def translate(self, target):
output = []
for seg in target.memory.segments:
if seg.code:
for addr in target.entry_points:
FIXDELAY = 0
branchQ = []
while addr+4 < seg.end:
try:
IR = self.disassemble(target.memory.getrange(addr,addr+4), addr)
except KeyError, e:
print "finishing early due to invalid disassembly", e
break
if type(IR) == str:
instrs = [ir.unhandled_instruction(IR)]
instrs[0].address = addr
else:
instrs = IR[1]
added = 0
#reordering the delay slot is kind of like the chicken and egg problem
if FIXDELAY:
FIXDELAY = 0
added = 1
#XXX even worse than normal code alert, flip the addresses also
a = instrs[0].address
b = branchQ[0].address
for n in instrs:
n.address = b
for n in branchQ:
n.address = a
output += instrs
output += branchQ
#MIPS requires delay slot re-ordering for branches
for n in instrs:
if n.type in ["jump","call","ret","branch_true"]:
#wait until next time around to add the branch
FIXDELAY = 1
branchQ = instrs
break
if not FIXDELAY and not added:
output += instrs
addr += 4
def makeIR(self, instruction, size, operands, TMEM_IR, OPmode, addr):
"""
instruction -> dictionary of instruction information
operands -> list of operand type and value
TMEM_IR -> IR for complex memory access
"""
#print "-----make IR ------"
m = instruction['mnemonic']
preload = False
poststore = False
#figure out if TMEM_IR is LOAD or STORE
if TMEM_IR:
#first operand is destination
if 'reg/mem' in operands[0][0]:
preload = True
poststore = True
elif 'reg/mem' in operands[1][0]:
preload = True
elif 'moffset' in operands[1][0]:
preload = True
IR = []
if m == "ADC":
IR = [ir.operation(operands[0][1],'=',operands[0][1],'+',operands[1][1],'+',self.DR("CF"))]
elif m == "ADD":
IR = [ir.operation(operands[0][1],'=',operands[0][1],'+',operands[1][1])]
elif m == "AND":
IR = [ir.operation(operands[0][1],'=',operands[0][1],'&',operands[1][1])]
elif m == "CALL":
if poststore:
poststore = False
IR = [ir.operation(self.DR("TVAL"),'=',self.DR("EIP"),"+",ir.constant_operand(size)),
ir.operation(self.DR("ESP",OPmode),'=',self.DR("ESP"),'-',ir.constant_operand(4)),
ir.store(self.DR("TVAL"),self.DR("ESP",OPmode))]
#absolute jump vs relative jump
if 'rel' in operands[0][0]:
IR += [ir.operation(self.DR("tval"),'=',self.DR("EIP"),"+",operands[0][1],'+',ir.constant_operand(size)),
ir.call(self.DR("tval"), relative=True)
]
else:
IR += [ir.call(operands[0][1])]
#controversial... analyzer must know callee _should_ do this
#IR += [ir.operation(self.DR('esp'),'=',self.DR('esp'),'+',ir.constant_operand(4))]
elif m == "CLC":
IR = [ir.operation(self.DR("CF"), '=', ir.constant_operand(0))]
elif m == "CLD":
IR = [ir.operation(self.DR("DF"), '=', ir.constant_operand(0))]
elif m == "CMP":
if poststore:
poststore = False
IR = [ir.operation(operands[0][1],'-',operands[1][1])]
elif m == "DEC":
IR = [ir.operation(operands[0][1],'=',operands[0][1],'-',ir.constant_operand(1))]
elif m == "INC":
IR = [ir.operation(operands[0][1],'=',operands[0][1],'+',ir.constant_operand(1))]
elif m == "IMUL":
#XXXXX TODO FIX ME
#EDX:EAX = a*b ||
IR = [ir.operation(operands[0][1], '=', operands[0][1], '*', operands[1][1])]
if operands[0][1].register_name == 'eax':
#TODO SIZE
IR += [ir.operation(self.DR("EDX"), '=', '(', operands[0][1], '*', operands[1][1], ')', '>>', ir.constant_operand(32))]
elif m == "JMP":
if poststore:
poststore = False
#absolute jump vs relative jump
if isinstance(operands[0][1], ir.constant_operand):
DEST = ir.constant_operand(int(size + operands[0][1].value))
else:
DEST = operands[0][1]
if 'rel' in operands[0][0]:
IR += [ir.jump(DEST,relative=True)]
else:
IR += [ir.jump(DEST)]
elif 'J' == m[0]:
#IR = [ir.operation(self.DR("tval"),'=',self.DR("EIP"),"+",operands[0][1])]
DEST = ir.constant_operand(int(size + operands[0][1].value))
IR = []
if m == "JO":
IR += [ir.operation(self.DR('OF'),'==',ir.constant_operand(1))]
elif m == "JNO":
IR += [ir.operation(self.DR('OF'),'==',ir.constant_operand(0))]
elif m == "JB":
IR += [ir.operation(self.DR('CF'),'==',ir.constant_operand(1))]
elif m == "JNC":
IR += [ir.operation(self.DR('CF'),'==',ir.constant_operand(0))]
elif m == "JBE":
IR += [ir.operation(self.DR('ZF'),'==',ir.constant_operand(1), '||', self.DR("CF"), '==', ir.constant_operand(1))]
elif m == "JNBE":
IR += [ir.operation(self.DR('ZF'),'==',ir.constant_operand(0), '&&', self.DR("CF"), '==', ir.constant_operand(0))]
elif m == "JS":
IR += [ir.operation(self.DR('SF'),'==',ir.constant_operand(1))]
elif m == "JNS":
IR += [ir.operation(self.DR('SF'),'==',ir.constant_operand(0))]
elif m == "JP":
IR += [ir.operation(self.DR('PF'),'==',ir.constant_operand(1))]
elif m == "JNP":
IR += [ir.operation(self.DR('PF'),'==',ir.constant_operand(0))]
elif m == "JL":
IR += [ir.operation(self.DR('SF'),'!=',self.DR('OF'))]
elif m == "JNL":
IR += [ir.operation(self.DR('SF'),'==',self.DR('OF'))]
elif m == "JLE":
IR += [ir.operation(self.DR('ZF'),'==',ir.constant_operand(1), '||', self.DR("SF"), '!=', self.DR("OF"))]
elif m == "JNLE":
IR += [ir.operation(self.DR('ZF'),'==',ir.constant_operand(0), '&&', self.DR("SF"), '==', self.DR("OF"))]
elif m == "JNZ":
IR += [ir.operation(self.DR('ZF'),'==',ir.constant_operand(1))]
elif m == "JZ":
IR += [ir.operation(self.DR('ZF'), '==', ir.constant_operand(0))]
if 'rel' in operands[0][0]:
IR += [ir.branch_true(DEST)]
else:
IR += [ir.branch_true(operands[0][1])]
elif m == "LEA":
preload = False
poststore = False
IR = [ir.operation(operands[0][1], '=', self.DR("TMEM"))]
elif m == "LEAVE":
# mov esp, ebp
# pop ebp
IR = [ir.operation(self.DR("ESP"),'=',self.DR("EBP")),
ir.load(self.DR("ESP"), self.DR("EBP")),
ir.operation(self.DR("ESP",OPmode), '=', self.DR("ESP",OPmode),"+",ir.constant_operand(4))
]
elif m == "MOV":
#print hex(addr), operands
if preload:
if 'moffset' not in operands[1][0]:
if operands[1][1].type != 'register' or operands[1][1].register_name != 'tval':
preload = False
if operands[0][1].type == 'register' and operands[0][1].register_name == 'tval':
IR = [ir.operation(operands[1][1])]
else:
IR = [ir.operation(operands[0][1], '=', operands[1][1])]
elif m == "MOVSX":
#XXXXXX TODO sign extend
IR = [ir.operation(operands[0][1], '=', operands[1][1])]
elif m == "MOVZX":
if '16' in operands[1][0]:
mask = 0xff
else:
mask = 0xffff
IR = [ir.operation(operands[0][1], '=', operands[1][1], '&', ir.constant_operand(mask))]
elif m == "NOT":
IR = [ir.operation(operands[0][1],'=','~',operands[0][1])]
elif m == "OR":
IR = [ir.operation(operands[0][1],'=',operands[0][1],'|',operands[1][1])]
elif m == "POP":
#TODO 0x20b6: pushl 0xff5b(%eax) no handled correctly
IR = [ir.load(self.DR("ESP",OPmode), operands[0][1]),
ir.operation(self.DR("ESP",OPmode), '=', self.DR("ESP",OPmode),"+",ir.constant_operand(4))]
elif m == "PUSH":
if type(operands[0][1]) == str:
IR = [ir.unhandled_instruction(instruction['mnemonic'])]
else:
IR = [ir.operation(self.DR("ESP",OPmode), '=', self.DR("ESP",OPmode),"-",ir.constant_operand(4)),
ir.store(operands[0][1], self.DR("ESP",OPmode)),
]
if operands[0][1].type == 'register' and operands[0][1].register_name != 'tval':
poststore = False
elif m == "RET":
#pop eip
preload = True
IR = [ir.load(self.DR("ESP",OPmode), self.DR('TVAL')),
ir.operation(self.DR('ESP',OPmode),'=',self.DR("ESP",OPmode),'+',ir.constant_operand(4)),
ir.ret(self.DR('TVAL'))]
elif m == "ROL":
#XXX TODO FIX sz here
sz = operands[0][1].size
IR = [ir.operation(operands[0][1], '=', operands[0][1], '<<', operands[1][1], '|', operands[0][1],'>>', '(', ir.constant_operand(sz),'-',operands[1][1],')')]
elif m == "ROR":
sz = operands[0][1].size
IR = [ir.operation(operands[0][1], '=', operands[0][1], '>>', operands[1][1], '|', operands[0][1],'<<', '(', ir.constant_operand(sz),'-',operands[1][1],')')]
elif m == "SAL":
IR = [ir.operation(operands[0][1],'=', '(', self.DR("CF"), '<<', ir.constant_operand(32), '+', operands[0][1], ')','>>',operands[1][1])]
elif m == "SAR":
IR = [ir.operation(operands[0][1],'=', '(', self.DR("CF"), '>>', ir.constant_operand(32), '+', operands[0][1], ')','<<',operands[1][1])]
elif m == "SHL":
IR = [ir.operation(operands[0][1],'=', operands[0][1],'<<',operands[1][1])]
elif m == "SHR":
IR = [ir.operation(operands[0][1],'=', operands[0][1],'>>',operands[1][1])]
elif m == "SUB":
IR = [ir.operation(operands[0][1],'=', operands[0][1],'-',operands[1][1])]
elif m == "TEST":
IR = [ir.operation(operands[0][1],'&',operands[1][1])]
elif m == "XCHG":
have_nop = 0
if operands[0][1].type == 'register' and operands[1][1].type == 'register':
if operands[0][1].register_name == "eax" and operands[1][1].register_name == "eax":
have_nop = 1
#TODO does not play well with TMEM
if have_nop:
IR = [ir.operation("NOP")]
else:
#XXXXX TODO TMEM
IR = [ir.operation(self.DR("tval"),'=',operands[0][1]),
ir.operation(operands[0][1],'=', operands[1][1]),
ir.operation(operands[1][1],'=', self.DR("TVAL"))]
elif m == "XOR":
IR = [ir.operation(operands[0][1],'=', operands[0][1],'^',operands[1][1])]
else:
IR = [ir.unhandled_instruction(instruction['mnemonic'])]
if TMEM_IR:
#print "@@"#, preload, poststore, TMEM_IR
out = []
if preload:
out += TMEM_IR + [ir.load(self.DR("TMEM"),self.DR("TVAL"))]
elif TMEM_IR:
if not poststore:
out += TMEM_IR
if poststore:
if IR[0].type == 'operation':
#XXX implicit load store hack
if len(IR[0].ops) == 1:
IR = [ir.operation(self.DR('tval'),'=', *(IR[0].ops))]
out += IR + TMEM_IR + [ir.store(self.DR("tval"), self.DR('TMEM'))]
else:
out += IR
return out
return IR