lyn 0.0.6

Python bindings for GNU Lightning JIT

Lyn provides Python bindings for GNU Lightning:

GNU lightning is a library that generates assembly language code at
run-time; it is very fast, making it ideal for Just-In-Time compilers, and
it abstracts over the target CPU, as it exposes to the clients a
standardized RISC instruction set inspired by the MIPS and SPARC chips.

The source code is on GitHub: https://github.com/cslarsen/lyn/ Releases are uploaded to PyPI: https://pypi.python.org/pypi/lyn/

“Lyn” is the Norwegian word for “lightning”.

Warning

This project is in early alpha! Many instructions have not been implemented yet, and tests are lacking for those that have This means that you shouldn’t be surprised to segfault the entire Python process (you will have to get used to that anyway, unless you happen to always write bug-free Lightning code).

But, you can use it right now to JIT-compile native machine code, straight from Python. To get a taste of Lyn and GNU Lightning, scroll down to the examples below.

Installation

>From PyPi:

$ pip install lyn

>From the bleeding edge:

$ git clone https://github.com/cslarsen/lyn
$ cd lyn
$ python setup.py test
$ python setup.py install

Non-Python Dependencies

You must install the following libraries using your favourite package manager:

• The GNU Lightning shared library v2.1.0 (later versions may also work), http://www.gnu.org/software/lightning/

• Optional: The Capstone Disassembler, http://www.capstone-engine.org

The last time I compiled GNU Lightning on Linux, I had to disable the disassembly options because of linker problems with libopcodes.so. This worked for me:

$ ./configure --enable-shared --disable-static --disable-disassembler

To use Capstone as a disassembler with Lyn, you have to install the Python modules and the C library. The module can be installed with pip install capstone.

Example: Multiply two numbers

In this example, we use with-blocks so that the GNU Lightning environment (along with the mul function) is reclaimed:

from lyn import Lightning, word_t, Register

with Lightning() as lib:
    with lib.state() as jit:
        jit.prolog()
        jit.getarg(Register.r0, jit.arg())
        jit.getarg(Register.r1, jit.arg())
        jit.mulr(Register.r0, Register.r0, Register.r1)
        jit.retr(Register.r0)
        jit.epilog()

        mul = jit.emit_function(word_t, [word_t, word_t])

        for a in xrange(-100, 100):
            for b in xrange(-100, 100):
                assert(mul(a,b) == a*b)

To use the mul function elsewhere in your program, you need to keep a reference to the state jit and the GNU Lightning environment lib. Both objects have release() methods for doing it manually:

lib = Lightning()
jit = lib.state()
# ...
jit.release()
lib.release()

The last two parts are order dependant, in that lib.release() must run after its associated states. If you don’t release them, it’s not a big deal, but you’ll waste memory. In such a case, OS will free up the memory at exit.

Example: Calling a C function

This example shows how to call C functions from GNU Lightning. In the example below, we create a function that takes a string argument and returns the result of passing it to strlen:

import lyn
from lyn import Register, Lightning

lightning = Lightning()
libc = lightning.load("c")

jit = lightning.state()
jit.prolog()

# Get the Python argument
jit.getarg(Register.r0, jit.arg())

# Call strlen with it
jit.pushargr(Register.r0)
jit.finishi(libc.strlen)

# Return strlen's return value
jit.retval(Register.r0)
jit.retr(Register.r0)
jit.epilog()

strlen = jit.emit_function(lyn.word_t, [lyn.char_p])

self.assertEqual(strlen(""), 0)
self.assertEqual(strlen("h"), 1)
self.assertEqual(strlen("he"), 2)
self.assertEqual(strlen("hello"), 5)

lightning.release()

Notice that we tell emit_function to create a function that returns a lyn.word_t. This is a datatype whose size equals the computer’s pointer width, or sizeof(void*). lyn.word_t will then be either ctypes.c_int64 or ctypes.c_int32.

The parameter type lyn.char_p is a subclass of ctypes.c_char_p that automatically converts strings to bytes objects. This is provided as a compatibility convenience for Python 2 and 3 users. Use this type instead of ctypes.c_char_p.

Example: Disassembling native code with Capstone

If you install Capstone, you can use it as a disassembler for the generated functions. At some point, I’ll integrate Capstone into Lyn:

from lyn import Lightning, Register, word_t
import capstone
import ctypes

lib = Lightning()
jit = lib.state()

# A function that returns one more than its integer input
start = jit.note()
jit.prolog()
arg = jit.arg()
jit.getarg(Register.r0, arg)
jit.addi(Register.r0, Register.r0, 1)
jit.retr(Register.r0)
jit.epilog()
end = jit.note()

# Bind function to Python: returns a word (native integer), takes a word.
incr = jit.emit_function(word_t, [word_t])

# Sanity check
assert(incr(1234) == 1235)

# This part should be obvious to C programmers: We need to read data from raw
# memory in to a Python iterable.
length = (jit.address(end) - jit.address(start)).value
codebuf = ctypes.create_string_buffer(length)
ctypes.memmove(codebuf, ctypes.c_char_p(incr.address.value), length)
print("Compiled %d bytes starting at 0x%x" % (length, incr.address))

def hexbytes(b):
    return "".join(map(lambda x: hex(x)[2:] + " ", b))

# Capstone is smart enough to stop at the first RET-like instruction.
md = capstone.Cs(capstone.CS_ARCH_X86, capstone.CS_MODE_64)
md.syntax = capstone.CS_OPT_SYNTAX_ATT # Change to Intel syntax if you want
for i in md.disasm(codebuf, incr.address.value):
    print("0x%x %-15s%s %s" % (i.address, hexbytes(i.bytes), i.mnemonic, i.op_str))

raw = "".join(map(lambda x: "\\x%02x" % x, map(ord, codebuf)))
print("\nRaw bytes: %s" % raw)

jit.release()
lib.release()

On my computer, this outputs:

Compiled 34 bytes starting at 0x105ed3000
0x105ed3000 48 83 ec 30    subq $0x30, %rsp
0x105ed3004 48 89 2c 24    movq %rbp, (%rsp)
0x105ed3008 48 89 e5       movq %rsp, %rbp
0x105ed300b 48 83 ec 18    subq $0x18, %rsp
0x105ed300f 48 89 f8       movq %rdi, %rax
0x105ed3012 48 83 c0 1     addq $1, %rax
0x105ed3016 48 89 ec       movq %rbp, %rsp
0x105ed3019 48 8b 2c 24    movq (%rsp), %rbp
0x105ed301d 48 83 c4 30    addq $0x30, %rsp
0x105ed3021 c3             retq

Raw bytes:
    \x48\x83\xec\x30\x48\x89\x2c\x24
    \x48\x89\xe5\x48\x83\xec\x18\x48
    \x89\xf8\x48\x83\xc0\x01\x48\x89
    \xec\x48\x8b\x2c\x24\x48\x83\xc4
    \x30\xc3

Capstone has a lot of neat features. I happen to favour AT&T assembly syntax, but you can easily change that in the above code. But if you set md.detail = True, you’ll be able to see implicit registers and a lot of other cool stuff.

Author and license

Copyright (C) 2015 Christian Stigen Larsen

Distributed under the LGPL v2.1 or later. You are allowed to change the license on a particular copy to the LGPL v3.0, the GPL v2.0 or the GPL v3.0.