ARM big endian not supported?

[michalmalik]

Does Unicorn have support for ARMEB?

from __future__ import print_function
from unicorn import *
from unicorn.arm_const import *

'''
mov %r0, $10
mov %r1, $20
add %r2, %r0, %r1
'''
CODE_LE = "\x0A\x00\xA0\xE3\x14\x10\xA0\xE3\x01\x20\x80\xE0"
CODE_BE = "\xE3\xA0\x00\x0A\xE3\xA0\x10\x14\xE0\x80\x20\x01"
BASE = 0x8000
EP = 0x8074
SIZE = 0x1000

try:
        little = Uc(UC_ARCH_ARM, UC_MODE_ARM | UC_MODE_LITTLE_ENDIAN)
        little.mem_map(BASE, SIZE)
        little.mem_write(EP, CODE_LE)
        little.emu_start(EP, EP + len(CODE_LE))
        print("LE R0: %d" % little.reg_read(UC_ARM_REG_R0))
        print("LE R1: %d" % little.reg_read(UC_ARM_REG_R1))
        print("LE R2: %d" % little.reg_read(UC_ARM_REG_R2))
except UcError as e:
        print("Little endian error: %s" % e)

try:
        big = Uc(UC_ARCH_ARM, UC_MODE_ARM | UC_MODE_BIG_ENDIAN)
        big.mem_map(BASE, SIZE)
        big.mem_write(EP, CODE_BE)
        big.emu_start(EP, EP + len(CODE_BE))
        print("BE R0: %d" % big.reg_read(UC_ARM_REG_R0))
        print("BE R1: %d" % big.reg_read(UC_ARM_REG_R1))
        print("BE R2: %d" % big.reg_read(UC_ARM_REG_R2))
except UcError as e:
        print("Big endian error: %s" % e)

Outputs:

LE R0: 10
LE R1: 20
LE R2: 30
Big endian error: Invalid mode (UC_ERR_MODE)

from __future__ import print_function
from unicorn import *

try:
    big = Uc(UC_ARCH_ARM, UC_MODE_ARM | UC_MODE_BIG_ENDIAN)
except UcError as e:
    print("Big endian error: %s" % e)

Outputs:

Big endian error: Invalid mode (UC_ERR_MODE)

[aquynh]

the reason is that Qemu does not support big endian for ARM, i dont know why. for Unicorn, you can look at how Mips supports big/little endian, and do the same thing to ARM. if you can do this, a pull request is welcome. otherwise, look forward to community support.

i will create a TODO list, and put this there.

[lunixbochs]

qemu-user has an armeb target

[aquynh]

this makes it even easier to port the big endian support to Unicorn.

[zachriggle]

Any updates on this before the v1.0 release?

[aquynh]

We are releasing 1.0 very soon, so will not have this feature in time. Apparently nobody works on this feature yet.