* wimlib; if not, see http://www.gnu.org/licenses/.
*/
-#include "buffer_io.h"
-#include "dentry.h"
-#include "lookup_table.h"
-#include "timestamp.h"
-#include "wimlib_internal.h"
+#ifdef HAVE_CONFIG_H
+# include "config.h"
+#endif
+
+#include "wimlib.h"
+#include "wimlib/dentry.h"
+#include "wimlib/encoding.h"
+#include "wimlib/endianness.h"
+#include "wimlib/error.h"
+#include "wimlib/lookup_table.h"
+#include "wimlib/metadata.h"
+#include "wimlib/resource.h"
+#include "wimlib/sha1.h"
+#include "wimlib/timestamp.h"
+
#include <errno.h>
+/* WIM alternate data stream entry (on-disk format) */
+struct wim_ads_entry_on_disk {
+ /* Length of the entry, in bytes. This apparently includes all
+ * fixed-length fields, plus the stream name and null terminator if
+ * present, and the padding up to an 8 byte boundary. wimlib is a
+ * little less strict when reading the entries, and only requires that
+ * the number of bytes from this field is at least as large as the size
+ * of the fixed length fields and stream name without null terminator.
+ * */
+ le64 length;
+
+ le64 reserved;
+
+ /* SHA1 message digest of the uncompressed stream; or, alternatively,
+ * can be all zeroes if the stream has zero length. */
+ u8 hash[SHA1_HASH_SIZE];
+
+ /* Length of the stream name, in bytes. 0 if the stream is unnamed. */
+ le16 stream_name_nbytes;
+
+ /* Stream name in UTF-16LE. It is @stream_name_nbytes bytes long,
+ * excluding the the null terminator. There is a null terminator
+ * character if @stream_name_nbytes != 0; i.e., if this stream is named.
+ * */
+ utf16lechar stream_name[];
+} _packed_attribute;
+
+#define WIM_ADS_ENTRY_DISK_SIZE 38
+
+/* 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 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 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;
+
+ /* The following three time fields should correspond to those gotten by
+ * calling GetFileTime() on Windows. */
+
+ /* Creation time, in 100-nanosecond intervals since January 1, 1601. */
+ le64 creation_time;
+
+ /* Last access time, in 100-nanosecond intervals since January 1, 1601. */
+ le64 last_access_time;
+
+ /* Last write time, in 100-nanosecond intervals since January 1, 1601. */
+ 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 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
+ * (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;
+ le32 reparse_tag;
+ le16 rp_unknown_2;
+ le16 not_rpfixed;
+ } _packed_attribute reparse;
+ 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, in UTF16-LE, if
+ * file_name_nbytes != 0. Includes null terminator. */
+ utf16lechar file_name[];
+
+ /* Followed by variable length short name, in UTF16-LE, if
+ * short_name_nbytes != 0. Includes null terminator. */
+ /*utf16lechar short_name[];*/
+} _packed_attribute;
+
+#define WIM_DENTRY_DISK_SIZE 102
+
/* Calculates the unaligned length, in bytes, of an on-disk WIM dentry that has
* a file name and short name that take the specified numbers of bytes. This
* excludes any alternate data stream entries that may follow the dentry. */
static u64
-__dentry_correct_length_unaligned(u16 file_name_nbytes, u16 short_name_nbytes)
+_dentry_correct_length_unaligned(u16 file_name_nbytes, u16 short_name_nbytes)
{
- u64 length = WIM_DENTRY_DISK_SIZE;
+ u64 length = sizeof(struct wim_dentry_on_disk);
if (file_name_nbytes)
length += file_name_nbytes + 2;
if (short_name_nbytes)
static u64
dentry_correct_length_unaligned(const struct wim_dentry *dentry)
{
- return __dentry_correct_length_unaligned(dentry->file_name_nbytes,
- dentry->short_name_nbytes);
-}
-
-/* Return the "correct" value to write in the length field of a WIM dentry,
- * based on the file name length and short name length. */
-static u64
-dentry_correct_length(const struct wim_dentry *dentry)
-{
- return (dentry_correct_length_unaligned(dentry) + 7) & ~7;
-}
-
-/* Return %true iff the alternate data stream entry @entry has the UTF-16LE
- * stream name @name that has length @name_nbytes bytes. */
-static inline bool
-ads_entry_has_name(const struct wim_ads_entry *entry,
- const utf16lechar *name, size_t name_nbytes)
-{
- return entry->stream_name_nbytes == name_nbytes &&
- memcmp(entry->stream_name, name, name_nbytes) == 0;
+ return _dentry_correct_length_unaligned(dentry->file_name_nbytes,
+ dentry->short_name_nbytes);
}
-/* Duplicates a multibyte string into a UTF-16LE string and returns the string
- * and its length, in bytes, in the pointer arguments. Frees any existing
- * string at the return location before overwriting it. */
+/* Duplicates a string of system-dependent encoding into a UTF-16LE string and
+ * returns the string and its length, in bytes, in the pointer arguments. Frees
+ * any existing string at the return location before overwriting it. */
static int
-get_utf16le_name(const mbchar *name, utf16lechar **name_utf16le_ret,
+get_utf16le_name(const tchar *name, utf16lechar **name_utf16le_ret,
u16 *name_utf16le_nbytes_ret)
{
utf16lechar *name_utf16le;
size_t name_utf16le_nbytes;
int ret;
+#if TCHAR_IS_UTF16LE
+ name_utf16le_nbytes = tstrlen(name) * sizeof(utf16lechar);
+ name_utf16le = MALLOC(name_utf16le_nbytes + sizeof(utf16lechar));
+ if (!name_utf16le)
+ return WIMLIB_ERR_NOMEM;
+ memcpy(name_utf16le, name, name_utf16le_nbytes + sizeof(utf16lechar));
+ ret = 0;
+#else
- ret = mbs_to_utf16le(name, strlen(name), &name_utf16le,
- &name_utf16le_nbytes);
- if (name_utf16le_nbytes > 0xffff) {
- FREE(name_utf16le);
- ERROR("Multibyte string \"%s\" is too long!", name);
- ret = WIMLIB_ERR_INVALID_UTF8_STRING;
+ ret = tstr_to_utf16le(name, tstrlen(name), &name_utf16le,
+ &name_utf16le_nbytes);
+ if (ret == 0) {
+ if (name_utf16le_nbytes > 0xffff) {
+ FREE(name_utf16le);
+ ERROR("Multibyte string \"%"TS"\" is too long!", name);
+ ret = WIMLIB_ERR_INVALID_UTF8_STRING;
+ }
}
+#endif
if (ret == 0) {
FREE(*name_utf16le_ret);
*name_utf16le_ret = name_utf16le;
/* Sets the name of a WIM dentry from a multibyte string. */
int
-set_dentry_name(struct wim_dentry *dentry, const mbchar *new_name)
+set_dentry_name(struct wim_dentry *dentry, const tchar *new_name)
{
int ret;
ret = get_utf16le_name(new_name, &dentry->file_name,
&dentry->file_name_nbytes);
if (ret == 0) {
/* Clear the short name and recalculate the dentry length */
- if (dentry->short_name_nbytes) {
+ if (dentry_has_short_name(dentry)) {
FREE(dentry->short_name);
dentry->short_name = NULL;
dentry->short_name_nbytes = 0;
}
- dentry->length = dentry_correct_length(dentry);
}
return ret;
}
static u64
ads_entry_total_length(const struct wim_ads_entry *entry)
{
- u64 len = WIM_ADS_ENTRY_DISK_SIZE;
+ u64 len = sizeof(struct wim_ads_entry_on_disk);
if (entry->stream_name_nbytes)
len += entry->stream_name_nbytes + 2;
return (len + 7) & ~7;
static u64
-__dentry_total_length(const struct wim_dentry *dentry, u64 length)
+_dentry_total_length(const struct wim_dentry *dentry, u64 length)
{
const struct wim_inode *inode = dentry->d_inode;
for (u16 i = 0; i < inode->i_num_ads; i++)
u64
dentry_correct_total_length(const struct wim_dentry *dentry)
{
- return __dentry_total_length(dentry,
- dentry_correct_length_unaligned(dentry));
+ return _dentry_total_length(dentry,
+ dentry_correct_length_unaligned(dentry));
}
/* Like dentry_correct_total_length(), but use the existing dentry->length field
static u64
dentry_total_length(const struct wim_dentry *dentry)
{
- return __dentry_total_length(dentry, dentry->length);
+ return _dentry_total_length(dentry, dentry->length);
}
int
int ret;
if (node) {
ret = for_dentry_tree_in_rbtree(node->rb_left, visitor, arg);
- if (ret != 0)
+ if (ret)
return ret;
ret = for_dentry_in_tree(rbnode_dentry(node), visitor, arg);
- if (ret != 0)
+ if (ret)
return ret;
ret = for_dentry_tree_in_rbtree(node->rb_right, visitor, arg);
- if (ret != 0)
+ if (ret)
return ret;
}
return 0;
for_dentry_in_tree(struct wim_dentry *root,
int (*visitor)(struct wim_dentry*, void*), void *arg)
{
- int ret = visitor(root, arg);
- if (ret == 0) {
- ret = for_dentry_tree_in_rbtree(root->d_inode->i_children.rb_node,
- visitor,
- arg);
- }
- return ret;
+ int ret;
+
+ if (!root)
+ return 0;
+ ret = (*visitor)(root, arg);
+ if (ret)
+ return ret;
+ return for_dentry_tree_in_rbtree(root->d_inode->i_children.rb_node,
+ visitor,
+ arg);
}
/* Like for_dentry_in_tree(), but the visitor function is always called on a
int (*visitor)(struct wim_dentry*, void*), void *arg)
{
int ret;
+
+ if (!root)
+ return 0;
ret = for_dentry_tree_in_rbtree_depth(root->d_inode->i_children.rb_node,
visitor, arg);
- if (ret == 0)
- ret = visitor(root, arg);
- return ret;
+ if (ret)
+ return ret;
+ return (*visitor)(root, arg);
}
/* Calculate the full path of @dentry. The full path of its parent must have
- * already been calculated. */
+ * already been calculated, or it must be the root dentry. */
int
-calculate_dentry_full_path(struct wim_dentry *dentry, void *ignore)
+calculate_dentry_full_path(struct wim_dentry *dentry)
{
- char *full_path;
+ tchar *full_path;
u32 full_path_nbytes;
+ int ret;
- wimlib_assert(dentry_is_root(dentry) ||
- dentry->parent->full_path != NULL);
+ if (dentry->_full_path)
+ return 0;
if (dentry_is_root(dentry)) {
- full_path = MALLOC(2);
+ static const tchar _root_path[] = {WIM_PATH_SEPARATOR, T('\0')};
+ full_path = TSTRDUP(_root_path);
if (!full_path)
return WIMLIB_ERR_NOMEM;
- full_path[0] = '/';
- full_path[1] = '\0';
- full_path_nbytes = 1;
+ full_path_nbytes = 1 * sizeof(tchar);
} else {
- char *parent_full_path;
+ struct wim_dentry *parent;
+ tchar *parent_full_path;
u32 parent_full_path_nbytes;
- const struct wim_dentry *parent;
- char *name_mbs;
- size_t name_mbs_nbytes;
- int ret;
-
- ret = utf16le_to_mbs_nbytes(dentry->file_name,
- dentry->file_name_nbytes,
- &name_mbs_nbytes);
- if (ret)
- return ret;
+ size_t filename_nbytes;
+
parent = dentry->parent;
if (dentry_is_root(parent)) {
- parent_full_path = "";
+ parent_full_path = T("");
parent_full_path_nbytes = 0;
} else {
- parent_full_path = parent->full_path;
+ if (!parent->_full_path) {
+ ret = calculate_dentry_full_path(parent);
+ if (ret)
+ return ret;
+ }
+ parent_full_path = parent->_full_path;
parent_full_path_nbytes = parent->full_path_nbytes;
}
- full_path_nbytes = parent_full_path_nbytes + 1 +
- name_mbs_nbytes;
- full_path = MALLOC(full_path_nbytes + 1);
+
+ /* Append this dentry's name as a tchar string to the full path
+ * of the parent followed by the path separator */
+ #if TCHAR_IS_UTF16LE
+ filename_nbytes = dentry->file_name_nbytes;
+ #else
+ {
+ int ret = utf16le_to_tstr_nbytes(dentry->file_name,
+ dentry->file_name_nbytes,
+ &filename_nbytes);
+ if (ret)
+ return ret;
+ }
+ #endif
+
+ full_path_nbytes = parent_full_path_nbytes + sizeof(tchar) +
+ filename_nbytes;
+ full_path = MALLOC(full_path_nbytes + sizeof(tchar));
if (!full_path)
return WIMLIB_ERR_NOMEM;
memcpy(full_path, parent_full_path, parent_full_path_nbytes);
- full_path[parent_full_path_nbytes] = '/';
-
- utf16le_to_mbs_buf(dentry->file_name,
- dentry->file_name_nbytes,
- &full_path[parent_full_path_nbytes + 1]);
+ full_path[parent_full_path_nbytes / sizeof(tchar)] = WIM_PATH_SEPARATOR;
+ #if TCHAR_IS_UTF16LE
+ memcpy(&full_path[parent_full_path_nbytes / sizeof(tchar) + 1],
+ dentry->file_name,
+ filename_nbytes + sizeof(tchar));
+ #else
+ utf16le_to_tstr_buf(dentry->file_name,
+ dentry->file_name_nbytes,
+ &full_path[parent_full_path_nbytes /
+ sizeof(tchar) + 1]);
+ #endif
}
- FREE(dentry->full_path);
- dentry->full_path = full_path;
+ dentry->_full_path = full_path;
dentry->full_path_nbytes= full_path_nbytes;
return 0;
}
+static int
+do_calculate_dentry_full_path(struct wim_dentry *dentry, void *_ignore)
+{
+ return calculate_dentry_full_path(dentry);
+}
+
+int
+calculate_dentry_tree_full_paths(struct wim_dentry *root)
+{
+ return for_dentry_in_tree(root, do_calculate_dentry_full_path, NULL);
+}
+
+tchar *
+dentry_full_path(struct wim_dentry *dentry)
+{
+ calculate_dentry_full_path(dentry);
+ return dentry->_full_path;
+}
+
static int
increment_subdir_offset(struct wim_dentry *dentry, void *subdir_offset_p)
{
}
}
+/* Case-sensitive UTF-16LE dentry or stream name comparison. Used on both UNIX
+ * (always) and Windows (sometimes) */
static int
-dentry_compare_names(const struct wim_dentry *d1, const struct wim_dentry *d2)
+compare_utf16le_names_case_sensitive(const utf16lechar *name1, size_t nbytes1,
+ const utf16lechar *name2, size_t nbytes2)
{
- int result = memcmp(d1->file_name, d2->file_name,
- min(d1->file_name_nbytes, d2->file_name_nbytes));
+ /* Return the result if the strings differ up to their minimum length.
+ * Note that we cannot use strcmp() or strncmp() here, as the strings
+ * are in UTF-16LE format. */
+ int result = memcmp(name1, name2, min(nbytes1, nbytes2));
if (result)
return result;
+
+ /* The strings are the same up to their minimum length, so return a
+ * result based on their lengths. */
+ if (nbytes1 < nbytes2)
+ return -1;
+ else if (nbytes1 > nbytes2)
+ return 1;
else
- return d1->file_name_nbytes - d2->file_name_nbytes;
+ return 0;
}
+#ifdef __WIN32__
+/* Windoze: Case-insensitive UTF-16LE dentry or stream name comparison */
+static int
+compare_utf16le_names_case_insensitive(const utf16lechar *name1, size_t nbytes1,
+ const utf16lechar *name2, size_t nbytes2)
+{
+ /* Return the result if the strings differ up to their minimum length.
+ * */
+ int result = _wcsnicmp((const wchar_t*)name1, (const wchar_t*)name2,
+ min(nbytes1 / 2, nbytes2 / 2));
+ if (result)
+ return result;
+ /* The strings are the same up to their minimum length, so return a
+ * result based on their lengths. */
+ if (nbytes1 < nbytes2)
+ return -1;
+ else if (nbytes1 > nbytes2)
+ return 1;
+ else
+ return 0;
+}
+#endif /* __WIN32__ */
+
+#ifdef __WIN32__
+# define compare_utf16le_names compare_utf16le_names_case_insensitive
+#else
+# define compare_utf16le_names compare_utf16le_names_case_sensitive
+#endif
+
+
+#ifdef __WIN32__
+static int
+dentry_compare_names_case_insensitive(const struct wim_dentry *d1,
+ const struct wim_dentry *d2)
+{
+ return compare_utf16le_names_case_insensitive(d1->file_name,
+ d1->file_name_nbytes,
+ d2->file_name,
+ d2->file_name_nbytes);
+}
+#endif /* __WIN32__ */
+
+static int
+dentry_compare_names_case_sensitive(const struct wim_dentry *d1,
+ const struct wim_dentry *d2)
+{
+ return compare_utf16le_names_case_sensitive(d1->file_name,
+ d1->file_name_nbytes,
+ d2->file_name,
+ d2->file_name_nbytes);
+}
+
+#ifdef __WIN32__
+# define dentry_compare_names dentry_compare_names_case_insensitive
+#else
+# define dentry_compare_names dentry_compare_names_case_sensitive
+#endif
+
+/* Return %true iff the alternate data stream entry @entry has the UTF-16LE
+ * stream name @name that has length @name_nbytes bytes. */
+static inline bool
+ads_entry_has_name(const struct wim_ads_entry *entry,
+ const utf16lechar *name, size_t name_nbytes)
+{
+ return !compare_utf16le_names(name, name_nbytes,
+ entry->stream_name,
+ entry->stream_name_nbytes);
+}
+
+/* Given a UTF-16LE filename and a directory, look up the dentry for the file.
+ * Return it if found, otherwise NULL. This is case-sensitive on UNIX and
+ * case-insensitive on Windows. */
struct wim_dentry *
get_dentry_child_with_utf16le_name(const struct wim_dentry *dentry,
- const utf16lechar *name)
+ const utf16lechar *name,
+ size_t name_nbytes)
{
- struct rb_node *node = dentry->d_inode->i_children.rb_node;
+ struct rb_node *node;
+
+#ifdef __WIN32__
+ node = dentry->d_inode->i_children_case_insensitive.rb_node;
+#else
+ node = dentry->d_inode->i_children.rb_node;
+#endif
+
+ struct wim_dentry *child;
while (node) {
- struct wim_dentry *child = rbnode_dentry(node);
- int result = dentry_compare_names(dentry, child);
+ #ifdef __WIN32__
+ child = rb_entry(node, struct wim_dentry, rb_node_case_insensitive);
+ #else
+ child = rbnode_dentry(node);
+ #endif
+ int result = compare_utf16le_names(name, name_nbytes,
+ child->file_name,
+ child->file_name_nbytes);
if (result < 0)
node = node->rb_left;
else if (result > 0)
node = node->rb_right;
- else
+ else {
+ #ifdef __WIN32__
+ if (!list_empty(&child->case_insensitive_conflict_list))
+ {
+ WARNING("Result of case-insensitive lookup is ambiguous "
+ "(returning \"%ls\" instead of \"%ls\")",
+ child->file_name,
+ container_of(child->case_insensitive_conflict_list.next,
+ struct wim_dentry,
+ case_insensitive_conflict_list)->file_name);
+ }
+ #endif
return child;
+ }
}
return NULL;
}
/* Returns the child of @dentry that has the file name @name. Returns NULL if
* no child has the name. */
struct wim_dentry *
-get_dentry_child_with_name(const struct wim_dentry *dentry, const mbchar *name)
+get_dentry_child_with_name(const struct wim_dentry *dentry, const tchar *name)
{
+#if TCHAR_IS_UTF16LE
+ return get_dentry_child_with_utf16le_name(dentry, name,
+ tstrlen(name) * sizeof(tchar));
+#else
utf16lechar *utf16le_name;
size_t utf16le_name_nbytes;
int ret;
struct wim_dentry *child;
- ret = mbs_to_utf16le(name, strlen(name),
- &utf16le_name, &utf16le_name_nbytes);
+ ret = tstr_to_utf16le(name, tstrlen(name) * sizeof(tchar),
+ &utf16le_name, &utf16le_name_nbytes);
if (ret) {
child = NULL;
} else {
child = get_dentry_child_with_utf16le_name(dentry,
- utf16le_name);
+ utf16le_name,
+ utf16le_name_nbytes);
FREE(utf16le_name);
}
return child;
+#endif
}
-/* Returns the dentry corresponding to the @path, or NULL if there is no such
- * dentry. */
-struct wim_dentry *
-get_dentry(WIMStruct *w, const mbchar *path)
+static struct wim_dentry *
+get_dentry_utf16le(WIMStruct *wim, const utf16lechar *path)
{
- utf16lechar *path_utf16le;
- size_t path_utf16le_nbytes;
- int ret;
struct wim_dentry *cur_dentry, *parent_dentry;
- utf16lechar *p, *pp;
- utf16lechar save;
+ const utf16lechar *p, *pp;
- ret = mbs_to_utf16le(path, strlen(path), &path_utf16le,
- &path_utf16le_nbytes);
- if (ret) {
- errno = ENOMEM;
+ cur_dentry = parent_dentry = wim_root_dentry(wim);
+ if (!cur_dentry) {
+ errno = ENOENT;
return NULL;
}
-
- parent_dentry = wim_root_dentry(w);
- p = path_utf16le;
-
+ p = path;
while (1) {
- while (*p == '/')
+ while (*p == cpu_to_le16(WIM_PATH_SEPARATOR))
p++;
- cur_dentry = parent_dentry;
- if (*p == '\0')
+ if (*p == cpu_to_le16('\0'))
break;
pp = p;
- while (*pp != '/' && *pp != '\0')
+ while (*pp != cpu_to_le16(WIM_PATH_SEPARATOR) &&
+ *pp != cpu_to_le16('\0'))
pp++;
- save = *pp;
- *pp = '\0';
- cur_dentry = get_dentry_child_with_utf16le_name(parent_dentry, p);
+ cur_dentry = get_dentry_child_with_utf16le_name(parent_dentry, p,
+ (void*)pp - (void*)p);
if (cur_dentry == NULL)
break;
- *pp = save;
p = pp;
+ parent_dentry = cur_dentry;
}
- FREE(path_utf16le);
if (cur_dentry == NULL) {
if (dentry_is_directory(parent_dentry))
errno = ENOENT;
return cur_dentry;
}
+/* Returns the dentry corresponding to the @path, or NULL if there is no such
+ * dentry. */
+struct wim_dentry *
+get_dentry(WIMStruct *wim, const tchar *path)
+{
+#if TCHAR_IS_UTF16LE
+ return get_dentry_utf16le(wim, path);
+#else
+ utf16lechar *path_utf16le;
+ size_t path_utf16le_nbytes;
+ int ret;
+ struct wim_dentry *dentry;
+
+ ret = tstr_to_utf16le(path, tstrlen(path) * sizeof(tchar),
+ &path_utf16le, &path_utf16le_nbytes);
+ if (ret)
+ return NULL;
+ dentry = get_dentry_utf16le(wim, path_utf16le);
+ FREE(path_utf16le);
+ return dentry;
+#endif
+}
+
struct wim_inode *
-wim_pathname_to_inode(WIMStruct *w, const mbchar *path)
+wim_pathname_to_inode(WIMStruct *wim, const tchar *path)
{
struct wim_dentry *dentry;
- dentry = get_dentry(w, path);
+ dentry = get_dentry(wim, path);
if (dentry)
return dentry->d_inode;
else
return NULL;
}
+/* Takes in a path of length @len in @buf, and transforms it into a string for
+ * the path of its parent directory. */
+static void
+to_parent_name(tchar *buf, size_t len)
+{
+ ssize_t i = (ssize_t)len - 1;
+ while (i >= 0 && buf[i] == WIM_PATH_SEPARATOR)
+ i--;
+ while (i >= 0 && buf[i] != WIM_PATH_SEPARATOR)
+ i--;
+ while (i >= 0 && buf[i] == WIM_PATH_SEPARATOR)
+ i--;
+ buf[i + 1] = T('\0');
+}
+
/* Returns the dentry that corresponds to the parent directory of @path, or NULL
* if the dentry is not found. */
-struct wim_dentry *get_parent_dentry(WIMStruct *w, const mbchar *path)
+struct wim_dentry *
+get_parent_dentry(WIMStruct *wim, const tchar *path)
{
- size_t path_len = strlen(path);
- mbchar buf[path_len + 1];
+ size_t path_len = tstrlen(path);
+ tchar buf[path_len + 1];
- memcpy(buf, path, path_len + 1);
+ tmemcpy(buf, path, path_len + 1);
to_parent_name(buf, path_len);
- return get_dentry(w, buf);
+ return get_dentry(wim, buf);
}
/* Prints the full path of a dentry. */
-int print_dentry_full_path(struct wim_dentry *dentry, void *ignore)
+int
+print_dentry_full_path(struct wim_dentry *dentry, void *_ignore)
{
- if (dentry->full_path)
- puts(dentry->full_path);
+ int ret = calculate_dentry_full_path(dentry);
+ if (ret)
+ return ret;
+ tprintf(T("%"TS"\n"), dentry->_full_path);
return 0;
}
* set. */
struct file_attr_flag {
u32 flag;
- const char *name;
+ const tchar *name;
};
struct file_attr_flag file_attr_flags[] = {
- {FILE_ATTRIBUTE_READONLY, "READONLY"},
- {FILE_ATTRIBUTE_HIDDEN, "HIDDEN"},
- {FILE_ATTRIBUTE_SYSTEM, "SYSTEM"},
- {FILE_ATTRIBUTE_DIRECTORY, "DIRECTORY"},
- {FILE_ATTRIBUTE_ARCHIVE, "ARCHIVE"},
- {FILE_ATTRIBUTE_DEVICE, "DEVICE"},
- {FILE_ATTRIBUTE_NORMAL, "NORMAL"},
- {FILE_ATTRIBUTE_TEMPORARY, "TEMPORARY"},
- {FILE_ATTRIBUTE_SPARSE_FILE, "SPARSE_FILE"},
- {FILE_ATTRIBUTE_REPARSE_POINT, "REPARSE_POINT"},
- {FILE_ATTRIBUTE_COMPRESSED, "COMPRESSED"},
- {FILE_ATTRIBUTE_OFFLINE, "OFFLINE"},
- {FILE_ATTRIBUTE_NOT_CONTENT_INDEXED,"NOT_CONTENT_INDEXED"},
- {FILE_ATTRIBUTE_ENCRYPTED, "ENCRYPTED"},
- {FILE_ATTRIBUTE_VIRTUAL, "VIRTUAL"},
+ {FILE_ATTRIBUTE_READONLY, T("READONLY")},
+ {FILE_ATTRIBUTE_HIDDEN, T("HIDDEN")},
+ {FILE_ATTRIBUTE_SYSTEM, T("SYSTEM")},
+ {FILE_ATTRIBUTE_DIRECTORY, T("DIRECTORY")},
+ {FILE_ATTRIBUTE_ARCHIVE, T("ARCHIVE")},
+ {FILE_ATTRIBUTE_DEVICE, T("DEVICE")},
+ {FILE_ATTRIBUTE_NORMAL, T("NORMAL")},
+ {FILE_ATTRIBUTE_TEMPORARY, T("TEMPORARY")},
+ {FILE_ATTRIBUTE_SPARSE_FILE, T("SPARSE_FILE")},
+ {FILE_ATTRIBUTE_REPARSE_POINT, T("REPARSE_POINT")},
+ {FILE_ATTRIBUTE_COMPRESSED, T("COMPRESSED")},
+ {FILE_ATTRIBUTE_OFFLINE, T("OFFLINE")},
+ {FILE_ATTRIBUTE_NOT_CONTENT_INDEXED,T("NOT_CONTENT_INDEXED")},
+ {FILE_ATTRIBUTE_ENCRYPTED, T("ENCRYPTED")},
+ {FILE_ATTRIBUTE_VIRTUAL, T("VIRTUAL")},
};
/* Prints a directory entry. @lookup_table is a pointer to the lookup table, if
* available. If the dentry is unresolved and the lookup table is NULL, the
* lookup table entries will not be printed. Otherwise, they will be. */
-int print_dentry(struct wim_dentry *dentry, void *lookup_table)
+int
+print_dentry(struct wim_dentry *dentry, void *lookup_table)
{
const u8 *hash;
struct wim_lookup_table_entry *lte;
const struct wim_inode *inode = dentry->d_inode;
- char buf[50];
+ tchar buf[50];
- printf("[DENTRY]\n");
- printf("Length = %"PRIu64"\n", dentry->length);
- printf("Attributes = 0x%x\n", inode->i_attributes);
+ tprintf(T("[DENTRY]\n"));
+ tprintf(T("Length = %"PRIu64"\n"), dentry->length);
+ tprintf(T("Attributes = 0x%x\n"), inode->i_attributes);
for (size_t i = 0; i < ARRAY_LEN(file_attr_flags); i++)
if (file_attr_flags[i].flag & inode->i_attributes)
- printf(" FILE_ATTRIBUTE_%s is set\n",
+ tprintf(T(" FILE_ATTRIBUTE_%"TS" is set\n"),
file_attr_flags[i].name);
- printf("Security ID = %d\n", inode->i_security_id);
- printf("Subdir offset = %"PRIu64"\n", dentry->subdir_offset);
+ tprintf(T("Security ID = %d\n"), inode->i_security_id);
+ tprintf(T("Subdir offset = %"PRIu64"\n"), dentry->subdir_offset);
wim_timestamp_to_str(inode->i_creation_time, buf, sizeof(buf));
- printf("Creation Time = %s\n", buf);
+ tprintf(T("Creation Time = %"TS"\n"), buf);
wim_timestamp_to_str(inode->i_last_access_time, buf, sizeof(buf));
- printf("Last Access Time = %s\n", buf);
+ tprintf(T("Last Access Time = %"TS"\n"), buf);
wim_timestamp_to_str(inode->i_last_write_time, buf, sizeof(buf));
- printf("Last Write Time = %s\n", buf);
+ tprintf(T("Last Write Time = %"TS"\n"), buf);
- printf("Reparse Tag = 0x%"PRIx32"\n", inode->i_reparse_tag);
- printf("Hard Link Group = 0x%"PRIx64"\n", inode->i_ino);
- printf("Hard Link Group Size = %"PRIu32"\n", inode->i_nlink);
- printf("Number of Alternate Data Streams = %hu\n", inode->i_num_ads);
- wimlib_printf("Filename = \"%W\"\n", dentry->file_name);
- wimlib_printf("Short Name \"%W\"\n", dentry->short_name);
- printf("Full Path = \"%s\"\n", dentry->full_path);
+ if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
+ tprintf(T("Reparse Tag = 0x%"PRIx32"\n"), inode->i_reparse_tag);
+ tprintf(T("Reparse Point Flags = 0x%"PRIx16"\n"),
+ inode->i_not_rpfixed);
+ tprintf(T("Reparse Point Unknown 2 = 0x%"PRIx32"\n"),
+ inode->i_rp_unknown_2);
+ }
+ tprintf(T("Reparse Point Unknown 1 = 0x%"PRIx32"\n"),
+ inode->i_rp_unknown_1);
+ tprintf(T("Hard Link Group = 0x%"PRIx64"\n"), inode->i_ino);
+ tprintf(T("Hard Link Group Size = %"PRIu32"\n"), inode->i_nlink);
+ tprintf(T("Number of Alternate Data Streams = %hu\n"), inode->i_num_ads);
+ if (dentry_has_long_name(dentry))
+ wimlib_printf(T("Filename = \"%"WS"\"\n"), dentry->file_name);
+ if (dentry_has_short_name(dentry))
+ wimlib_printf(T("Short Name \"%"WS"\"\n"), dentry->short_name);
+ if (dentry->_full_path)
+ tprintf(T("Full Path = \"%"TS"\"\n"), dentry->_full_path);
lte = inode_stream_lte(dentry->d_inode, 0, lookup_table);
if (lte) {
} else {
hash = inode_stream_hash(inode, 0);
if (hash) {
- printf("Hash = 0x");
- print_hash(hash);
- putchar('\n');
- putchar('\n');
+ tprintf(T("Hash = 0x"));
+ print_hash(hash, stdout);
+ tputchar(T('\n'));
+ tputchar(T('\n'));
}
}
for (u16 i = 0; i < inode->i_num_ads; i++) {
- printf("[Alternate Stream Entry %u]\n", i);
- wimlib_printf("Name = \"%W\"\n", inode->i_ads_entries[i].stream_name);
- printf("Name Length (UTF16) = %u\n",
+ tprintf(T("[Alternate Stream Entry %u]\n"), i);
+ wimlib_printf(T("Name = \"%"WS"\"\n"),
+ inode->i_ads_entries[i].stream_name);
+ tprintf(T("Name Length (UTF16 bytes) = %hu\n"),
inode->i_ads_entries[i].stream_name_nbytes);
hash = inode_stream_hash(inode, i + 1);
if (hash) {
- printf("Hash = 0x");
- print_hash(hash);
- putchar('\n');
+ tprintf(T("Hash = 0x"));
+ print_hash(hash, stdout);
+ tputchar(T('\n'));
}
print_lookup_table_entry(inode_stream_lte(inode, i + 1, lookup_table),
stdout);
}
/* Initializations done on every `struct wim_dentry'. */
-static void dentry_common_init(struct wim_dentry *dentry)
+static void
+dentry_common_init(struct wim_dentry *dentry)
{
memset(dentry, 0, sizeof(struct wim_dentry));
- dentry->refcnt = 1;
}
-static struct wim_inode *new_timeless_inode()
+struct wim_inode *
+new_timeless_inode(void)
{
struct wim_inode *inode = CALLOC(1, sizeof(struct wim_inode));
if (inode) {
inode->i_security_id = -1;
inode->i_nlink = 1;
- #ifdef WITH_FUSE
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 inode;
}
-static struct wim_inode *new_inode()
+static struct wim_inode *
+new_inode(void)
{
struct wim_inode *inode = new_timeless_inode();
if (inode) {
return inode;
}
-/*
- * Creates an unlinked directory entry.
- *
- * @name: The UTF-8 filename of the new dentry.
- *
- * Returns a pointer to the new dentry, or NULL if out of memory.
- */
-struct wim_dentry *new_dentry(const mbchar *name)
+/* Creates an unlinked directory entry. */
+int
+new_dentry(const tchar *name, struct wim_dentry **dentry_ret)
{
struct wim_dentry *dentry;
+ int ret;
dentry = MALLOC(sizeof(struct wim_dentry));
if (!dentry)
- goto err;
+ return WIMLIB_ERR_NOMEM;
dentry_common_init(dentry);
- if (set_dentry_name(dentry, name) != 0)
- goto err;
-
- dentry->parent = dentry;
-
- return dentry;
-err:
- FREE(dentry);
- ERROR_WITH_ERRNO("Failed to create new dentry with name \"%s\"", name);
- return NULL;
+ ret = set_dentry_name(dentry, name);
+ if (ret == 0) {
+ dentry->parent = dentry;
+ *dentry_ret = dentry;
+ } else {
+ FREE(dentry);
+ ERROR("Failed to set name on new dentry with name \"%"TS"\"",
+ name);
+ }
+ return ret;
}
-static struct wim_dentry *
-__new_dentry_with_inode(const mbchar *name, bool timeless)
+static int
+_new_dentry_with_inode(const tchar *name, struct wim_dentry **dentry_ret,
+ bool timeless)
{
struct wim_dentry *dentry;
- dentry = new_dentry(name);
- if (dentry) {
- if (timeless)
- dentry->d_inode = new_timeless_inode();
- else
- dentry->d_inode = new_inode();
- if (dentry->d_inode) {
- inode_add_dentry(dentry, dentry->d_inode);
- } else {
- free_dentry(dentry);
- dentry = NULL;
- }
+ int ret;
+
+ ret = new_dentry(name, &dentry);
+ if (ret)
+ return ret;
+
+ if (timeless)
+ dentry->d_inode = new_timeless_inode();
+ else
+ dentry->d_inode = new_inode();
+ if (!dentry->d_inode) {
+ free_dentry(dentry);
+ return WIMLIB_ERR_NOMEM;
}
- return dentry;
+
+ inode_add_dentry(dentry, dentry->d_inode);
+ *dentry_ret = dentry;
+ return 0;
}
-struct wim_dentry *
-new_dentry_with_timeless_inode(const mbchar *name)
+int
+new_dentry_with_timeless_inode(const tchar *name, struct wim_dentry **dentry_ret)
{
- return __new_dentry_with_inode(name, true);
+ return _new_dentry_with_inode(name, dentry_ret, true);
}
-struct wim_dentry *
-new_dentry_with_inode(const mbchar *name)
+int
+new_dentry_with_inode(const tchar *name, struct wim_dentry **dentry_ret)
{
- return __new_dentry_with_inode(name, false);
+ return _new_dentry_with_inode(name, dentry_ret, false);
}
+int
+new_filler_directory(const tchar *name, struct wim_dentry **dentry_ret)
+{
+ int ret;
+ struct wim_dentry *dentry;
+
+ DEBUG("Creating filler directory \"%"TS"\"", name);
+ ret = new_dentry_with_inode(name, &dentry);
+ if (ret)
+ return ret;
+ /* Leave the inode number as 0; this is allowed for non
+ * hard-linked files. */
+ dentry->d_inode->i_resolved = 1;
+ dentry->d_inode->i_attributes = FILE_ATTRIBUTE_DIRECTORY;
+ *dentry_ret = dentry;
+ return 0;
+}
static int
-init_ads_entry(struct wim_ads_entry *ads_entry, const mbchar *name)
+init_ads_entry(struct wim_ads_entry *ads_entry, const void *name,
+ size_t name_nbytes, bool is_utf16le)
{
int ret = 0;
memset(ads_entry, 0, sizeof(*ads_entry));
- if (name && *name) {
- ret = get_utf16le_name(name, &ads_entry->stream_name,
- &ads_entry->stream_name_nbytes);
+
+ if (is_utf16le) {
+ utf16lechar *p = MALLOC(name_nbytes + sizeof(utf16lechar));
+ if (!p)
+ return WIMLIB_ERR_NOMEM;
+ memcpy(p, name, name_nbytes);
+ p[name_nbytes / 2] = cpu_to_le16(0);
+ ads_entry->stream_name = p;
+ ads_entry->stream_name_nbytes = name_nbytes;
+ } else {
+ if (name && *(const tchar*)name != T('\0')) {
+ ret = get_utf16le_name(name, &ads_entry->stream_name,
+ &ads_entry->stream_name_nbytes);
+ }
}
return ret;
}
}
/* Frees an inode. */
-void free_inode(struct wim_inode *inode)
+void
+free_inode(struct wim_inode *inode)
{
if (inode) {
if (inode->i_ads_entries) {
wimlib_assert(inode->i_num_opened_fds == 0);
FREE(inode->i_fds);
pthread_mutex_destroy(&inode->i_mutex);
- if (inode->i_hlist.pprev)
- hlist_del(&inode->i_hlist);
#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);
}
/* Decrements link count on an inode and frees it if the link count reaches 0.
* */
-static void put_inode(struct wim_inode *inode)
+static void
+put_inode(struct wim_inode *inode)
{
wimlib_assert(inode->i_nlink != 0);
if (--inode->i_nlink == 0) {
* The corresponding inode (if any) is freed only if its link count is
* decremented to 0.
*/
-void free_dentry(struct wim_dentry *dentry)
-{
- FREE(dentry->file_name);
- FREE(dentry->short_name);
- FREE(dentry->full_path);
- if (dentry->d_inode)
- put_inode(dentry->d_inode);
- FREE(dentry);
-}
-
-void put_dentry(struct wim_dentry *dentry)
+void
+free_dentry(struct wim_dentry *dentry)
{
- wimlib_assert(dentry->refcnt != 0);
- if (--dentry->refcnt == 0)
- free_dentry(dentry);
+ if (dentry) {
+ FREE(dentry->file_name);
+ FREE(dentry->short_name);
+ FREE(dentry->_full_path);
+ if (dentry->d_inode)
+ put_inode(dentry->d_inode);
+ FREE(dentry);
+ }
}
/* This function is passed as an argument to for_dentry_in_tree_depth() in order
* to free a directory tree. */
-static int do_free_dentry(struct wim_dentry *dentry, void *__lookup_table)
+static int
+do_free_dentry(struct wim_dentry *dentry, void *_lookup_table)
{
- struct wim_lookup_table *lookup_table = __lookup_table;
- unsigned i;
+ struct wim_lookup_table *lookup_table = _lookup_table;
if (lookup_table) {
- struct wim_lookup_table_entry *lte;
struct wim_inode *inode = dentry->d_inode;
- wimlib_assert(inode->i_nlink != 0);
- for (i = 0; i <= inode->i_num_ads; i++) {
+ for (unsigned i = 0; i <= inode->i_num_ads; i++) {
+ struct wim_lookup_table_entry *lte;
+
lte = inode_stream_lte(inode, i, lookup_table);
if (lte)
lte_decrement_refcnt(lte, lookup_table);
}
}
-
- put_dentry(dentry);
+ free_dentry(dentry);
return 0;
}
* table entries corresponding to the dentries will be
* decremented.
*/
-void free_dentry_tree(struct wim_dentry *root, struct wim_lookup_table *lookup_table)
+void
+free_dentry_tree(struct wim_dentry *root, struct wim_lookup_table *lookup_table)
{
- if (root)
- for_dentry_in_tree_depth(root, do_free_dentry, lookup_table);
+ for_dentry_in_tree_depth(root, do_free_dentry, lookup_table);
}
-int increment_dentry_refcnt(struct wim_dentry *dentry, void *ignore)
+#ifdef __WIN32__
+
+/* Insert a dentry into the case insensitive index for a directory.
+ *
+ * This is a red-black tree, but when multiple dentries share the same
+ * case-insensitive name, only one is inserted into the tree itself; the rest
+ * are connected in a list.
+ */
+static struct wim_dentry *
+dentry_add_child_case_insensitive(struct wim_dentry *parent,
+ struct wim_dentry *child)
{
- dentry->refcnt++;
- return 0;
+ struct rb_root *root;
+ struct rb_node **new;
+ struct rb_node *rb_parent;
+
+ root = &parent->d_inode->i_children_case_insensitive;
+ new = &root->rb_node;
+ rb_parent = NULL;
+ while (*new) {
+ struct wim_dentry *this = container_of(*new, struct wim_dentry,
+ rb_node_case_insensitive);
+ int result = dentry_compare_names_case_insensitive(child, this);
+
+ rb_parent = *new;
+
+ if (result < 0)
+ new = &((*new)->rb_left);
+ else if (result > 0)
+ new = &((*new)->rb_right);
+ else
+ return this;
+ }
+ rb_link_node(&child->rb_node_case_insensitive, rb_parent, new);
+ rb_insert_color(&child->rb_node_case_insensitive, root);
+ return NULL;
}
+#endif
/*
* Links a dentry into the directory tree.
*
- * @parent: The dentry that will be the parent of @dentry.
- * @dentry: The dentry to link.
+ * @parent: The dentry that will be the parent of @child.
+ * @child: The dentry to link.
+ *
+ * Returns NULL if successful. If @parent already contains a dentry with the
+ * same case-sensitive name as @child, the pointer to this duplicate dentry is
+ * returned.
*/
-bool
+struct wim_dentry *
dentry_add_child(struct wim_dentry * restrict parent,
struct wim_dentry * restrict child)
{
- wimlib_assert(dentry_is_directory(parent));
+ struct rb_root *root;
+ struct rb_node **new;
+ struct rb_node *rb_parent;
- struct rb_root *root = &parent->d_inode->i_children;
- struct rb_node **new = &(root->rb_node);
- struct rb_node *rb_parent = NULL;
+ wimlib_assert(dentry_is_directory(parent));
+ wimlib_assert(parent != child);
+ /* Case sensitive child dentry index */
+ root = &parent->d_inode->i_children;
+ new = &root->rb_node;
+ rb_parent = NULL;
while (*new) {
struct wim_dentry *this = rbnode_dentry(*new);
- int result = dentry_compare_names(child, this);
+ int result = dentry_compare_names_case_sensitive(child, this);
rb_parent = *new;
else if (result > 0)
new = &((*new)->rb_right);
else
- return false;
+ return this;
}
child->parent = parent;
rb_link_node(&child->rb_node, rb_parent, new);
rb_insert_color(&child->rb_node, root);
- return true;
+
+#ifdef __WIN32__
+ {
+ struct wim_dentry *existing;
+ existing = dentry_add_child_case_insensitive(parent, child);
+ if (existing) {
+ list_add(&child->case_insensitive_conflict_list,
+ &existing->case_insensitive_conflict_list);
+ child->rb_node_case_insensitive.__rb_parent_color = 0;
+ } else {
+ INIT_LIST_HEAD(&child->case_insensitive_conflict_list);
+ }
+ }
+#endif
+ return NULL;
}
/* Unlink a WIM dentry from the directory entry tree. */
unlink_dentry(struct wim_dentry *dentry)
{
struct wim_dentry *parent = dentry->parent;
+
if (parent == dentry)
return;
rb_erase(&dentry->rb_node, &parent->d_inode->i_children);
+#ifdef __WIN32__
+ if (dentry->rb_node_case_insensitive.__rb_parent_color) {
+ /* This dentry was in the case-insensitive red-black tree. */
+ rb_erase(&dentry->rb_node_case_insensitive,
+ &parent->d_inode->i_children_case_insensitive);
+ if (!list_empty(&dentry->case_insensitive_conflict_list)) {
+ /* Make a different case-insensitively-the-same dentry
+ * be the "representative" in the red-black tree. */
+ struct list_head *next;
+ struct wim_dentry *other;
+ struct wim_dentry *existing;
+
+ next = dentry->case_insensitive_conflict_list.next;
+ other = list_entry(next, struct wim_dentry, case_insensitive_conflict_list);
+ existing = dentry_add_child_case_insensitive(parent, other);
+ wimlib_assert(existing == NULL);
+ }
+ }
+ list_del(&dentry->case_insensitive_conflict_list);
+#endif
}
/*
* stream name @stream_name.
*/
struct wim_ads_entry *
-inode_get_ads_entry(struct wim_inode *inode, const mbchar *stream_name,
+inode_get_ads_entry(struct wim_inode *inode, const tchar *stream_name,
u16 *idx_ret)
{
if (inode->i_num_ads == 0) {
return NULL;
} else {
- int ret;
- utf16lechar *stream_name_utf16le;
size_t stream_name_utf16le_nbytes;
u16 i;
struct wim_ads_entry *result;
- ret = mbs_to_utf16le(stream_name, strlen(stream_name),
- &stream_name_utf16le,
- &stream_name_utf16le_nbytes);
- if (ret)
- return NULL;
+ #if TCHAR_IS_UTF16LE
+ const utf16lechar *stream_name_utf16le;
+ stream_name_utf16le = stream_name;
+ stream_name_utf16le_nbytes = tstrlen(stream_name) * sizeof(tchar);
+ #else
+ utf16lechar *stream_name_utf16le;
+
+ {
+ int ret = tstr_to_utf16le(stream_name,
+ tstrlen(stream_name) *
+ sizeof(tchar),
+ &stream_name_utf16le,
+ &stream_name_utf16le_nbytes);
+ if (ret)
+ return NULL;
+ }
+ #endif
i = 0;
result = NULL;
do {
break;
}
} while (++i != inode->i_num_ads);
+ #if !TCHAR_IS_UTF16LE
FREE(stream_name_utf16le);
+ #endif
return result;
}
}
-/*
- * Add an alternate stream entry to a WIM inode and return a pointer to it, or
- * NULL if memory could not be allocated.
- */
-struct wim_ads_entry *
-inode_add_ads(struct wim_inode *inode, const char *stream_name)
+static struct wim_ads_entry *
+do_inode_add_ads(struct wim_inode *inode, const void *stream_name,
+ size_t stream_name_nbytes, bool is_utf16le)
{
u16 num_ads;
struct wim_ads_entry *ads_entries;
struct wim_ads_entry *new_entry;
- DEBUG("Add alternate data stream \"%s\"", stream_name);
-
if (inode->i_num_ads >= 0xfffe) {
ERROR("Too many alternate data streams in one inode!");
return NULL;
inode->i_ads_entries = ads_entries;
new_entry = &inode->i_ads_entries[num_ads - 1];
- if (init_ads_entry(new_entry, stream_name) != 0)
+ if (init_ads_entry(new_entry, stream_name, stream_name_nbytes, is_utf16le))
return NULL;
-#ifdef WITH_FUSE
new_entry->stream_id = inode->i_next_stream_id++;
-#endif
inode->i_num_ads = num_ads;
return new_entry;
}
-int
-inode_add_ads_with_data(struct wim_inode *inode, const mbchar *name,
- const void *value, size_t size,
- struct wim_lookup_table *lookup_table)
+struct wim_ads_entry *
+inode_add_ads_utf16le(struct wim_inode *inode,
+ const utf16lechar *stream_name,
+ size_t stream_name_nbytes)
{
- int ret = WIMLIB_ERR_NOMEM;
- struct wim_ads_entry *new_ads_entry;
- struct wim_lookup_table_entry *existing_lte;
- struct wim_lookup_table_entry *lte;
- u8 value_hash[SHA1_HASH_SIZE];
+ DEBUG("Add alternate data stream \"%"WS"\"", stream_name);
+ return do_inode_add_ads(inode, stream_name, stream_name_nbytes, true);
+}
- wimlib_assert(inode->i_resolved);
- new_ads_entry = inode_add_ads(inode, name);
- if (!new_ads_entry)
- goto out;
- sha1_buffer((const u8*)value, size, value_hash);
- existing_lte = __lookup_resource(lookup_table, value_hash);
+/*
+ * Add an alternate stream entry to a WIM inode and return a pointer to it, or
+ * NULL if memory could not be allocated.
+ */
+struct wim_ads_entry *
+inode_add_ads(struct wim_inode *inode, const tchar *stream_name)
+{
+ DEBUG("Add alternate data stream \"%"TS"\"", stream_name);
+ return do_inode_add_ads(inode, stream_name,
+ tstrlen(stream_name) * sizeof(tchar),
+ TCHAR_IS_UTF16LE);
+}
+
+static struct wim_lookup_table_entry *
+add_stream_from_data_buffer(const void *buffer, size_t size,
+ struct wim_lookup_table *lookup_table)
+{
+ u8 hash[SHA1_HASH_SIZE];
+ struct wim_lookup_table_entry *lte, *existing_lte;
+
+ sha1_buffer(buffer, size, hash);
+ existing_lte = __lookup_resource(lookup_table, hash);
if (existing_lte) {
+ wimlib_assert(wim_resource_size(existing_lte) == size);
lte = existing_lte;
lte->refcnt++;
} else {
- u8 *value_copy;
+ void *buffer_copy;
lte = new_lookup_table_entry();
if (!lte)
- goto out_free_ads_entry;
- value_copy = MALLOC(size);
- if (!value_copy) {
- FREE(lte);
- goto out_free_ads_entry;
+ return NULL;
+ buffer_copy = memdup(buffer, size);
+ if (!buffer_copy) {
+ free_lookup_table_entry(lte);
+ return NULL;
}
- memcpy(value_copy, value, size);
lte->resource_location = RESOURCE_IN_ATTACHED_BUFFER;
- lte->attached_buffer = value_copy;
+ lte->attached_buffer = buffer_copy;
lte->resource_entry.original_size = size;
- lte->resource_entry.size = size;
- lte->resource_entry.flags = 0;
- copy_hash(lte->hash, value_hash);
+ copy_hash(lte->hash, hash);
lookup_table_insert(lookup_table, lte);
}
- new_ads_entry->lte = lte;
- ret = 0;
- goto out;
-out_free_ads_entry:
- inode_remove_ads(inode, new_ads_entry - inode->i_ads_entries,
- lookup_table);
-out:
- return ret;
+ return lte;
+}
+
+int
+inode_add_ads_with_data(struct wim_inode *inode, const tchar *name,
+ const void *value, size_t size,
+ struct wim_lookup_table *lookup_table)
+{
+ struct wim_ads_entry *new_ads_entry;
+
+ wimlib_assert(inode->i_resolved);
+
+ new_ads_entry = inode_add_ads(inode, name);
+ if (!new_ads_entry)
+ return WIMLIB_ERR_NOMEM;
+
+ new_ads_entry->lte = add_stream_from_data_buffer(value, size,
+ lookup_table);
+ if (!new_ads_entry->lte) {
+ inode_remove_ads(inode, new_ads_entry - inode->i_ads_entries,
+ lookup_table);
+ return WIMLIB_ERR_NOMEM;
+ }
+ return 0;
+}
+
+/* Set the unnamed stream of a WIM inode, given a data buffer containing the
+ * stream contents. */
+int
+inode_set_unnamed_stream(struct wim_inode *inode, const void *data, size_t len,
+ struct wim_lookup_table *lookup_table)
+{
+ inode->i_lte = add_stream_from_data_buffer(data, len, lookup_table);
+ if (!inode->i_lte)
+ return WIMLIB_ERR_NOMEM;
+ inode->i_resolved = 1;
+ return 0;
}
/* Remove an alternate data stream from a WIM inode */
ads_entry = &inode->i_ads_entries[idx];
- DEBUG("Remove alternate data stream \"%W\"", ads_entry->stream_name);
+ DEBUG("Remove alternate data stream \"%"WS"\"", ads_entry->stream_name);
lte = ads_entry->lte;
if (lte)
if (size != sizeof(struct wimlib_unix_data))
return BAD_UNIX_DATA;
- ret = read_full_wim_resource(lte, (u8*)unix_data, 0);
+ ret = read_full_resource_into_buf(lte, unix_data);
if (ret)
return ret;
if (which & UNIX_DATA_MODE || !have_good_unix_data)
unix_data.mode = mode;
ret = inode_add_ads_with_data(inode, WIMLIB_UNIX_DATA_TAG,
- (const u8*)&unix_data,
+ &unix_data,
sizeof(struct wimlib_unix_data),
lookup_table);
if (ret == 0 && have_unix_data)
* @p: Pointer to buffer that starts with the first alternate stream entry.
*
* @inode: Inode to load the alternate data streams into.
- * @inode->i_num_ads must have been set to the number of
- * alternate data streams that are expected.
+ * @inode->i_num_ads must have been set to the number of
+ * alternate data streams that are expected.
*
* @remaining_size: Number of bytes of data remaining in the buffer pointed
- * to by @p.
- *
- * The format of the on-disk alternate stream entries is as follows:
+ * to by @p.
*
- * struct wim_ads_entry_on_disk {
- * u64 length; // Length of the entry, in bytes. This includes
- * all fields (including the stream name and
- * null terminator if present, AND the padding!).
- * u64 reserved; // Seems to be unused
- * u8 hash[20]; // SHA1 message digest of the uncompressed stream
- * u16 stream_name_len; // Length of the stream name, in bytes
- * char stream_name[]; // Stream name in UTF-16LE, @stream_name_len bytes long,
- * not including null terminator
- * u16 zero; // UTF-16 null terminator for the stream name, NOT
- * included in @stream_name_len. Based on what
- * I've observed from filenames in dentries,
- * this field should not exist when
- * (@stream_name_len == 0), but you can't
- * actually tell because of the padding anyway
- * (provided that the padding is zeroed, which
- * it always seems to be).
- * char padding[]; // Padding to make the size a multiple of 8 bytes.
- * };
- *
- * In addition, the entries are 8-byte aligned.
*
* Return 0 on success or nonzero on failure. On success, inode->i_ads_entries
* is set to an array of `struct wim_ads_entry's of length inode->i_num_ads. On
* failure, @inode is not modified.
*/
static int
-read_ads_entries(const u8 *p, struct wim_inode *inode, u64 remaining_size)
+read_ads_entries(const u8 * restrict p, struct wim_inode * restrict inode,
+ size_t nbytes_remaining)
{
u16 num_ads;
struct wim_ads_entry *ads_entries;
int ret;
+ BUILD_BUG_ON(sizeof(struct wim_ads_entry_on_disk) != WIM_ADS_ENTRY_DISK_SIZE);
+
+ /* Allocate an array for our in-memory representation of the alternate
+ * data stream entries. */
num_ads = inode->i_num_ads;
ads_entries = CALLOC(num_ads, sizeof(inode->i_ads_entries[0]));
- if (!ads_entries) {
- ERROR("Could not allocate memory for %"PRIu16" "
- "alternate data stream entries", num_ads);
- return WIMLIB_ERR_NOMEM;
- }
+ if (!ads_entries)
+ goto out_of_memory;
+ /* Read the entries into our newly allocated buffer. */
for (u16 i = 0; i < num_ads; i++) {
- struct wim_ads_entry *cur_entry;
u64 length;
- u64 length_no_padding;
- u64 total_length;
- size_t utf8_len;
- const u8 *p_save = p;
+ struct wim_ads_entry *cur_entry;
+ const struct wim_ads_entry_on_disk *disk_entry =
+ (const struct wim_ads_entry_on_disk*)p;
cur_entry = &ads_entries[i];
-
- #ifdef WITH_FUSE
ads_entries[i].stream_id = i + 1;
- #endif
-
- /* Read the base stream entry, excluding the stream name. */
- if (remaining_size < WIM_ADS_ENTRY_DISK_SIZE) {
- ERROR("Stream entries go past end of metadata resource");
- ERROR("(remaining_size = %"PRIu64")", remaining_size);
- ret = WIMLIB_ERR_INVALID_DENTRY;
- goto out_free_ads_entries;
- }
-
- p = get_u64(p, &length);
- p += 8; /* Skip the reserved field */
- p = get_bytes(p, SHA1_HASH_SIZE, (u8*)cur_entry->hash);
- p = get_u16(p, &cur_entry->stream_name_nbytes);
-
- cur_entry->stream_name = NULL;
-
- /* Length including neither the null terminator nor the padding
- * */
- length_no_padding = WIM_ADS_ENTRY_DISK_SIZE +
- cur_entry->stream_name_nbytes;
-
- /* Length including the null terminator and the padding */
- total_length = ((length_no_padding + 2) + 7) & ~7;
-
- wimlib_assert(total_length == ads_entry_total_length(cur_entry));
-
- if (remaining_size < length_no_padding) {
- ERROR("Stream entries go past end of metadata resource");
- ERROR("(remaining_size = %"PRIu64" bytes, "
- "length_no_padding = %"PRIu64" bytes)",
- remaining_size, length_no_padding);
- ret = WIMLIB_ERR_INVALID_DENTRY;
- goto out_free_ads_entries;
- }
-
- /* The @length field in the on-disk ADS entry is expected to be
- * equal to @total_length, which includes all of the entry and
- * the padding that follows it to align the next ADS entry to an
- * 8-byte boundary. However, to be safe, we'll accept the
- * length field as long as it's not less than the un-padded
- * total length and not more than the padded total length. */
- if (length < length_no_padding || length > total_length) {
- ERROR("Stream entry has unexpected length "
- "field (length field = %"PRIu64", "
- "unpadded total length = %"PRIu64", "
- "padded total length = %"PRIu64")",
- length, length_no_padding, total_length);
- ret = WIMLIB_ERR_INVALID_DENTRY;
- goto out_free_ads_entries;
- }
+ /* Do we have at least the size of the fixed-length data we know
+ * need? */
+ if (nbytes_remaining < sizeof(struct wim_ads_entry_on_disk))
+ goto out_invalid;
+
+ /* Read the length field */
+ length = le64_to_cpu(disk_entry->length);
+
+ /* Make sure the length field is neither so small it doesn't
+ * include all the fixed-length data nor so large it overflows
+ * the metadata resource buffer. */
+ if (length < sizeof(struct wim_ads_entry_on_disk) ||
+ length > nbytes_remaining)
+ goto out_invalid;
+
+ /* Read the rest of the fixed-length data. */
+
+ cur_entry->reserved = le64_to_cpu(disk_entry->reserved);
+ copy_hash(cur_entry->hash, disk_entry->hash);
+ cur_entry->stream_name_nbytes = le16_to_cpu(disk_entry->stream_name_nbytes);
+
+ /* If stream_name_nbytes != 0, this is a named stream.
+ * Otherwise this is an unnamed stream, or in some cases (bugs
+ * in Microsoft's software I guess) a meaningless entry
+ * distinguished from the real unnamed stream entry, if any, by
+ * the fact that the real unnamed stream entry has a nonzero
+ * hash field. */
if (cur_entry->stream_name_nbytes) {
- cur_entry->stream_name = MALLOC((size_t)
- cur_entry->stream_name_nbytes + 2);
- if (!cur_entry->stream_name) {
- ret = WIMLIB_ERR_NOMEM;
- goto out_free_ads_entries;
- }
- get_bytes(p, cur_entry->stream_name_nbytes,
- cur_entry->stream_name);
- cur_entry->stream_name[cur_entry->stream_name_nbytes / 2] = 0;
+ /* The name is encoded in UTF16-LE, which uses 2-byte
+ * coding units, so the length of the name had better be
+ * an even number of bytes... */
+ if (cur_entry->stream_name_nbytes & 1)
+ goto out_invalid;
+
+ /* Add the length of the stream name to get the length
+ * we actually need to read. Make sure this isn't more
+ * than the specified length of the entry. */
+ if (sizeof(struct wim_ads_entry_on_disk) +
+ cur_entry->stream_name_nbytes > length)
+ goto out_invalid;
+
+ cur_entry->stream_name = MALLOC(cur_entry->stream_name_nbytes + 2);
+ if (!cur_entry->stream_name)
+ goto out_of_memory;
+
+ memcpy(cur_entry->stream_name,
+ disk_entry->stream_name,
+ cur_entry->stream_name_nbytes);
+ cur_entry->stream_name[cur_entry->stream_name_nbytes / 2] = cpu_to_le16(0);
}
+
/* It's expected that the size of every ADS entry is a multiple
* of 8. However, to be safe, I'm allowing the possibility of
* an ADS entry at the very end of the metadata resource ending
* un-aligned. So although we still need to increment the input
- * pointer by @total_length to reach the next ADS entry, it's
- * possible that less than @total_length is actually remaining
- * in the metadata resource. We should set the remaining size to
- * 0 bytes if this happens. */
- p = p_save + total_length;
- if (remaining_size < total_length)
- remaining_size = 0;
+ * pointer by @length to reach the next ADS entry, it's possible
+ * that less than @length is actually remaining in the metadata
+ * resource. We should set the remaining bytes to 0 if this
+ * happens. */
+ length = (length + 7) & ~(u64)7;
+ p += length;
+ if (nbytes_remaining < length)
+ nbytes_remaining = 0;
else
- remaining_size -= total_length;
+ nbytes_remaining -= length;
}
inode->i_ads_entries = ads_entries;
-#ifdef WITH_FUSE
inode->i_next_stream_id = inode->i_num_ads + 1;
-#endif
- return 0;
+ ret = 0;
+ goto out;
+out_of_memory:
+ ret = WIMLIB_ERR_NOMEM;
+ goto out_free_ads_entries;
+out_invalid:
+ ERROR("An alternate data stream entry is invalid");
+ ret = WIMLIB_ERR_INVALID_DENTRY;
out_free_ads_entries:
- for (u16 i = 0; i < num_ads; i++)
- destroy_ads_entry(&ads_entries[i]);
- FREE(ads_entries);
+ if (ads_entries) {
+ for (u16 i = 0; i < num_ads; i++)
+ destroy_ads_entry(&ads_entries[i]);
+ FREE(ads_entries);
+ }
+out:
return ret;
}
* Reads a WIM directory entry, including all alternate data stream entries that
* follow it, from the WIM image's metadata resource.
*
- * @metadata_resource: Buffer containing the uncompressed metadata resource.
- * @metadata_resource_len: Length of the metadata resource.
- * @offset: Offset of this directory entry in the metadata resource.
+ * @metadata_resource:
+ * Pointer to the metadata resource buffer.
+ *
+ * @metadata_resource_len:
+ * Length of the metadata resource buffer, in bytes.
+ *
+ * @offset: Offset of the dentry within the metadata resource.
+ *
* @dentry: A `struct wim_dentry' that will be filled in by this function.
*
* Return 0 on success or nonzero on failure. On failure, @dentry will have
* buffers. On success, the dentry->length field must be examined. If zero,
* this was a special "end of directory" dentry and not a real dentry. If
* nonzero, this was a real dentry.
+ *
+ * Possible errors include:
+ * WIMLIB_ERR_NOMEM
+ * WIMLIB_ERR_INVALID_DENTRY
*/
int
-read_dentry(const u8 metadata_resource[], u64 metadata_resource_len,
- u64 offset, struct wim_dentry *dentry)
+read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len,
+ u64 offset, struct wim_dentry * restrict dentry)
{
- const u8 *p;
+
u64 calculated_size;
- utf16lechar *file_name = NULL;
- utf16lechar *short_name = NULL;
+ utf16lechar *file_name;
+ utf16lechar *short_name;
u16 short_name_nbytes;
u16 file_name_nbytes;
int ret;
- struct wim_inode *inode = NULL;
+ struct wim_inode *inode;
+ const u8 *p = &metadata_resource[offset];
+ const struct wim_dentry_on_disk *disk_dentry =
+ (const struct wim_dentry_on_disk*)p;
+
+ BUILD_BUG_ON(sizeof(struct wim_dentry_on_disk) != WIM_DENTRY_DISK_SIZE);
+
+ if ((uintptr_t)p & 7)
+ WARNING("WIM dentry is not 8-byte aligned");
dentry_common_init(dentry);
- /*Make sure the dentry really fits into the metadata resource.*/
- if (offset + 8 > metadata_resource_len || offset + 8 < offset) {
+ /* Before reading the whole dentry, we need to read just the length.
+ * This is because a dentry of length 8 (that is, just the length field)
+ * terminates the list of sibling directory entries. */
+ if (offset + sizeof(u64) > metadata_resource_len ||
+ offset + sizeof(u64) < offset)
+ {
ERROR("Directory entry starting at %"PRIu64" ends past the "
"end of the metadata resource (size %"PRIu64")",
offset, metadata_resource_len);
return WIMLIB_ERR_INVALID_DENTRY;
}
-
- /* Before reading the whole dentry, we need to read just the length.
- * This is because a dentry of length 8 (that is, just the length field)
- * terminates the list of sibling directory entries. */
-
- p = get_u64(&metadata_resource[offset], &dentry->length);
+ dentry->length = le64_to_cpu(disk_dentry->length);
/* A zero length field (really a length of 8, since that's how big the
* directory entry is...) indicates that this is the end of directory
* dentry. We do not read it into memory as an actual dentry, so just
- * return successfully in that case. */
+ * return successfully in this case. */
+ if (dentry->length == 8)
+ dentry->length = 0;
if (dentry->length == 0)
return 0;
- /* If the dentry does not overflow the metadata resource buffer and is
- * not too short, read the rest of it (excluding the alternate data
- * streams, but including the file name and short name variable-length
- * fields) into memory. */
- if (offset + dentry->length >= metadata_resource_len
- || offset + dentry->length < offset)
+ /* Now that we have the actual length provided in the on-disk structure,
+ * again make sure it doesn't overflow the metadata resource buffer. */
+ if (offset + dentry->length > metadata_resource_len ||
+ offset + dentry->length < offset)
{
ERROR("Directory entry at offset %"PRIu64" and with size "
"%"PRIu64" ends past the end of the metadata resource "
return WIMLIB_ERR_INVALID_DENTRY;
}
- if (dentry->length < WIM_DENTRY_DISK_SIZE) {
+ /* Make sure the dentry length is at least as large as the number of
+ * fixed-length fields */
+ if (dentry->length < sizeof(struct wim_dentry_on_disk)) {
ERROR("Directory entry has invalid length of %"PRIu64" bytes",
- dentry->length);
+ entry->length);
return WIMLIB_ERR_INVALID_DENTRY;
}
+ /* Allocate a `struct wim_inode' for this `struct wim_dentry'. */
inode = new_timeless_inode();
if (!inode)
return WIMLIB_ERR_NOMEM;
- p = get_u32(p, &inode->i_attributes);
- p = get_u32(p, (u32*)&inode->i_security_id);
- p = get_u64(p, &dentry->subdir_offset);
+ /* Read more fields; some into the dentry, and some into the inode. */
- /* 2 unused fields */
- p += 2 * sizeof(u64);
- /*p = get_u64(p, &dentry->unused1);*/
- /*p = get_u64(p, &dentry->unused2);*/
+ inode->i_attributes = le32_to_cpu(disk_dentry->attributes);
+ inode->i_security_id = le32_to_cpu(disk_dentry->security_id);
+ dentry->subdir_offset = le64_to_cpu(disk_dentry->subdir_offset);
+ dentry->d_unused_1 = le64_to_cpu(disk_dentry->unused_1);
+ dentry->d_unused_2 = le64_to_cpu(disk_dentry->unused_2);
+ inode->i_creation_time = le64_to_cpu(disk_dentry->creation_time);
+ inode->i_last_access_time = le64_to_cpu(disk_dentry->last_access_time);
+ inode->i_last_write_time = le64_to_cpu(disk_dentry->last_write_time);
+ copy_hash(inode->i_hash, disk_dentry->unnamed_stream_hash);
- p = get_u64(p, &inode->i_creation_time);
- p = get_u64(p, &inode->i_last_access_time);
- p = get_u64(p, &inode->i_last_write_time);
-
- p = get_bytes(p, SHA1_HASH_SIZE, inode->i_hash);
-
- /*
- * I don't know what's going on here. It seems like M$ screwed up the
+ /* I don't know what's going on here. It seems like M$ screwed up the
* reparse points, then put the fields in the same place and didn't
- * document it. The WIM_HDR_FLAG_RP_FIX flag in the WIM header might
- * have something to do with this, but it's not documented.
- */
+ * document it. So we have some fields we read for reparse points, and
+ * some fields in the same place for non-reparse-point.s */
if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
- /* ??? */
- p += 4;
- p = get_u32(p, &inode->i_reparse_tag);
- p += 4;
+ inode->i_rp_unknown_1 = le32_to_cpu(disk_dentry->reparse.rp_unknown_1);
+ inode->i_reparse_tag = le32_to_cpu(disk_dentry->reparse.reparse_tag);
+ inode->i_rp_unknown_2 = le16_to_cpu(disk_dentry->reparse.rp_unknown_2);
+ inode->i_not_rpfixed = le16_to_cpu(disk_dentry->reparse.not_rpfixed);
+ /* Leave inode->i_ino at 0. Note that this means the WIM file
+ * cannot archive hard-linked reparse points. Such a thing
+ * doesn't really make sense anyway, although I believe it's
+ * theoretically possible to have them on NTFS. */
} else {
- p = get_u32(p, &inode->i_reparse_tag);
- p = get_u64(p, &inode->i_ino);
+ inode->i_rp_unknown_1 = le32_to_cpu(disk_dentry->nonreparse.rp_unknown_1);
+ inode->i_ino = le64_to_cpu(disk_dentry->nonreparse.hard_link_group_id);
}
- /* By the way, the reparse_reserved field does not actually exist (at
- * least when the file is not a reparse point) */
+ inode->i_num_ads = le16_to_cpu(disk_dentry->num_alternate_data_streams);
- p = get_u16(p, &inode->i_num_ads);
+ short_name_nbytes = le16_to_cpu(disk_dentry->short_name_nbytes);
+ file_name_nbytes = le16_to_cpu(disk_dentry->file_name_nbytes);
- p = get_u16(p, &short_name_nbytes);
- p = get_u16(p, &file_name_nbytes);
+ if ((short_name_nbytes & 1) | (file_name_nbytes & 1))
+ {
+ ERROR("Dentry name is not valid UTF-16LE (odd number of bytes)!");
+ ret = WIMLIB_ERR_INVALID_DENTRY;
+ goto out_free_inode;
+ }
/* We now know the length of the file name and short name. Make sure
* the length of the dentry is large enough to actually hold them.
* The calculated length here is unaligned to allow for the possibility
* that the dentry->length names an unaligned length, although this
* would be unexpected. */
- calculated_size = __dentry_correct_length_unaligned(file_name_nbytes,
- short_name_nbytes);
+ calculated_size = _dentry_correct_length_unaligned(file_name_nbytes,
+ short_name_nbytes);
if (dentry->length < calculated_size) {
ERROR("Unexpected end of directory entry! (Expected "
- "at least %"PRIu64" bytes, got %"PRIu64" bytes. "
- "short_name_nbytes = %hu, file_name_nbytes = %hu)",
- calculated_size, dentry->length,
- short_name_nbytes, file_name_nbytes);
+ "at least %"PRIu64" bytes, got %"PRIu64" bytes.)",
+ calculated_size, dentry->length);
ret = WIMLIB_ERR_INVALID_DENTRY;
goto out_free_inode;
}
+ p += sizeof(struct wim_dentry_on_disk);
+
/* Read the filename if present. Note: if the filename is empty, there
* is no null terminator following it. */
if (file_name_nbytes) {
- file_name = MALLOC((size_t)file_name_nbytes + 2);
+ file_name = MALLOC(file_name_nbytes + 2);
if (!file_name) {
- ERROR("Failed to allocate %zu bytes for dentry file name",
- (size_t)file_name_nbytes + 2);
+ ERROR("Failed to allocate %d bytes for dentry file name",
+ file_name_nbytes + 2);
ret = WIMLIB_ERR_NOMEM;
goto out_free_inode;
}
- p = get_bytes(p, (size_t)file_name_nbytes + 2, file_name);
- if (file_name[file_name_nbytes / 2] != 0) {
- file_name[file_name_nbytes / 2] = 0;
- WARNING("File name in WIM dentry \"%W\" is not "
- "null-terminated!", file_name);
- }
+ memcpy(file_name, p, file_name_nbytes);
+ p += file_name_nbytes + 2;
+ file_name[file_name_nbytes / 2] = cpu_to_le16(0);
+ } else {
+ file_name = NULL;
}
- /* Align the calculated size */
- calculated_size = (calculated_size + 7) & ~7;
-
- if (dentry->length > calculated_size) {
- /* Weird; the dentry says it's longer than it should be. Note
- * that the length field does NOT include the size of the
- * alternate stream entries. */
-
- /* Strangely, some directory entries inexplicably have a little
- * over 70 bytes of extra data. The exact amount of data seems
- * to be 72 bytes, but it is aligned on the next 8-byte
- * boundary. It does NOT seem to be alternate data stream
- * entries. Here's an example of the aligned data:
- *
- * 01000000 40000000 6c786bba c58ede11 b0bb0026 1870892a b6adb76f
- * e63a3e46 8fca8653 0d2effa1 6c786bba c58ede11 b0bb0026 1870892a
- * 00000000 00000000 00000000 00000000
- *
- * Here's one interpretation of how the data is laid out.
- *
- * struct unknown {
- * u32 field1; (always 0x00000001)
- * u32 field2; (always 0x40000000)
- * u8 data[48]; (???)
- * u64 reserved1; (always 0)
- * u64 reserved2; (always 0)
- * };*/
- /*DEBUG("Dentry for file or directory `%s' has %"PRIu64" extra "*/
- /*"bytes of data",*/
- /*file_name_utf8, dentry->length - calculated_size);*/
- }
/* Read the short filename if present. Note: if there is no short
* filename, there is no null terminator following it. */
if (short_name_nbytes) {
- short_name = MALLOC((size_t)short_name_nbytes + 2);
+ short_name = MALLOC(short_name_nbytes + 2);
if (!short_name) {
- ERROR("Failed to allocate %zu bytes for dentry short name",
- (size_t)short_name_nbytes + 2);
+ ERROR("Failed to allocate %d bytes for dentry short name",
+ short_name_nbytes + 2);
ret = WIMLIB_ERR_NOMEM;
goto out_free_file_name;
}
- p = get_bytes(p, (size_t)short_name_nbytes + 2, short_name);
- if (short_name[short_name_nbytes / 2] != 0) {
- short_name[short_name_nbytes / 2] = 0;
- WARNING("Short name in WIM dentry \"%W\" is not "
- "null-terminated!", file_name);
- }
+ memcpy(short_name, p, short_name_nbytes);
+ p += short_name_nbytes + 2;
+ short_name[short_name_nbytes / 2] = cpu_to_le16(0);
+ } else {
+ short_name = NULL;
}
+ /* Align the dentry length */
+ dentry->length = (dentry->length + 7) & ~7;
+
/*
* Read the alternate data streams, if present. dentry->num_ads tells
* us how many they are, and they will directly follow the dentry
* on-disk.
*
* Note that each alternate data stream entry begins on an 8-byte
- * aligned boundary, and the alternate data stream entries are NOT
- * included in the dentry->length field for some reason.
+ * aligned boundary, and the alternate data stream entries seem to NOT
+ * be included in the dentry->length field for some reason.
*/
if (inode->i_num_ads != 0) {
-
- /* Trying different lengths is just a hack to make sure we have
- * a chance of reading the ADS entries correctly despite the
- * poor documentation. */
-
- if (calculated_size != dentry->length) {
- WARNING("Trying calculated dentry length (%"PRIu64") "
- "instead of dentry->length field (%"PRIu64") "
- "to read ADS entries",
- calculated_size, dentry->length);
- }
- u64 lengths_to_try[3] = {calculated_size,
- (dentry->length + 7) & ~7,
- dentry->length};
ret = WIMLIB_ERR_INVALID_DENTRY;
- for (size_t i = 0; i < ARRAY_LEN(lengths_to_try); i++) {
- if (lengths_to_try[i] > metadata_resource_len - offset)
- continue;
- ret = read_ads_entries(&metadata_resource[offset + lengths_to_try[i]],
- inode,
- metadata_resource_len - offset - lengths_to_try[i]);
- if (ret == 0)
- goto out;
+ if (offset + dentry->length > metadata_resource_len ||
+ (ret = read_ads_entries(&metadata_resource[offset + dentry->length],
+ inode,
+ metadata_resource_len - offset - dentry->length)))
+ {
+ ERROR("Failed to read alternate data stream "
+ "entries of WIM dentry \"%"WS"\"", file_name);
+ goto out_free_short_name;
}
- ERROR("Failed to read alternate data stream "
- "entries of WIM dentry \"%W\"", file_name);
- goto out_free_short_name;
}
-out:
/* We've read all the data for this dentry. Set the names and their
* lengths, and we've done. */
dentry->d_inode = inode;
dentry->short_name = short_name;
dentry->file_name_nbytes = file_name_nbytes;
dentry->short_name_nbytes = short_name_nbytes;
- return 0;
+ ret = 0;
+ goto out;
out_free_short_name:
FREE(short_name);
out_free_file_name:
FREE(file_name);
out_free_inode:
free_inode(inode);
+out:
return ret;
}
+static const tchar *
+dentry_get_file_type_string(const struct wim_dentry *dentry)
+{
+ const struct wim_inode *inode = dentry->d_inode;
+ if (inode_is_directory(inode))
+ return T("directory");
+ else if (inode_is_symlink(inode))
+ return T("symbolic link");
+ else
+ return T("file");
+}
+
/* Reads the children of a dentry, and all their children, ..., etc. from the
* metadata resource and into the dentry tree.
*
{
u64 cur_offset = dentry->subdir_offset;
struct wim_dentry *child;
+ struct wim_dentry *duplicate;
+ struct wim_dentry *parent;
struct wim_dentry cur_child;
int ret;
if (cur_offset == 0)
return 0;
+ /* Check for cyclic directory structure */
+ for (parent = dentry->parent; !dentry_is_root(parent); parent = parent->parent)
+ {
+ if (unlikely(parent->subdir_offset == cur_offset)) {
+ ERROR("Cyclic directory structure directed: children "
+ "of \"%"TS"\" coincide with children of \"%"TS"\"",
+ dentry_full_path(dentry),
+ dentry_full_path(parent));
+ return WIMLIB_ERR_INVALID_DENTRY;
+ }
+ }
+
/* Find and read all the children of @dentry. */
- while (1) {
+ for (;;) {
/* Read next child of @dentry into @cur_child. */
ret = read_dentry(metadata_resource, metadata_resource_len,
cur_offset, &cur_child);
- if (ret != 0)
+ if (ret)
break;
/* Check for end of directory. */
/* Not end of directory. Allocate this child permanently and
* link it to the parent and previous child. */
- child = MALLOC(sizeof(struct wim_dentry));
+ child = memdup(&cur_child, sizeof(struct wim_dentry));
if (!child) {
- ERROR("Failed to allocate %zu bytes for new dentry",
- sizeof(struct wim_dentry));
+ ERROR("Failed to allocate new dentry!");
ret = WIMLIB_ERR_NOMEM;
break;
}
- memcpy(child, &cur_child, sizeof(struct wim_dentry));
- dentry_add_child(dentry, child);
- inode_add_dentry(child, child->d_inode);
-
- /* If there are children of this child, call this procedure
- * recursively. */
- if (child->subdir_offset != 0) {
- ret = read_dentry_tree(metadata_resource,
- metadata_resource_len, child);
- if (ret != 0)
- break;
- }
/* Advance to the offset of the next child. Note: We need to
* advance by the TOTAL length of the dentry, not by the length
- * child->length, which although it does take into account the
- * padding, it DOES NOT take into account alternate stream
+ * cur_child.length, which although it does take into account
+ * the padding, it DOES NOT take into account alternate stream
* 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_get_file_type_string(child);
+ duplicate_type = dentry_get_file_type_string(duplicate);
+ WARNING("Ignoring duplicate %"TS" \"%"TS"\" "
+ "(the WIM image already contains a %"TS" "
+ "at that path with the exact same name)",
+ child_type, dentry_full_path(duplicate),
+ duplicate_type);
+ free_dentry(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;
}
*
* @dentry: The dentry structure.
* @p: The memory location to write the data to.
- * @return: Pointer to the byte after the last byte we wrote as part of the
- * dentry.
+ *
+ * Returns the pointer to the byte after the last byte we wrote as part of the
+ * dentry, including any alternate data stream entries.
*/
static u8 *
-write_dentry(const struct wim_dentry *dentry, u8 *p)
+write_dentry(const struct wim_dentry * restrict dentry, u8 * restrict p)
{
- u8 *orig_p = p;
+ const struct wim_inode *inode;
+ struct wim_dentry_on_disk *disk_dentry;
+ const u8 *orig_p;
const u8 *hash;
- const struct wim_inode *inode = dentry->d_inode;
- /* We calculate the correct length of the dentry ourselves because the
- * dentry->length field may been set to an unexpected value from when we
- * read the dentry in (for example, there may have been unknown data
- * appended to the end of the dentry...) */
- u64 length = dentry_correct_length(dentry);
-
- p = put_u64(p, length);
- p = put_u32(p, inode->i_attributes);
- p = put_u32(p, inode->i_security_id);
- p = put_u64(p, dentry->subdir_offset);
- p = put_u64(p, 0); /* unused1 */
- p = put_u64(p, 0); /* unused2 */
- p = put_u64(p, inode->i_creation_time);
- p = put_u64(p, inode->i_last_access_time);
- p = put_u64(p, inode->i_last_write_time);
+ wimlib_assert(((uintptr_t)p & 7) == 0); /* 8 byte aligned */
+ orig_p = p;
+
+ inode = dentry->d_inode;
+ disk_dentry = (struct wim_dentry_on_disk*)p;
+
+ disk_dentry->attributes = cpu_to_le32(inode->i_attributes);
+ disk_dentry->security_id = cpu_to_le32(inode->i_security_id);
+ disk_dentry->subdir_offset = cpu_to_le64(dentry->subdir_offset);
+ disk_dentry->unused_1 = cpu_to_le64(dentry->d_unused_1);
+ disk_dentry->unused_2 = cpu_to_le64(dentry->d_unused_2);
+ disk_dentry->creation_time = cpu_to_le64(inode->i_creation_time);
+ disk_dentry->last_access_time = cpu_to_le64(inode->i_last_access_time);
+ disk_dentry->last_write_time = cpu_to_le64(inode->i_last_write_time);
hash = inode_stream_hash(inode, 0);
- p = put_bytes(p, SHA1_HASH_SIZE, hash);
+ copy_hash(disk_dentry->unnamed_stream_hash, hash);
if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
- p = put_zeroes(p, 4);
- p = put_u32(p, inode->i_reparse_tag);
- p = put_zeroes(p, 4);
+ disk_dentry->reparse.rp_unknown_1 = cpu_to_le32(inode->i_rp_unknown_1);
+ disk_dentry->reparse.reparse_tag = cpu_to_le32(inode->i_reparse_tag);
+ disk_dentry->reparse.rp_unknown_2 = cpu_to_le16(inode->i_rp_unknown_2);
+ disk_dentry->reparse.not_rpfixed = cpu_to_le16(inode->i_not_rpfixed);
} else {
- u64 link_group_id;
- p = put_u32(p, 0);
- if (inode->i_nlink == 1)
- link_group_id = 0;
- else
- link_group_id = inode->i_ino;
- p = put_u64(p, link_group_id);
- }
- p = put_u16(p, inode->i_num_ads);
- p = put_u16(p, dentry->short_name_nbytes);
- p = put_u16(p, dentry->file_name_nbytes);
- if (dentry->file_name_nbytes) {
- p = put_bytes(p, dentry->file_name_nbytes + 2,
- dentry->file_name);
- }
- if (dentry->short_name_nbytes) {
- p = put_bytes(p, dentry->short_name_nbytes + 2,
- dentry->short_name);
+ disk_dentry->nonreparse.rp_unknown_1 = cpu_to_le32(inode->i_rp_unknown_1);
+ disk_dentry->nonreparse.hard_link_group_id =
+ cpu_to_le64((inode->i_nlink == 1) ? 0 : inode->i_ino);
}
+ disk_dentry->num_alternate_data_streams = cpu_to_le16(inode->i_num_ads);
+ disk_dentry->short_name_nbytes = cpu_to_le16(dentry->short_name_nbytes);
+ 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);
+
+ if (dentry_has_short_name(dentry))
+ p = mempcpy(p, dentry->short_name, dentry->short_name_nbytes + 2);
/* Align to 8-byte boundary */
- wimlib_assert(length >= (p - orig_p) && length - (p - orig_p) <= 7);
- p = put_zeroes(p, length - (p - orig_p));
+ while ((uintptr_t)p & 7)
+ *p++ = 0;
- /* Write the alternate data streams, if there are any. Please see
- * read_ads_entries() for comments about the format of the on-disk
- * alternate data stream entries. */
+ /* We calculate the correct length of the dentry ourselves because the
+ * dentry->length field may been set to an unexpected value from when we
+ * read the dentry in (for example, there may have been unknown data
+ * appended to the end of the dentry...). Furthermore, the dentry may
+ * have been renamed, thus changing its needed length. */
+ disk_dentry->length = cpu_to_le64(p - orig_p);
+
+ /* Write the alternate data streams entries, if any. */
for (u16 i = 0; i < inode->i_num_ads; i++) {
- p = put_u64(p, ads_entry_total_length(&inode->i_ads_entries[i]));
- p = put_u64(p, 0); /* Unused */
+ const struct wim_ads_entry *ads_entry =
+ &inode->i_ads_entries[i];
+ struct wim_ads_entry_on_disk *disk_ads_entry =
+ (struct wim_ads_entry_on_disk*)p;
+ orig_p = p;
+
+ disk_ads_entry->reserved = cpu_to_le64(ads_entry->reserved);
+
hash = inode_stream_hash(inode, i + 1);
- p = put_bytes(p, SHA1_HASH_SIZE, hash);
- p = put_u16(p, inode->i_ads_entries[i].stream_name_nbytes);
- if (inode->i_ads_entries[i].stream_name_nbytes) {
- p = put_bytes(p,
- inode->i_ads_entries[i].stream_name_nbytes + 2,
- inode->i_ads_entries[i].stream_name);
+ copy_hash(disk_ads_entry->hash, hash);
+ disk_ads_entry->stream_name_nbytes = cpu_to_le16(ads_entry->stream_name_nbytes);
+ p += sizeof(struct wim_ads_entry_on_disk);
+ if (ads_entry->stream_name_nbytes) {
+ p = mempcpy(p, ads_entry->stream_name,
+ ads_entry->stream_name_nbytes + 2);
}
- p = put_zeroes(p, (8 - (p - orig_p) % 8) % 8);
+ /* Align to 8-byte boundary */
+ while ((uintptr_t)p & 7)
+ *p++ = 0;
+ disk_ads_entry->length = cpu_to_le64(p - orig_p);
}
- wimlib_assert(p - orig_p == __dentry_total_length(dentry, length));
return p;
}
for_dentry_child(parent, write_dentry_cb, &p);
/* write end of directory entry */
- p = put_u64(p, 0);
+ *(le64*)p = cpu_to_le64(0);
+ p += 8;
/* Recurse on children. */
for_dentry_child(parent, write_dentry_tree_recursive_cb, &p);
/* Write end of directory entry after the root dentry just to be safe;
* however the root dentry obviously cannot have any siblings. */
- p = put_u64(p, 0);
+ *(le64*)p = cpu_to_le64(0);
+ p += 8;
/* Recursively write the rest of the dentry tree. */
return write_dentry_tree_recursive(root, p);