Adjust Qiling to comply with latest changed in Unicorn 2.1.2

[elicn]

[Merge only after Unicorn 2.1.2 is officially released]

Highlights:

• Introduced co-processor register accessors to AArch32 and AArch64 (ql.arch.cpr)
• Replaced deprecated pseudo registers with cpr definitions to be used with the new accessors
• Separated x86 64-bit only registers from the rest as Unicorn became less tolerant to accessing unsupported registers
• SSE instructions are no longer supported on the default Intel CPU model. x86 tests that use SSE instructions are now set to use Intel Haswell CPU model to make sure SSE is supported
• Various opportunistic styling and annotation fixes

Fixes: #1495, #1329

[wtdcode]

The next release will be 2.1.2 and it seems only Separated x86 64-bit only registers from the rest as Unicorn became less tolerant to accessing unsupported registers needs 2.2?

[elicn]

Ah, no, that's probably my mistake. I'll fix the title accordingly.

[wtdcode]

2.1.2 is alive. I will give this PR a shot recently.

[xwings]

@elicn looks like we need few more fix ?

[elicn]

@xwings, this is mostly because Qiling is still set to use the older Unicorn release.
We can start by setting Qiling to use the most recent one, and then see if it fails. From a quick smoke test it looks like there are a few additional fixes to be made.

[wtdcode]

The mips breakage (testing test_mips32eb_fake_urandom) breaks with 2.0.0, 2.0.1, 2.1.0, 2.1.1, 2.1.2 with:

[=] 	brk(inp = 0x0) = 0x56568000
[=] 	mmap2(addr = 0x0, length = 0x2000, prot = 0x3, flags = 0x802, fd = 0xffffffff, pgoffset = 0x0) = 0x1000000
[=] 	uname(buf = 0x7ff3c674) = 0x0
[=] 	access(path = 0x47dcb04, mode = 0x4) = -0x2 (ENOENT)
E
======================================================================
ERROR: test_mips32eb_fake_urandom (__main__.ELFTest)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/mio/opensource/qiling/tests/test_elf.py", line 462, in test_mips32eb_fake_urandom
    ql.run()
  File "/home/mio/opensource/qiling/qiling/core.py", line 588, in run
    self.os.run()
  File "/home/mio/opensource/qiling/qiling/os/linux/linux.py", line 178, in run
    self.ql.emu_start(self.ql.loader.entry_point, entry_address, self.ql.timeout)
  File "/home/mio/opensource/qiling/qiling/core.py", line 768, in emu_start
    raise self.internal_exception
  File "/home/mio/opensource/qiling/qiling/core_hooks.py", line 141, in wrapper
    return callback(*args, **kwargs)
  File "/home/mio/opensource/qiling/qiling/core_hooks.py", line 190, in _hook_intr_cb
    raise QlErrorCoreHook("_hook_intr_cb : not handled")
qiling.exception.QlErrorCoreHook: _hook_intr_cb : not handled

Is it possible that the breakage was introduced in the Qiling side?

[wtdcode]

Additional context:

	2.0.0	2.0.1	2.1.0	2.1.2
dev	y	y	UC_ERR_ARG	UC_ERR_WRITE_UNMAPPED
uc2.2-adjustments	_hook_intr_cb : not handled	_hook_intr_cb : not handled	_hook_intr_cb : not handled	UC_ERR_WRITE_UNALIGNED

I think this suggests a breakage on Qiling side? Note UC_ERR_WRITE_UNMAPPED seems from the lack of correct regiser access.

[wtdcode]

md = Cs(CS_ARCH_MIPS, CS_MODE_32 | CS_MODE_BIG_ENDIAN)
def instruction_count(ql: Qiling, address: int, size: int):
    a0 = ql.arch.regs.a0
    code = ql.mem.read(address, size)
    insn = list(md.disasm(code, size))[0]
    print(f"address={address:x} a0={a0:x} code={insn.mnemonic}")
    if insn.mnemonic == "sdc1":
        pc = ql.arch.regs.arch_pc
        pc += size
        ql.arch.regs.arch_pc = pc

This hooks fixes mips tests, no idea why ld doesn't respect stack alignment.

[wtdcode]

The left breakages:

x86 multithread: Bug of Unicorn side, fixed on master already
Android: works well dev branch, maybe check history again
mips32eb: The error is different between dev and this branch. The error message of this branch looks like: ../examples/rootfs/mips32_linux/bin/mips32_multithreading error while loading shared libraries /lib/libc.so.6 unexpected reloc type 0x03 while it gets UC_ERR_WRITE_UNMAPPED on dev branch. According to this, the binary itself is statically linked but Qiling tries to execute it via ld. Is that the reason?

[elicn]

Everything should be OK now, depending on the fixes applied in Unicorn 2.1.3.

[wtdcode]

Local test shows test_onlinux.sh passes with unicorn-engine/unicorn@56ba347 except m0hamed_rootkit.ko which I don't see the link to download.

Note this include the fix to MIPS delay slot so no longer need to tune the count of emulation. =)

Hooray!