def _parse_indirect_block(stream, value):
# This is an indirect datablock (Bigger than 16344, therefor we handle it differently)
# The value inside the vk entry actually contains a pointer to the buffers containing the data
big_data_block_header = BIG_DATA_BLOCK.parse(value.value)
# Go to the start of the segment offset list
stream.seek(REGF_HEADER_SIZE +
big_data_block_header.offset_to_list_of_segments)
buffer = BytesIO()
# Read them sequentially until we got all the size of the VK
value_size = value.size
while value_size > 0:
data_segment_offset = Int32ul.parse_stream(stream)
with boomerang_stream(stream) as tmpstream:
tmpstream.seek(REGF_HEADER_SIZE + 4 + data_segment_offset)
tmpbuffer = tmpstream.read(min(0x3fd8, value_size))
value_size -= len(tmpbuffer)
buffer.write(tmpbuffer)
buffer.seek(0)
return buffer.read()
def iter_values(self, as_json=False, max_len=MAX_LEN):
"""
Get the values of a subkey. Will raise if no values exist
:param as_json: Whether to normalize the data as JSON or not
:param max_len: Max length of value to return
:return: List of values for the subkey
"""
if not self.values_count:
return
# Get the offset of the values key. We skip 4 because of Cell Header
target_offset = REGF_HEADER_SIZE + 4 + self.header.values_list_offset
self._stream.seek(target_offset)
for _ in range(self.values_count):
is_corrupted = False
try:
vk_offset = Int32ul.parse_stream(self._stream)
except StreamError:
logger.info(
f'Skipping bad registry VK at {self._stream.tell()}')
raise RegistryParsingException(
f'Bad registry VK at {self._stream.tell()}')
with boomerang_stream(self._stream) as substream:
actual_vk_offset = REGF_HEADER_SIZE + 4 + vk_offset
substream.seek(actual_vk_offset)
try:
vk = VALUE_KEY.parse_stream(substream)
except (ConstError, StreamError):
logger.error(
f'Could not parse VK at {substream.tell()}, registry hive is probably corrupted.'
)
return
value = self.read_value(vk, substream)
if vk.name_size == 0:
value_name = '(default)'
elif vk.flags.VALUE_COMP_NAME:
# Compressed (ASCII) value name
value_name = vk.name.decode('ascii', errors='replace')
else:
# Unicode (UTF-16) value name
value_name = vk.name.decode('utf-16-le', errors='replace')
logger.debug(
f'Unicode value name identified: "{value_name}"')
# If the value is bigger than this value, it means this is a DEVPROP structure
# https://doxygen.reactos.org/d0/dba/devpropdef_8h_source.html
# https://sourceforge.net/p/mingw-w64/mingw-w64/ci/668a1d3e85042c409e0c292e621b3dc0aa26177c/tree/
# mingw-w64-headers/include/devpropdef.h?diff=dd86a3b7594dadeef9d6a37c4b6be3ca42ef7e94
# We currently do not support these, but also wouldn't like to yield this as binary data
# This int casting will always work because the data_type is construct's EnumIntegerString
# TODO: Add actual parsing
if int(vk.data_type) > 0xffff0000:
data_type = int(vk.data_type) & 0xffff
continue
# Skip this unknown data type, research pending :)
# TODO: Add actual parsing
if int(vk.data_type) == 0x200000:
continue
data_type = str(vk.data_type)
if data_type in ['REG_SZ', 'REG_EXPAND', 'REG_EXPAND_SZ']:
if vk.data_size >= 0x80000000:
# data is contained in the data_offset field
value.size -= 0x80000000
actual_value = vk.data_offset
elif vk.data_size > 0x3fd8 and value.value[:2] == b'db':
data = self._parse_indirect_block(substream, value)
actual_value = try_decode_binary(data, as_json=as_json)
else:
actual_value = try_decode_binary(value.value,
as_json=as_json)
elif data_type in ['REG_BINARY', 'REG_NONE']:
if vk.data_size >= 0x80000000:
# data is contained in the data_offset field
actual_value = vk.data_offset
elif vk.data_size > 0x3fd8 and value.value[:2] == b'db':
try:
actual_value = self._parse_indirect_block(
substream, value)
actual_value = try_decode_binary(
actual_value,
as_json=True) if as_json else actual_value
except ConstError:
logger.error(f'Bad value at {actual_vk_offset}')
continue
else:
# Return the actual data
actual_value = binascii.b2a_hex(value.value).decode(
)[:max_len] if as_json else value.value
elif data_type == 'REG_SZ':
actual_value = try_decode_binary(value.value,
as_json=as_json)
elif data_type == 'REG_DWORD':
# If the data size is bigger than 0x80000000, data is actually stored in the VK data offset.
actual_value = vk.data_offset if vk.data_size >= 0x80000000 else Int32ul.parse(
value.value)
elif data_type == 'REG_QWORD':
actual_value = vk.data_offset if vk.data_size >= 0x80000000 else Int64ul.parse(
value.value)
elif data_type == 'REG_MULTI_SZ':
parsed_value = GreedyRange(CString('utf-16-le')).parse(
value.value)
# Because the ListContainer object returned by Construct cannot be turned into a list,
# we do this trick
actual_value = [x for x in parsed_value if x]
# We currently dumps this as hex string or raw
# TODO: Add actual parsing
elif data_type in [
'REG_RESOURCE_REQUIREMENTS_LIST', 'REG_RESOURCE_LIST'
]:
actual_value = binascii.b2a_hex(value.value).decode(
)[:max_len] if as_json else value.value
else:
actual_value = try_decode_binary(value.value,
as_json=as_json)
yield Value(name=value_name,
value_type=str(value.value_type),
value=actual_value,
is_corrupted=is_corrupted)