Source code for pyecsca.codegen.render

import os
import shutil
import subprocess
import tempfile
from os import path
from os import makedirs
from typing import Optional, List, Set, Mapping, MutableMapping, Any, Tuple

from jinja2 import Environment, PackageLoader
from importlib_resources import files
from public import public
from pyecsca.ec.configuration import HashType, RandomMod, Reduction, Multiplication, Squaring
from pyecsca.ec.coordinates import CoordinateModel
from pyecsca.ec.formula import Formula
from pyecsca.ec.model import CurveModel
from pyecsca.ec.mult import (
    ScalarMultiplier,
    LTRMultiplier,
    RTLMultiplier,
    CoronMultiplier,
    LadderMultiplier,
    SimpleLadderMultiplier,
    DifferentialLadderMultiplier,
    BinaryNAFMultiplier,
    WindowNAFMultiplier,
    SlidingWindowMultiplier,
    FixedWindowLTRMultiplier,
    FullPrecompMultiplier,
    BGMWMultiplier,
    CombMultiplier,
    AccumulationOrder,
    ProcessingDirection
)
from pyecsca.ec.op import OpType, CodeOp

from pyecsca.codegen.common import Platform, DeviceConfiguration

env = Environment(
        loader=PackageLoader("pyecsca.codegen")
)

env.globals["isinstance"] = isinstance
env.globals["bin"] = bin
env.globals["AccumulationOrder"] = AccumulationOrder
env.globals["ProcessingDirection"] = ProcessingDirection


def render_op(op: OpType, result: str, left: str, right: str, mod: str, red: str) -> Optional[str]:
    """
    Render an operation `op` (add, sub, neg, ...) into a C string.

    The operation will use `left` (optional) and `right` operands, be over `mod` modulus,
    use the `red` reduction context and leave the result in `result`. All of these need to
    be valid C variable names.
    """
    if op == OpType.Add:
        return "bn_red_add(&{}, &{}, &{}, &{}, &{});".format(left, right, mod, red, result)
    elif op == OpType.Sub:
        return "bn_red_sub(&{}, &{}, &{}, &{}, &{});".format(left, right, mod, red, result)
    elif op == OpType.Neg:
        return "bn_red_neg(&{}, &{}, &{}, &{});".format(right, mod, red, result)
    elif op == OpType.Mult:
        return "bn_red_mul(&{}, &{}, &{}, &{}, &{});".format(left, right, mod, red, result)
    elif op == OpType.Div or op == OpType.Inv:
        return "bn_red_div(&{}, &{}, &{}, &{}, &{});".format(left, right, mod, red, result)
    elif op == OpType.Sqr:
        return "bn_red_sqr(&{}, &{}, &{}, &{});".format(left, mod, red, result)
    elif op == OpType.Pow:
        return "bn_red_pow(&{}, &{}, &{}, &{}, &{});".format(left, right, mod, red, result)
    elif op == OpType.Id:
        return "bn_copy(&{}, &{});".format(left, result)
    else:
        print(op, result, left, right, mod)
        return None


env.globals["render_op"] = render_op


def render_defs(model: CurveModel, coords: CoordinateModel) -> str:
    """
    Render the defs.h file with definitions of the curve and point structures.

    These change depending on the number (and name) of curve parameters and number
    (and name) of point coordinates.
    """
    return env.get_template("defs.h").render(params=model.parameter_names,
                                             variables=coords.variables)


def render_curve_impl(model: CurveModel) -> str:
    """
    Render the defs.h file with basic curve operations.

    These change depending on the number (and name) of curve parameters.
    """
    return env.get_template("curve.c").render(params=model.parameter_names)


def transform_ops(ops: List[CodeOp], parameters: List[str], outputs: Set[str],
                  renames: Mapping[str, str] = None) -> MutableMapping[Any, Any]:
    """
    Transform a list of CodeOps, parameters, outputs and renames into a mapping
    that will be used by the ops template macros to render the ops.

    This tracks allocations and frees, also creates a mapping of constants
    to variable names.
    """
    def rename(name: str):
        if renames is not None and name not in outputs:
            return renames.get(name, name)
        return name

    allocations: List[str] = []
    initializations = {}
    const_mapping = {}
    operations = []
    frees = []
    # Go over the ops, track allocations needed for intermediates and constants.
    # There are two kinds of constants, encoded and not-encoded. The encoded
    # ones are default, the non-encoded ones are only in the exponents, where
    # you don't want to encode using the reduction object (i.e. Montgomery form).
    # Also constructs a mapping from raw constants to their variable names.
    for op in ops:
        if op.result not in allocations:
            allocations.append(op.result)
            frees.append(op.result)
        for param in op.parameters:
            if param not in allocations and param not in parameters:
                raise ValueError("Should be allocated or parameter: {}".format(param))
        for const in op.constants:
            if op.operator == OpType.Pow:
                name = "cu" + str(const)
                encode = False
            else:
                name = "c" + str(const)
                encode = True
            if name not in allocations:
                allocations.append(name)
                initializations[name] = (const, encode)
                const_mapping[(const, encode)] = name
                frees.append(name)
        operations.append((op.operator, op.result, rename(op.left), rename(op.right)))
    mapped = []
    # Go over the operations and map constants to their variable names,
    # make sure the encoded/non-encoded version is used where appropriate.
    for op in operations:
        o2 = op[2]
        if (o2, True) in const_mapping:
            o2 = const_mapping[(o2, True)]
        o3 = op[3]
        o3_enc = op[0] != OpType.Pow
        if (o3, o3_enc) in const_mapping:
            o3 = const_mapping[(o3, o3_enc)]
        mapped.append((op[0], op[1], o2, o3))
    returns = {}
    # Handle renames in the returns.
    if renames:
        for r_from, r_to in renames.items():
            if r_from in outputs:
                returns[r_from] = r_to

    return dict(allocations=allocations,
                initializations=initializations,
                const_mapping=const_mapping, operations=mapped,
                frees=frees, returns=returns)


def render_coords_impl(coords: CoordinateModel, accumulation_order: Optional[AccumulationOrder]) -> str:
    ops = []
    for s in coords.satisfying:
        try:
            ops.append(CodeOp(s))
        except Exception:
            pass
    renames = {"x": "out_x", "y": "out_y"}
    for variable in coords.variables:
        renames[variable] = "point->{}".format(variable)
    for param in coords.curve_model.parameter_names:
        renames[param] = "curve->{}".format(param)
    namespace = transform_ops(ops, coords.curve_model.parameter_names,
                              coords.curve_model.coordinate_names, renames)
    returns = namespace["returns"]
    namespace["returns"] = {}
    frees = namespace["frees"]
    namespace["frees"] = {}

    return env.get_template("point.c").render(variables=coords.variables, **namespace,
                                              to_affine_rets=returns, to_affine_frees=frees,
                                              accumulation_order=accumulation_order)


def render_formulas_impl(formulas: Set[Formula]) -> str:
    names = {formula.shortname for formula in formulas}
    return env.get_template("formulas.c").render(names=names)


def render_formula_impl(formula: Formula, short_circuit: bool = False) -> str:
    template = env.get_template(f"formula_{formula.shortname}.c")
    inputs = ["one", "other", "diff"]
    outputs = ["out_one", "out_other"]
    renames = {}
    for input in formula.inputs:
        var = input[0]
        num = int(input[1:]) - formula.input_index
        renames[input] = "{}->{}".format(inputs[num], var)
    for param in formula.coordinate_model.curve_model.parameter_names:
        renames[param] = "curve->{}".format(param)
    for output in formula.outputs:
        var = output[0]
        num = int(output[1:]) - formula.output_index
        renames[output] = "{}->{}".format(outputs[num], var)
    namespace = transform_ops(formula.code, formula.coordinate_model.curve_model.parameter_names,
                              formula.outputs, renames)
    namespace["short_circuit"] = short_circuit
    namespace["formula"] = formula
    return template.render(namespace)


def render_scalarmult_impl(scalarmult: ScalarMultiplier) -> str:
    return env.get_template("mult.c").render(scalarmult=scalarmult, LTRMultiplier=LTRMultiplier,
                                             RTLMultiplier=RTLMultiplier,
                                             CoronMultiplier=CoronMultiplier,
                                             LadderMultiplier=LadderMultiplier,
                                             SimpleLadderMultiplier=SimpleLadderMultiplier,
                                             DifferentialLadderMultiplier=DifferentialLadderMultiplier,
                                             BinaryNAFMultiplier=BinaryNAFMultiplier,
                                             WindowNAFMultiplier=WindowNAFMultiplier,
                                             SlidingWindowMultiplier=SlidingWindowMultiplier,
                                             FixedWindowLTRMultiplier=FixedWindowLTRMultiplier,
                                             FullPrecompMultiplier=FullPrecompMultiplier,
                                             BGMWMultiplier=BGMWMultiplier,
                                             CombMultiplier=CombMultiplier)


def render_action() -> str:
    return env.get_template("action.c").render()


def render_rand() -> str:
    return env.get_template("rand.c").render()


def render_main(model: CurveModel, coords: CoordinateModel, keygen: bool, ecdh: bool,
                ecdsa: bool) -> str:
    return env.get_template("main.c").render(model=model, coords=coords,
                                             curve_variables=coords.variables,
                                             curve_parameters=model.parameter_names,
                                             keygen=keygen, ecdh=ecdh, ecdsa=ecdsa)


def render_makefile(platform: Platform, hash_type: HashType, mod_rand: RandomMod,
                    reduction: Reduction, mul: Multiplication, sqr: Squaring, defines: Optional[MutableMapping[str, Any]]) -> str:
    return env.get_template("Makefile").render(platform=str(platform), hash_type=str(hash_type),
                                               mod_rand=str(mod_rand), reduction=str(reduction),
                                               mul=str(mul), sqr=str(sqr), defines=defines)


def save_render(dir: str, fname: str, rendered: str):
    with open(path.join(dir, fname), "w") as f:
        f.write(rendered)


[docs] @public def render(config: DeviceConfiguration) -> Tuple[str, str, str]: """ Render the `config` into a temporary directory. :param config: The configuration to render. :return: The temporary directory, the elf-file name, the hex-file name. """ temp = tempfile.mkdtemp() symlinks = ["asn1", "bn", "hal", "hash", "prng", "simpleserial", "tommath", "fat.h", "rand.h", "point.h", "curve.h", "mult.h", "formulas.h", "action.h", "Makefile.inc"] for sym in symlinks: os.symlink(files("pyecsca.codegen").joinpath(sym), path.join(temp, sym)) gen_dir = path.join(temp, "gen") makedirs(gen_dir, exist_ok=True) save_render(temp, "Makefile", render_makefile(config.platform, config.hash_type, config.mod_rand, config.red, config.mult, config.sqr, config.defines)) save_render(temp, "main.c", render_main(config.model, config.coords, config.keygen, config.ecdh, config.ecdsa)) save_render(gen_dir, "defs.h", render_defs(config.model, config.coords)) save_render(gen_dir, "point.c", render_coords_impl(config.coords, getattr(config.scalarmult, "accumulation_order", None))) save_render(gen_dir, "formulas.c", render_formulas_impl(config.formulas)) for formula in config.formulas: save_render(gen_dir, f"formula_{formula.shortname}.c", render_formula_impl(formula, config.scalarmult.short_circuit)) save_render(gen_dir, "action.c", render_action()) save_render(gen_dir, "rand.c", render_rand()) save_render(gen_dir, "curve.c", render_curve_impl(config.model)) save_render(gen_dir, "mult.c", render_scalarmult_impl(config.scalarmult)) return temp, "pyecsca-codegen-{}.elf".format( str(config.platform)), "pyecsca-codegen-{}.hex".format(str(config.platform))
[docs] @public def build(dir: str, elf_file: str, hex_file: str, outdir: str, strip: bool = False, remove: bool = True) -> subprocess.CompletedProcess: """ Build a rendered configuration. :param dir: Directory with a rendered implementation. :param elf_file: Elf-file name. :param hex_file: Hex-file name. :param outdir: Output directory to copy the elf and hex files into. :param strip: Whether to strip the resulting binary of debug symbols. :param remove: Whether to remove the original directory after build. :return: The subprocess that ran the build (make). """ res = subprocess.run(["make"], cwd=dir, capture_output=True) if res.returncode != 0: raise ValueError("Build failed!") if strip: subprocess.run(["make", "strip"], cwd=dir, capture_output=True) full_elf_path = path.join(dir, elf_file) full_hex_path = path.join(dir, hex_file) makedirs(outdir, exist_ok=True) shutil.copy(full_elf_path, outdir) shutil.copy(full_hex_path, outdir) if remove: shutil.rmtree(dir) return res
[docs] @public def render_and_build(config: DeviceConfiguration, outdir: str, strip: bool = False, remove: bool = True) -> Tuple[str, str, str, subprocess.CompletedProcess]: """ Render and build a `config` in one go. :param config: The configuration to build. :param outdir: Output directory to copy the elf and hex files into. :param strip: Whether to strip the resulting binary of debug symbols. :param remove: Whether to remove the original directory after build. :return: The subprocess that ran the build (make). """ dir, elf_file, hex_file = render(config) res = build(dir, elf_file, hex_file, outdir, strip, remove) return dir, elf_file, hex_file, res