def slice_nibbles(data: bytes, start_nibble: int, size: int = 1) -> int:
"""Slice out integer value of bytes indexed by nibble instead of byte.
This function is only designed to work with current instruction formats. It
makes a number of assumptions about byte order and positioning for these
specific cases.
"""
if size == 1:
# Single nibble
return int((data[start_nibble // 2] >> (((start_nibble + 1) % 2) * 4))
& 0xF)
elif size == 2:
# Single byte, assuming byte-alignment
return data[start_nibble // 2]
elif size == 4:
# Normal 2-byte value, assuming byte-alignment
return (data[start_nibble // 2] << 8) + data[start_nibble // 2 + 1]
elif size == 8 or size == 16:
# The 2-byte values are ordered from low to high
res = 0
for i, nibble in enumerate(range(start_nibble, start_nibble + size,
4)):
res += (
(data[nibble // 2] << 8) + data[nibble // 2 + 1]) << (i * 16)
return res
else:
log_error(
f"slice_nibbles called with unexpected size: {size}. Returning 0")
return 0
def parse_proto_ids(
self, data: bytes, size: int, offset_to_section: FileOffset
) -> None:
i_offset = (4 - (offset_to_section)) % 4
self.proto_ids: List[DexProtoId] = [
DexProtoId(
shorty=self.strings[self._parse_uint(data[i : i + 4])],
return_type=self.type_ids[self._parse_uint(data[i + 4 : i + 8])],
parameters=self.type_lists[
cast(FileOffset, self._parse_uint(data[i + 8 : i + 12]))
]
if self._parse_uint(data[i + 8 : i + 12])
else list(),
)
for i in range(i_offset, size * 12 + i_offset, 12)
]
for proto in self.proto_ids:
if len(proto.shorty) - 1 != len(proto.parameters):
log_error("Shorty does not match parameters")
def make_strings(self, data: bytes) -> None:
self.strings: List[str] = list()
for string_data_off in self.string_ids:
utf16_size, off = parse_uleb128(data[string_data_off : string_data_off + 5])
try:
string, string_size_off = parse_mutf8(data[string_data_off + off :])
except UnicodeDecodeError:
# This should never be reached
t = data[
string_data_off
+ off : string_data_off
+ off
+ data[string_data_off + off :].index(b"\x00")
]
log_error(f"Failed to decode MUTF8: {t!r}")
raise
self.strings.append(string)
plen = len(string.encode("utf-16", "surrogatepass")) // 2 - 1
if plen != utf16_size:
# This should never be reached
log_error(
f'String {repr(string)} at string offset "{string_data_off}" Python length {plen} does not match expected length {utf16_size}'
)
def disassemble_pseudoinstructions(data: bytes,
addr: "FileOffset") -> PseudoInstructions:
# Static variable
if "insns" not in disassemble.__dict__:
disassemble.insns = load_insns() # type: ignore[attr-defined]
pseudoinstructions: PseudoInstructions = cast(PseudoInstructions, dict())
code_offset = 0
while code_offset < len(data):
if data[code_offset + 1] == 0 and data[code_offset] != 0:
# Pseudo-instruction
# TODO performance benchmark swapping here vs. doing it once at
# beginning of function
data_swapped = endian_swap_shorts(data[code_offset + 2:])
if data[code_offset] == 1:
# packed-switch-payload
size = unpack("<H", data_swapped[:2])[0]
pseudoinstructions[cast(
"FileOffset",
addr + code_offset)] = SmaliPackedSwitchPayload(
_total_size=size * 4 + 8,
size=size,
first_key=unpack("<i", data_swapped[2:6])[0],
targets=[
unpack("<i", data_swapped[i:i + 4])[0]
for i in range(6, 6 + size * 4, 4)
],
)
code_offset += size * 4 + 8
elif data[code_offset] == 2:
# sparse-switch-payload
size = unpack("<H", data_swapped[:2])[0]
pseudoinstructions[cast(
"FileOffset",
addr + code_offset)] = SmaliSparseSwitchPayload(
_total_size=size * 8 + 4,
size=size,
keys=[
unpack("<i", data_swapped[i:i + 4])[0]
for i in range(2, 2 + size * 4, 4)
],
targets=[
unpack("<i", data_swapped[i:i + 4])[0]
for i in range(2 + size * 4, 2 + size * 8, 4)
],
)
code_offset += size * 8 + 4
elif data[code_offset] == 3:
# fill-array-data-payload
element_width = unpack("<H", data_swapped[:2])[0]
size = unpack("<I", data_swapped[2:6])[0]
pseudoinstructions[cast(
"FileOffset",
addr + code_offset)] = SmaliFillArrayDataPayload(
_total_size=((size * element_width + 1) // 2) * 2 + 8,
element_width=element_width,
size=size,
data=data_swapped[6:8 + (
(element_width * size + 1) // 2) * 2],
)
code_offset += ((size * element_width + 1) // 2) * 2 + 8
else:
log_error(
f"Unknown pseudoinstruction {data[code_offset:code_offset+2]!r} at {addr + code_offset} in code block at {addr}"
)
code_offset += 2
else:
# Normal instruction
insn_info = disassemble.insns[data[
code_offset + 1]] # type: ignore[attr-defined]
code_offset += insn_info.fmt.insn_len * 2
return pseudoinstructions
def disassemble(df: "DexFile", data: bytes,
addr: "FileOffset") -> Tuple[List[InstructionTextToken], int]:
# Static variable
if "insns" not in disassemble.__dict__:
# https://github.com/python/mypy/issues/708
disassemble.insns = load_insns() # type: ignore[attr-defined]
if len(data) < 2:
log_warn(
f"Trying to disassemble data of length {len(data)} at {addr}: {data!r}"
)
# Fun fact: if you return -1 here, binja segfaults
return [], 0
# Handle pseudo-instructions first
if data[0] == 0 and data[1] != 0:
if data[1] == 1:
# packed-switch
ps = cast(SmaliPackedSwitchPayload, df.pseudoinstructions[addr])
text = f".packed-switch {hex(ps.first_key)}\n"
text += "".join([
f" :pswitch_offset_{target:x}\n"
for target in ps.targets
])
text += " .end packed-switch"
elif data[1] == 2:
# sparse-switch
# FIXME why do these casts not work?
ps = cast(SmaliSparseSwitchPayload, df.pseudoinstructions[addr])
text = ".sparse-switch\n"
text += "".join([
f" {hex(ps.keys[i])} -> :sswitch_offset_{ps.targets[i]:x}\n"
for i in range(ps.size)
])
text += " .end sparse-switch"
elif data[1] == 3:
ps = cast(SmaliFillArrayDataPayload, df.pseudoinstructions[addr])
text = f"pseudo-instruction: {ps}"
else:
raise ValueError(f"Invalid pseudo-instruction with type {data[1]}")
return (
[
InstructionTextToken(
token_type=InstructionTextTokenType.InstructionToken,
text=text,
),
],
df.pseudoinstructions[addr]._total_size,
)
# Now handle normal instructions
tokens = list()
insn_info = disassemble.insns[data[0]] # type: ignore[attr-defined]
tokens.append(
InstructionTextToken(InstructionTextTokenType.InstructionToken,
insn_info.mnemonic))
data_to_parse = endian_swap_shorts(data[:2 * insn_info.fmt.insn_len])
if len(data_to_parse) != insn_info.fmt.insn_len * 2:
log_error(
"Disassembly failed. Too few bytes part of instruction available to parse"
)
return list(), insn_info.fmt.insn_len * 2
args = parse_with_format(data_to_parse, insn_info.fmt.format_)
if "r" in insn_info._formatid:
# Range instructions
args["N"] = args["A"] + args["C"] - 1
# Fix up syntax
if insn_info._formatid == "35c":
# 35c is weird for a couple reasons
# 1. It uses "kind" instead of the actual kind of the name of the
# constant pool
# 2. It forgets about "kind" for A=5 and lists them all out
m = re.search("\\s([a-z]+)@", insn_info.syntax)
if m is None:
log_error(f"Failed to parse 35c at {addr}")
else:
kind = m.group(1)
if args["A"] == 5:
syntax = f"{{vC, vD, vE, vF, vG}}, {kind}@BBBB"
elif args["A"] == 4:
syntax = f"{{vC, vD, vE, vF}}, {kind}@BBBB"
elif args["A"] == 3:
syntax = f"{{vC, vD, vE}}, {kind}@BBBB"
elif args["A"] == 2:
syntax = f"{{vC, vD}}, {kind}@BBBB"
elif args["A"] == 1:
syntax = f"{{vC}}, {kind}@BBBB"
elif args["A"] == 0:
syntax = f"{{}}, {kind}@BBBB"
else:
log_error(f"Failed to parse syntax for 35c instruction at {addr}")
syntax = "error (35c)"
elif "[A=" in insn_info.fmt.syntax:
for line in insn_info.fmt.syntax.split("[A="):
line = line.strip()
if line and line[0] == str(args["A"]):
syntax = line[6:]
break
else:
log_error(f"Failed to parse syntax for instruction at {addr}")
syntax = "error"
else:
syntax = insn_info.syntax
for word in syntax.split(" "):
if not word or word.isspace():
continue
tokens += tokenize_syntax(df, word, args)
return tokens, insn_info.fmt.insn_len * 2
def tokenize_syntax(df: "DexFile", word: str,
args: Dict[str, int]) -> List[InstructionTextToken]:
tokens = list()
tokens.append(InstructionTextToken(InstructionTextTokenType.TextToken,
" "))
# Check for prefixes and suffixes
trailing_comma = False
trailing_curly_brace = False
if word[-1] == ",":
trailing_comma = True
word = word[:-1]
if word[-1] == "}": # Needs to be after ',' check
trailing_curly_brace = True
word = word[:-1]
if word[0] == "{":
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken, "{"))
word = word[1:]
# Format operand with numbers where the placeholders are
word_formatted = format_args_with_syntax(args, word)
# Add operand token
if word_formatted == "":
# {}
pass
elif word_formatted[0] == "v":
# Register e.g. v01
val = int(word_formatted[1:], 16)
if val >= 256:
# TODO add link to issue. See comment in Smali
log_warn(
f"Rendering v{val}, but Binary Ninja only knows about registers up to 255 for analysis."
)
tokens.append(
InstructionTextToken(InstructionTextTokenType.RegisterToken,
f"v{val}"))
elif word_formatted[:2] == "#+":
# Literal e.g. #+0001
tokens.append(
InstructionTextToken(InstructionTextTokenType.IntegerToken,
hex(int(word_formatted[2:], 16))))
elif "@" in word_formatted:
# Lookup value e.g. call_site@0001
# Possible lookup types: call_site, field, method, method_handle, proto, string, type
lookup_type, lookup_index_str = word_formatted.split("@")
lookup_index = int(lookup_index_str, 16)
if lookup_type == "call_site":
log_warn(lookup_type + " isn't implemented yet")
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken,
word_formatted))
elif lookup_type == "field":
field = df.field_ids[lookup_index]
# Class name
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken,
field.class_))
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken, "->"))
# Field name
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken,
field.name))
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken, ":"))
# Type
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken,
field.type_))
elif lookup_type == "meth":
meth = df.method_ids[lookup_index]
# Class and method names
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken,
meth.class_))
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken, "->"))
if meth._insns_off is not None:
tokens.append(
InstructionTextToken(
InstructionTextTokenType.PossibleAddressToken,
meth.name,
value=meth._insns_off,
))
else:
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken,
meth.name))
# Parameters
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken, "("))
for param in meth.proto.parameters:
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken,
param))
# if meth.proto.parameters:
# # Remove trailing semicolon
# tokens.pop()
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken, ")"))
# Return type
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken,
meth.proto.return_type))
elif lookup_type == "method_handle":
log_warn(lookup_type + " isn't implemented yet")
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken,
word_formatted))
elif lookup_type == "proto":
log_warn(lookup_type + " isn't implemented yet")
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken,
word_formatted))
elif lookup_type == "string":
string_ = df.strings[lookup_index]
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken, '"'))
tokens.append(
# Escape e.g \n -> \\n or binja will render literal newline
InstructionTextToken(
InstructionTextTokenType.TextToken,
string_.encode("unicode-escape").decode(),
))
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken, '"'))
elif lookup_type == "type":
type_ = df.type_ids[lookup_index]
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken,
type_))
else:
log_error(f"Unknown lookup type: {word_formatted}")
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken,
word_formatted))
elif word_formatted[0] == "+":
# Address offset e.g. +0011
if int(word_formatted[1:], 16) >= 0:
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken, "+"))
tokens.append(
InstructionTextToken(InstructionTextTokenType.PossibleAddressToken,
word_formatted[1:]))
elif word_formatted == "..":
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken, ".."))
else:
# Other tokens. Investigate these
log_warn(
f'Formatting unknown token with syntax: "{word}": {word_formatted}'
)
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken,
word_formatted))
# Add suffixes
if trailing_curly_brace:
tokens.append(
InstructionTextToken(InstructionTextTokenType.TextToken, "}"))
if trailing_comma:
tokens.append(
InstructionTextToken(
InstructionTextTokenType.OperandSeparatorToken, ","))
return tokens
def __init__(self, data: bytes) -> None:
endian_bytes = data[40:44]
if endian_bytes == b"\x12\x34\x56\x78":
self.endianness = Endianness.BigEndian
elif endian_bytes == b"\x78\x56\x34\x12":
self.endianness = Endianness.LittleEndian
else:
raise ValueError(f"Invalid endianness found: {endian_bytes!r}")
if self.endianness == Endianness.BigEndian:
# It is likely that these do not exist at all, but who knows
log_warn(
"This is a big-endian file. The author was unable to find one of these to test with, so there will probably be errors. Please open an issue with a copy of this file!"
)
map_off = self._parse_uint(data[52:56])
map_size = self._parse_uint(data[map_off : map_off + 4])
# Parse map list items. First we collect them all, and then we parse
# them in an order that satisfies dependency relationships. For
# example, string_ids/strings need to be parsed first, and type_ids
# need to be parsed before type_lists, which need to be parsed before
# protos. Strings are the first items in the map list, but protos come
# before type_lists, so we can't just go in order.
map_list = dict()
for i in range(map_off + 4, 4 + map_off + map_size * 12, 12):
item_type = self._parse_ushort(data[i : i + 2])
item_size = self._parse_uint(data[i + 4 : i + 8])
item_offset = cast(FileOffset, self._parse_uint(data[i + 8 : i + 12]))
# log_debug(f'found type: "{item_type}", "{MapType(item_type).name}"')
map_list[item_type] = MapListItem(size=item_size, offset=item_offset)
# Ignore sections we don't need to reparse
map_list.pop(MapType.TYPE_HEADER_ITEM)
# The map list is what this part is parsing. No recursion
map_list.pop(MapType.TYPE_MAP_LIST)
# string_ids and strings
mi = map_list.pop(MapType.TYPE_STRING_ID_ITEM)
self.parse_string_ids(
data[mi.offset : mi.offset + 4 * mi.size], mi.size, mi.offset
)
self.make_strings(data)
del self.string_ids
map_list.pop(MapType.TYPE_STRING_DATA_ITEM) # Already handled
# Then, type_ids and type_lists
mi = map_list.pop(MapType.TYPE_TYPE_ID_ITEM)
self.parse_type_ids(
data[mi.offset : mi.offset + 4 * mi.size], mi.size, mi.offset
)
try:
mi = map_list.pop(MapType.TYPE_TYPE_LIST)
self.parse_type_lists(data[mi.offset :], mi.size, mi.offset)
except KeyError:
log_warn("No type list section")
# Need proto ids before method ids and both method ids and field
# ids before class data before class definitions
mi = map_list.pop(MapType.TYPE_PROTO_ID_ITEM)
self.parse_proto_ids(
data[mi.offset : mi.offset + 12 * mi.size], mi.size, mi.offset
)
mi = map_list.pop(MapType.TYPE_METHOD_ID_ITEM)
self.parse_method_ids(data[mi.offset :], mi.size, mi.offset)
try:
mi = map_list.pop(MapType.TYPE_FIELD_ID_ITEM)
self.parse_field_ids(
data[mi.offset : mi.offset + 8 * mi.size], mi.size, mi.offset
)
except KeyError:
log_warn("No field id section.")
mi = map_list.pop(MapType.TYPE_CODE_ITEM)
self.parse_code_items(data[mi.offset :], mi.size, mi.offset)
mi = map_list.pop(MapType.TYPE_CLASS_DATA_ITEM)
self.parse_class_data(data[mi.offset :], mi.size, mi.offset)
del self.code_items
# Need encoded_array_items before class_defs
try:
mi = map_list.pop(MapType.TYPE_ENCODED_ARRAY_ITEM)
self.parse_encoded_array_items(data[mi.offset :], mi.size, mi.offset)
except KeyError:
log_warn("No encoded array section.")
del self.proto_ids
# Rest are in order of MapType constant
mi = map_list.pop(MapType.TYPE_CLASS_DEF_ITEM)
self.parse_class_defs(
data[mi.offset : mi.offset + 32 * mi.size], mi.size, mi.offset
)
try:
del self.type_lists
del self.class_data_items
except AttributeError:
pass
try:
mi = map_list.pop(MapType.TYPE_CALL_SITE_ID_ITEM)
self.parse_call_site_ids(
data[mi.offset : mi.offset + 4 * mi.size], mi.size, mi.offset
)
except KeyError:
log_warn("No calls")
try:
mi = map_list.pop(MapType.TYPE_METHOD_HANDLE_ITEM)
self.parse_method_handles(
data[mi.offset : mi.offset + 8 * mi.size], mi.size, mi.offset
)
except KeyError:
log_warn("No methods")
# TODO annotations
try:
mi = map_list.pop(MapType.TYPE_ANNOTATION_ITEM)
mi = map_list.pop(MapType.TYPE_ANNOTATIONS_DIRECTORY_ITEM)
# self.parse_annotation_set_refs(data[mi.offset:mi.offset+4+mi.size*4], mi.size)
mi = map_list.pop(MapType.TYPE_ANNOTATION_SET_ITEM)
except KeyError:
log_warn("No annotations")
try:
mi = map_list.pop(MapType.TYPE_ANNOTATION_SET_REF_LIST)
# self.parse_annotation_sets(data[mi.offset:mi.offset+4+mi.size*4], mi.size)
except KeyError:
log_warn("No annotations set refs")
# TODO debug info
try:
mi = map_list.pop(MapType.TYPE_DEBUG_INFO_ITEM)
except KeyError:
log_warn("No debug info items")
for item_type in map_list:
log_error(f"unknown type {hex(item_type)}")