def argc(self, function):
'''
Counts the number of arguments used by the specified function.
'''
argv = set()
notargv = set()
ea = ida_shims.start_ea(function)
end_ea = ida_shims.end_ea(function)
if self.arch.unknown:
return 0
while ea < end_ea:
insn = ida_shims.decode_insn(ea)
features = ida_shims.get_canon_feature(insn)
for n in range(0, len(self.USE_OPND)):
ops = ida_shims.get_operands(insn)
if ops[n].type in [
idaapi.o_reg, idaapi.o_displ, idaapi.o_phrase
]:
try:
regname = self.arch.registers[ops[n].reg]
index = self.arch.argv.index(regname)
except ValueError:
continue
if features & self.USE_OPND[n] and regname not in notargv:
argv.update(self.arch.argv[:index + 1])
for n in range(0, len(self.CHANGE_OPND)):
ops = ida_shims.get_operands(insn)
if ops[n].type in [
idaapi.o_reg, idaapi.o_displ, idaapi.o_phrase
]:
try:
regname = self.arch.registers[ops[n].reg]
index = self.arch.argv.index(regname)
except ValueError:
continue
if regname not in argv:
notargv.update(self.arch.argv[index:])
if argv.union(notargv) == set(self.arch.argv):
break
# TODO: Use idc.NextHead(ea) instead...
ea += self.arch.insn_size
return len(argv)
def trace(self, ea):
'''
Given an EA where an argument register is set, attempt to trace what
function call that argument is passed to.
@ea - The address of an instruction that modifies a function argument
register.
Returns a tuple of (function EA, argv index, argument register name) on
success.
Returns None on failure.
'''
insn = ida_shims.decode_insn(ea)
features = ida_shims.get_canon_feature(insn)
if self.arch.unknown:
return (None, None, None)
for n in range(0, len(self.CHANGE_OPND)):
ops = ida_shims.get_operands(insn)
if ops[n].type in [idaapi.o_reg, idaapi.o_displ, idaapi.o_phrase]:
try:
regname = self.arch.registers[ops[n].reg]
index = self.arch.argv.index(regname)
except ValueError:
continue
if features & self.CHANGE_OPND[n]:
ea = ea - (self.arch.delay_slot * self.arch.insn_size)
while True:
insn = ida_shims.decode_insn(ea)
if idaapi.is_call_insn(ea):
for xref in idautils.XrefsFrom(ea):
if xref.type in [idaapi.fl_CF, idaapi.fl_CN]:
return (xref.to, index, regname)
# If we couldn't figure out where the function call
# was going to, just quit
break
try:
is_block_end = idaapi.is_basic_block_end(ea)
except TypeError:
is_block_end = idaapi.is_basic_block_end(ea, True)
if is_block_end:
break
# TODO: Use idc.NextHead(ea) instead...
ea += self.arch.insn_size
return (None, None, None)
def argv(self, func):
'''
Attempts to identify what types of arguments are passed to a given
function. Currently unused.
'''
args = [None for x in self.arch.argv]
if not self.arch.unknown:
start_ea = ida_shims.start_ea(func)
for xref in idautils.XrefsTo(start_ea):
if idaapi.is_call_insn(xref.frm):
insn = ida_shims.decode_insn(xref.frm)
ea = xref.frm + (self.arch.delay_slot *
self.arch.insn_size)
end_ea = (xref.frm - (self.arch.insn_size * 10))
while ea >= end_ea:
if idaapi.is_basic_block_end(ea) or \
(ea != xref.frm and idaapi.is_call_insn(ea)):
break
insn = ida_shims.decode_insn(ea)
features = ida_shims.get_canon_feature(insn)
for n in range(0, len(self.CHANGE_OPND)):
ops = ida_shims.get_operands(insn)
if ops[n].type in [
idaapi.o_reg, idaapi.o_displ,
idaapi.o_phrase
]:
try:
regname = self.arch.registers[ops[n].reg]
index = self.arch.argv.index(regname)
except ValueError:
continue
if features & self.CHANGE_OPND[n]:
for xref in idautils.XrefsFrom(ea):
# TODO: Where is this xref type defined?
if xref.type == 1:
string = \
ida_shims.get_strlit_contents(
xref.to)
if string and len(string) > 4:
args[index] = str
break
ea -= self.arch.insn_size
yield args
def _fix_data_offsets(self):
ea = 0
count = 0
print("Fixing unresolved offset xrefs...", end=' ')
while ea != idaapi.BADADDR:
(ea, n) = idaapi.find_notype(ea, idaapi.SEARCH_DOWN)
if ida_shims.can_decode(ea):
insn = ida_shims.decode_insn(ea)
ops = ida_shims.get_operands(insn)
for i in range(0, len(ops)):
op = ops[i]
if op.type == idaapi.o_imm and idaapi.getseg(op.value):
idaapi.add_dref(ea, op.value,
(idaapi.dr_O | idaapi.XREF_USER))
count += 1
print("created %d new data xrefs" % count)
def block(self, block):
'''
Returns a tuple:
([formal, block, signatures], [fuzzy, block, signatures],
set([unique, immediate, values]), [called, function, names])
'''
formal = []
fuzzy = []
functions = []
immediates = []
ea = ida_shims.start_ea(block)
while ea < ida_shims.end_ea(block):
insn = ida_shims.decode_insn(ea)
# Get a list of all data/code refs from the current instruction
drefs = [x for x in idautils.DataRefsFrom(ea)]
crefs = [x for x in idautils.CodeRefsFrom(ea, False)]
# Add all instruction mnemonics to the formal block hash
formal.append(ida_shims.print_insn_mnem(ea))
# If this is a call instruction, be sure to note the name of the
# function being called. This is used to apply call-based
# signatures to functions.
#
# For fuzzy signatures, we can't use the actual name or EA of the
# function, but rather just want to note that a function call was
# made.
#
# Formal signatures already have the call instruction mnemonic,
# which is more specific than just saying that a call was made.
if idaapi.is_call_insn(ea):
for cref in crefs:
func_name = ida_shims.get_name(cref)
if func_name:
functions.append(func_name)
fuzzy.append("funcref")
# If there are data references from the instruction, check to see
# if any of them are strings. These are looked up in the
# pre-generated strings dictionary.
#
# String values are easily identifiable, and are used as part of
# both the fuzzy and the formal signatures.
#
# It is more difficult to determine if non-string values are
# constants or not; for both fuzzy and formal signatures, just use
# "data" to indicate that some data was referenced.
elif drefs:
for dref in drefs:
if self.strings.has_key(dref):
formal.append(self.strings[dref].value)
fuzzy.append(self.strings[dref].value)
else:
formal.append("dataref")
fuzzy.append("dataref")
# If there are no data or code references from the instruction, use
# every operand as part of the formal signature.
#
# Fuzzy signatures are only concerned with interesting immediate
# values, that is, values that are greater than 65,535, are not
# memory addresses, and are not displayed as negative values.
elif not drefs and not crefs:
ops = ida_shims.get_operands(insn)
for n in range(0, len(ops)):
opnd_text = ida_shims.print_operand(ea, n)
formal.append(opnd_text)
if ops[n].type == idaapi.o_imm and \
not opnd_text.startswith('-'):
if ops[n].value >= 0xFFFF:
if ida_shims.get_full_flags(ops[n].value) == 0:
fuzzy.append(str(ops[n].value))
immediates.append(ops[n].value)
ea = ida_shims.next_head(ea)
return (self.sighash(''.join(formal)), self.sighash(''.join(fuzzy)),
immediates, functions)