def register_kernel_call(self):
# expect in buffer: the address to call and all dword to push on the stack
CCall_IOCTL = x64.MultipleInstr()
CCall_IOCTL += x64.Mov('RAX', self.IO_STACK_INPUT_BUFFER_LEN)
CCall_IOCTL += x64.Cmp('RAX', 0)
CCall_IOCTL += x64.Jz(":FAIL") # Need at least the function to call
CCall_IOCTL += x64.Mov('R15', 4 * 8) # Size to pop on the stack at the end (4 * push RDI)
CCall_IOCTL += x64.Mov('R10', self.IO_STACK_INPUT_BUFFER)
CCall_IOCTL += x64.Label(':PUSH_NEXT_ARG')
CCall_IOCTL += x64.Cmp('RAX', (8 * 5))
CCall_IOCTL += x64.Jbe(":SETUP_REG_ARGS")
CCall_IOCTL += x64.Sub('RAX', 8)
INPUT_BUFFER_NEXT_ARG = x64.create_displacement(base='R10', index='RAX')
CCall_IOCTL += x64.Mov('RBX', INPUT_BUFFER_NEXT_ARG)
CCall_IOCTL += x64.Push('RBX')
CCall_IOCTL += x64.Add('R15', 8) # Add at Size to pop on the stack at the end
CCall_IOCTL += x64.Jmp(':PUSH_NEXT_ARG')
CCall_IOCTL += x64.Label(":SETUP_REG_ARGS")
# Could be done in a loop
# But do I really want to generate x86 in a loop..
CCall_IOCTL += x64.Cmp('RAX', (8 * 5))
CCall_IOCTL += x64.Jz(":SETUP_4_ARGS")
CCall_IOCTL += x64.Cmp('RAX', (8 * 4))
CCall_IOCTL += x64.Jz(":SETUP_3_ARGS")
CCall_IOCTL += x64.Cmp('RAX', (8 * 3))
CCall_IOCTL += x64.Jz(":SETUP_2_ARGS")
CCall_IOCTL += x64.Cmp('RAX', (8 * 2))
CCall_IOCTL += x64.Jz(":SETUP_1_ARGS")
CCall_IOCTL += x64.Jmp(":SETUP_0_ARGS")
CCall_IOCTL += x64.Label(":SETUP_4_ARGS")
CCall_IOCTL += x64.Mov('R9', x64.mem('[R10 + 0x20]'))
CCall_IOCTL += x64.Label(":SETUP_3_ARGS")
CCall_IOCTL += x64.Mov('R8', x64.mem('[R10 + 0x18]'))
CCall_IOCTL += x64.Label(":SETUP_2_ARGS")
CCall_IOCTL += x64.Mov('RDX', x64.mem('[R10 + 0x10]'))
CCall_IOCTL += x64.Label(":SETUP_1_ARGS")
CCall_IOCTL += x64.Mov('RCX', x64.mem('[R10 + 8]'))
CCall_IOCTL += x64.Label(":SETUP_0_ARGS")
CCall_IOCTL += x64.Mov('RAX', x64.mem('[R10]'))
# Fix Reserve space (calling convention)
CCall_IOCTL += x64.Push('RDI')
CCall_IOCTL += x64.Push('RDI')
CCall_IOCTL += x64.Push('RDI')
CCall_IOCTL += x64.Push('RDI')
CCall_IOCTL += x64.Call('RAX')
CCall_IOCTL += x64.Mov('RDX', self.IRP_OUTPUT_BUFFER)
CCall_IOCTL += x64.Mov(x64.mem('[RDX]'), 'RAX')
CCall_IOCTL += x64.Xor('RAX', 'RAX')
CCall_IOCTL += x64.Add('RSP', 'R15')
CCall_IOCTL += x64.Ret()
CCall_IOCTL += x64.Label(":FAIL")
CCall_IOCTL += x64.Mov('RAX', 0x0C000000D)
CCall_IOCTL += x64.Ret()
self.upgrade_driver_add_new_ioctl_handler(DU_KCALL_IOCTL, CCall_IOCTL.get_code())
def NtCreateThreadEx_32_to_64(process, addr, param):
NtCreateThreadEx = get_NtCreateThreadEx_syswow_addr()
create_thread = x64.MultipleInstr()
# Save registers
create_thread += x64.Push('RBX')
create_thread += x64.Push('RCX')
create_thread += x64.Push('RDX')
create_thread += x64.Push('RSI')
create_thread += x64.Push('RDI')
create_thread += x64.Push('R8')
create_thread += x64.Push('R9')
create_thread += x64.Push('R10')
create_thread += x64.Push('R11')
create_thread += x64.Push('R12')
create_thread += x64.Push('R13')
# Setup args
create_thread += x64.Push(0)
create_thread += x64.Mov('RCX', 'RSP') # Arg1
create_thread += x64.Mov('RDX', 0x1fffff) # Arg2
create_thread += x64.Mov('R8', 0) # Arg3
create_thread += x64.Mov('R9', process.handle) # Arg4
create_thread += x64.Mov('RAX', 0)
create_thread += x64.Push('RAX') # Arg11
create_thread += x64.Push('RAX') # Arg10
create_thread += x64.Push('RAX') # Arg9
create_thread += x64.Push('RAX') # Arg8
create_thread += x64.Push('RAX') # Arg7
create_thread += x64.Mov('RAX', param)
create_thread += x64.Push('RAX') # Arg6
create_thread += x64.Mov('RAX', addr)
create_thread += x64.Push('RAX') # Arg5
# reserve space for register (calling convention)
create_thread += x64.Push('R9')
create_thread += x64.Push('R8')
create_thread += x64.Push('RDX')
create_thread += x64.Push('RCX')
# Call
create_thread += x64.Mov('R13', NtCreateThreadEx)
create_thread += x64.Call('R13')
# Clean stack
create_thread += x64.Add('RSP', 12 * 8)
create_thread += x64.Pop('R13')
create_thread += x64.Pop('R12')
create_thread += x64.Pop('R11')
create_thread += x64.Pop('R10')
create_thread += x64.Pop('R9')
create_thread += x64.Pop('R8')
create_thread += x64.Pop('RDI')
create_thread += x64.Pop('RSI')
create_thread += x64.Pop('RDX')
create_thread += x64.Pop('RCX')
create_thread += x64.Pop('RBX')
return execute_64bits_code_from_syswow(create_thread.get_code())
def _upgrade_driver_inject_base_upgrade(self):
kldbgdrv = self.kldbgdrv
upgrade = x64.MultipleInstr()
# R14 : IOCODE
# RSI -> IO_STACK_LOCATION
# RDI -> IRP
upgrade = x64.MultipleInstr()
upgrade += x64.Cmp('R14', self.NORMAL_IO_CODE)
upgrade += x64.Jz(self.normal_io_offset - (self.init_driver_offset + len(upgrade.get_code())))
upgrade += x64.Mov('Rax', x64.create_displacement(disp=kldbgdrv + self.HANDLE_ARRAY_ADDR))
upgrade += x64.Label(":LOOP")
upgrade += x64.Mov('RCX', x64.create_displacement('RAX'))
upgrade += x64.Cmp('R14', 'RCX')
upgrade += x64.Jnz(':END')
upgrade += x64.Mov('RAX', x64.create_displacement('RAX', disp=8))
upgrade += x64.Call('RAX')
upgrade += x64.Mov('RBX', 'RAX')
upgrade += x64.JmpAt(kldbgdrv + self.normal_end_offset)
upgrade += x64.Label(":END")
upgrade += x64.Cmp('RCX', 0)
upgrade += x64.Jnz(':NEXT')
upgrade += x64.JmpAt(kldbgdrv + self.fail_offset)
upgrade += x64.Label(":NEXT")
upgrade += x64.Add('RAX', 0x10)
upgrade += x64.Jmp(':LOOP')
self.kdbg.write_pfv_memory(kldbgdrv + self.init_driver_offset, str(upgrade.get_code()))
# Write first array dest
self.write_pfv_ptr(kldbgdrv + self.HANDLE_ARRAY_ADDR, kldbgdrv + self.FIRST_ARRAY_ADDR)
self.write_pfv_ptr(kldbgdrv + self.FIRST_ARRAY_ADDR, 0)
self.write_pfv_ptr(kldbgdrv + self.FIRST_ARRAY_ADDR + 8, 0)
# Jump hijack
jump_init_function = x64.Jmp(self.init_driver_offset - (self.hijack_offset))
self.kdbg.write_pfv_memory(kldbgdrv + self.hijack_offset, str(jump_init_function.get_code()))
self.ioctl_array = kldbgdrv + self.FIRST_ARRAY_ADDR
self.ioctl_array_ptr = kldbgdrv + self.HANDLE_ARRAY_ADDR
self.next_code_addr = kldbgdrv + self.init_driver_offset + len(upgrade.get_code())
self.is_upgraded = True
def generate_syswow64_call(target, errcheck=None):
nb_args = len(target.prototype._argtypes_)
target_addr = get_syswow_ntdll_exports()[target.__name__]
argument_buffer_len = (nb_args * 8)
argument_buffer = windows.current_process.allocator.reserve_size(argument_buffer_len)
alignement_information = windows.current_process.allocator.reserve_size(8)
nb_args_on_stack = max(nb_args - 4, 0)
code_64b = x64.MultipleInstr()
# Save registers
code_64b += x64.Push('RBX')
code_64b += x64.Push('RCX')
code_64b += x64.Push('RDX')
code_64b += x64.Push('RSI')
code_64b += x64.Push('RDI')
code_64b += x64.Push('R8')
code_64b += x64.Push('R9')
code_64b += x64.Push('R10')
code_64b += x64.Push('R11')
code_64b += x64.Push('R12')
code_64b += x64.Push('R13')
# Alignment stuff :)
code_64b += x64.Mov('RCX', 'RSP')
code_64b += x64.And('RCX', 0x0f)
code_64b += x64.Mov(x64.deref(alignement_information), 'RCX')
code_64b += x64.Sub('RSP', 'RCX')
# retrieve argument from the argument buffer
if nb_args >= 1:
code_64b += x64.Mov('RCX', x64.create_displacement(disp=argument_buffer))
if nb_args >= 2:
code_64b += x64.Mov('RDX', x64.create_displacement(disp=argument_buffer + (8 * 1)))
if nb_args >= 3:
code_64b += x64.Mov('R8', x64.create_displacement(disp=argument_buffer + (8 * 2)))
if nb_args >= 4:
code_64b += x64.Mov('R9', x64.create_displacement(disp=argument_buffer + (8 * 3)))
for i in range(nb_args_on_stack):
code_64b += x64.Mov('RAX', x64.create_displacement(disp=argument_buffer + 8 * (nb_args - 1 - i)))
code_64b += x64.Push('RAX')
# reserve space for register (calling convention)
code_64b += x64.Push('R9')
code_64b += x64.Push('R8')
code_64b += x64.Push('RDX')
code_64b += x64.Push('RCX')
# Call
code_64b += x64.Mov('R13', target_addr)
code_64b += x64.Call('R13')
# Realign stack :)
code_64b += x64.Add('RSP', x64.deref(alignement_information))
# Clean stack
code_64b += x64.Add('RSP', (4 + nb_args_on_stack) * 8)
code_64b += x64.Pop('R13')
code_64b += x64.Pop('R12')
code_64b += x64.Pop('R11')
code_64b += x64.Pop('R10')
code_64b += x64.Pop('R9')
code_64b += x64.Pop('R8')
code_64b += x64.Pop('RDI')
code_64b += x64.Pop('RSI')
code_64b += x64.Pop('RDX')
code_64b += x64.Pop('RCX')
code_64b += x64.Pop('RBX')
code_64b += x64.Ret()
return try_generate_stub_target(code_64b.get_code(), argument_buffer, target, errcheck=errcheck)
GetProcAddress64 += x64.Jz(":DLL_NOT_FOUND")
GetProcAddress64 += x64.Mov(
"RCX", x64.mem("[RAX + 80]")) # RCX = NAME (UNICODE_STRING.Buffer)
GetProcAddress64 += x64.Call(":FUNC_STRLENW64")
GetProcAddress64 += x64.Mov("RDI", "RCX")
GetProcAddress64 += x64.Mov("RCX", "RAX")
GetProcAddress64 += x64.Mov("RSI", "R11")
GetProcAddress64 += x64.Rep + x64.CmpsW(
) #;cmp with current dll name (unicode)
GetProcAddress64 += x64.Test("RCX", "RCX")
GetProcAddress64 += x64.Jz(":DLL_FOUND")
GetProcAddress64 += x64.Mov("RDX", x64.mem("[RDX]"))
GetProcAddress64 += x64.Jmp(":a_dest")
GetProcAddress64 += x64.Label(":DLL_FOUND") # here rbx = base
GetProcAddress64 += x64.Mov("EAX", x64.mem("[RBX + 60]")) # rax = PEBASE RVA
GetProcAddress64 += x64.Add("RAX", "RBX") # RAX = PEBASE
GetProcAddress64 += x64.Add("RAX", 24) # ;OPTIONAL HEADER
GetProcAddress64 += x64.Mov("ECX",
x64.mem("[rax + 112]")) # ;rcx = RVA export dir
GetProcAddress64 += x64.Add("RCX", "RBX") # ;rcx = export_dir
GetProcAddress64 += x64.Mov("RAX", "RCX") # ;RAX = export_dir
GetProcAddress64 += x64.Push("RAX") # ;Save it for after function search
# ; EBX = BASE | EAX = EXPORT DIR
GetProcAddress64 += x64.Mov("ECX", x64.mem("[RAX + 24] "))
GetProcAddress64 += x64.Mov("R13", "RCX") # ;r13 = NB names
GetProcAddress64 += x64.Mov("EDX",
x64.mem("[RAX + 32] ")) # EDX = names array RVA
GetProcAddress64 += x64.Add("RDX", "RBX") # RDX = names array
GetProcAddress64 += x64.Xor("RCX", "RCX")
GetProcAddress64 += x64.Label(":SEARCH_LOOP")
GetProcAddress64 += x64.Cmp("RCX", "R13")