_is_directory(child))) {
- ret = read_dentry_tree(metadata_resource,
- metadata_resource_len,
- child);
- if (ret)
- break;
- } else {
- WARNING("Ignoring children of non-directory \"%"TS"\"",
- dentry_full_path(child));
- }
- }
+ continue;
+ }
+ inode_add_dentry(child, child->d_inode);
+ /* If there are children of this child, call this
+ * procedure recursively. */
+ if (child->subdir_offset != 0) {
+ if (likely(dentry_is_directory(child))) {
+ ret = read_dentry_tree(metadata_resource,
+ metadata_resource_len,
+ child);
+ if (ret)
+ break;
+ } else {
+ WARNING("Ignoring children of non-directory \"%"TS"\"",
+ dentry_full_path(child));
+ }
}
}
return ret;
disk_dentry->file_name_nbytes = cpu_to_le16(dentry->file_name_nbytes);
p += sizeof(struct wim_dentry_on_disk);
+ wimlib_assert(dentry_is_root(dentry) != dentry_has_long_name(dentry));
+
if (dentry_has_long_name(dentry))
p = mempcpy(p, dentry->file_name, dentry->file_name_nbytes + 2);
* Returns pointer to the byte after the last byte we wrote.
*/
u8 *
-write_dentry_tree(const struct wim_dentry *root, u8 *p)
+write_dentry_tree(const struct wim_dentry * restrict root, u8 * restrict p)
{
DEBUG("Writing dentry tree.");
wimlib_assert(dentry_is_root(root));
/* Recursively write the rest of the dentry tree. */
return write_dentry_tree_recursive(root, p);
}
+
+
+static int
+init_wimlib_dentry(struct wimlib_dir_entry *wdentry,
+ struct wim_dentry *dentry,
+ const WIMStruct *wim)
+{
+ int ret;
+ size_t dummy;
+ const struct wim_inode *inode = dentry->d_inode;
+ struct wim_lookup_table_entry *lte;
+
+#if TCHAR_IS_UTF16LE
+ wdentry->filename = dentry->file_name;
+ wdentry->dos_name = dentry->short_name;
+#else
+ if (dentry_has_long_name(dentry)) {
+ ret = utf16le_to_tstr(dentry->file_name,
+ dentry->file_name_nbytes,
+ (tchar**)&wdentry->filename,
+ &dummy);
+ if (ret)
+ return ret;
+ }
+ if (dentry_has_short_name(dentry)) {
+ ret = utf16le_to_tstr(dentry->short_name,
+ dentry->short_name_nbytes,
+ (tchar**)&wdentry->dos_name,
+ &dummy);
+ if (ret)
+ return ret;
+ }
+#endif
+ ret = calculate_dentry_full_path(dentry);
+ if (ret)
+ return ret;
+ wdentry->full_path = dentry->_full_path;
+
+ for (struct wim_dentry *d = dentry; !dentry_is_root(d); d = d->parent)
+ wdentry->depth++;
+
+ if (inode->i_security_id >= 0) {
+ const struct wim_security_data *sd = wim_const_security_data(wim);
+ wdentry->security_descriptor = sd->descriptors[inode->i_security_id];
+ wdentry->security_descriptor_size = sd->sizes[inode->i_security_id];
+ }
+ wdentry->reparse_tag = inode->i_reparse_tag;
+ wdentry->num_links = inode->i_nlink;
+ wdentry->attributes = inode->i_attributes;
+ wdentry->hard_link_group_id = inode->i_ino;
+ wdentry->creation_time = wim_timestamp_to_timespec(inode->i_creation_time);
+ wdentry->last_write_time = wim_timestamp_to_timespec(inode->i_last_write_time);
+ wdentry->last_access_time = wim_timestamp_to_timespec(inode->i_last_access_time);
+
+ lte = inode_unnamed_lte(inode, wim->lookup_table);
+ if (lte)
+ lte_to_wimlib_resource_entry(lte, &wdentry->streams[0].resource);
+
+ for (unsigned i = 0; i < inode->i_num_ads; i++) {
+ if (inode->i_ads_entries[i].stream_name == NULL)
+ continue;
+ lte = inode_stream_lte(inode, i + 1, wim->lookup_table);
+ wdentry->num_named_streams++;
+ if (lte) {
+ lte_to_wimlib_resource_entry(lte, &wdentry->streams[
+ wdentry->num_named_streams].resource);
+ }
+ #if TCHAR_IS_UTF16LE
+ wdentry->streams[wdentry->num_named_streams].stream_name =
+ inode->i_ads_entries[i].stream_name;
+ #else
+ size_t dummy;
+
+ ret = utf16le_to_tstr(inode->i_ads_entries[i].stream_name,
+ inode->i_ads_entries[i].stream_name_nbytes,
+ (tchar**)&wdentry->streams[
+ wdentry->num_named_streams].stream_name,
+ &dummy);
+ if (ret)
+ return ret;
+ #endif
+ }
+ return 0;
+}
+
+static void
+free_wimlib_dentry(struct wimlib_dir_entry *wdentry)
+{
+#if !TCHAR_IS_UTF16LE
+ FREE((tchar*)wdentry->filename);
+ FREE((tchar*)wdentry->dos_name);
+ for (unsigned i = 1; i <= wdentry->num_named_streams; i++)
+ FREE((tchar*)wdentry->streams[i].stream_name);
+#endif
+ FREE(wdentry);
+}
+
+struct iterate_dir_tree_ctx {
+ WIMStruct *wim;
+ int flags;
+ wimlib_iterate_dir_tree_callback_t cb;
+ void *user_ctx;
+};
+
+static int
+do_iterate_dir_tree(WIMStruct *wim,
+ struct wim_dentry *dentry, int flags,
+ wimlib_iterate_dir_tree_callback_t cb,
+ void *user_ctx);
+
+static int
+call_do_iterate_dir_tree(struct wim_dentry *dentry, void *_ctx)
+{
+ struct iterate_dir_tree_ctx *ctx = _ctx;
+ return do_iterate_dir_tree(ctx->wim, dentry, ctx->flags,
+ ctx->cb, ctx->user_ctx);
+}
+
+static int
+do_iterate_dir_tree(WIMStruct *wim,
+ struct wim_dentry *dentry, int flags,
+ wimlib_iterate_dir_tree_callback_t cb,
+ void *user_ctx)
+{
+ struct wimlib_dir_entry *wdentry;
+ int ret = WIMLIB_ERR_NOMEM;
+
+
+ wdentry = CALLOC(1, sizeof(struct wimlib_dir_entry) +
+ (1 + dentry->d_inode->i_num_ads) *
+ sizeof(struct wimlib_stream_entry));
+ if (!wdentry)
+ goto out;
+
+ ret = init_wimlib_dentry(wdentry, dentry, wim);
+ if (ret)
+ goto out_free_wimlib_dentry;
+
+ if (!(flags & WIMLIB_ITERATE_DIR_TREE_FLAG_CHILDREN)) {
+ ret = (*cb)(wdentry, user_ctx);
+ if (ret)
+ goto out_free_wimlib_dentry;
+ }
+
+ if (flags & (WIMLIB_ITERATE_DIR_TREE_FLAG_RECURSIVE |
+ WIMLIB_ITERATE_DIR_TREE_FLAG_CHILDREN))
+ {
+ struct iterate_dir_tree_ctx ctx = {
+ .wim = wim,
+ .flags = flags &= ~WIMLIB_ITERATE_DIR_TREE_FLAG_CHILDREN,
+ .cb = cb,
+ .user_ctx = user_ctx,
+ };
+ ret = for_dentry_child(dentry, call_do_iterate_dir_tree, &ctx);
+ }
+out_free_wimlib_dentry:
+ free_wimlib_dentry(wdentry);
+out:
+ return ret;
+}
+
+struct image_iterate_dir_tree_ctx {
+ const tchar *path;
+ int flags;
+ wimlib_iterate_dir_tree_callback_t cb;
+ void *user_ctx;
+};
+
+
+static int
+image_do_iterate_dir_tree(WIMStruct *wim)
+{
+ struct image_iterate_dir_tree_ctx *ctx = wim->private;
+ struct wim_dentry *dentry;
+
+ dentry = get_dentry(wim, ctx->path);
+ if (!dentry)
+ return WIMLIB_ERR_PATH_DOES_NOT_EXIST;
+ return do_iterate_dir_tree(wim, dentry, ctx->flags, ctx->cb, ctx->user_ctx);
+}
+
+WIMLIBAPI int
+wimlib_iterate_dir_tree(WIMStruct *wim, int image, const tchar *path,
+ int flags,
+ wimlib_iterate_dir_tree_callback_t cb, void *user_ctx)
+{
+ struct image_iterate_dir_tree_ctx ctx = {
+ .path = path,
+ .flags = flags,
+ .cb = cb,
+ .user_ctx = user_ctx,
+ };
+ wim->private = &ctx;
+ return for_image(wim, image, image_do_iterate_dir_tree);
+}
git://wimlib.net / wimlib / blobdiff