Re: [PATCH v2 11/15] scripts/oss-fuzz: Add general-fuzzer build script

[Darren Kenny <darren.kenny@oracle.com>]

On Wednesday, 2020-08-19 at 02:11:06 -04, Alexander Bulekov wrote:
> This parses a yaml file containing general-fuzzer configs and builds a
> separate oss-fuzz wrapper binary for each one, changing some
> preprocessor macros for each configuration. To avoid dealing with
> escaping and stringifying, convert each string into a byte-array
> representation
>
> Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
> ---
>  scripts/oss-fuzz/build_general_fuzzers.py | 62 +++++++++++++++++++++++
>  1 file changed, 62 insertions(+)
>  create mode 100755 scripts/oss-fuzz/build_general_fuzzers.py
>
> diff --git a/scripts/oss-fuzz/build_general_fuzzers.py b/scripts/oss-fuzz/build_general_fuzzers.py
> new file mode 100755
> index 0000000000..79f4664117
> --- /dev/null
> +++ b/scripts/oss-fuzz/build_general_fuzzers.py
> @@ -0,0 +1,62 @@
> +#!/usr/bin/env python3
> +# -*- coding: utf-8 -*-
> +
> +"""
> +This script creates wrapper binaries that invoke the general-device-fuzzer with
> +configurations specified in a yaml config file.
> +"""
> +import sys
> +import os
> +import yaml
> +import tempfile
> +
> +CC = ""
> +TEMPLATE = ""
> +
> +
> +def usage():
> +    print("Usage: CC=COMPILER {} CONFIG_PATH \
> +OUTPUT_PATH_PREFIX".format(sys.argv[0]))
> +    sys.exit(0)
> +
> +
> +def str_to_c_byte_array(s):
> +    """
> +    Convert strings to byte-arrays so we don't worry about formatting
> +    strings to play nicely with cc -DQEMU_FUZZARGS etc
> +    """
> +    return ','.join('0x{:02x}'.format(ord(x)) for x in s)
> +
> +
> +def compile_wrapper(cfg, path):
> +    os.system('$CC -DQEMU_FUZZ_ARGS="{}" -DQEMU_FUZZ_OBJECTS="{}" \
> +                {} -o {}'.format(
> +                    str_to_c_byte_array(cfg["args"].replace("\n", " ")),
> +                    str_to_c_byte_array(cfg["objects"].replace("\n", " ")),
> +                    TEMPLATE, path))

NIT: When using multiple placeholders, it is nicer to use names for
them, so that reordering, or adding new ones is easier too.

> +
> +
> +def main():
> +    global CC
> +    global TEMPLATE
> +
> +    if len(sys.argv) != 3:
> +        usage()
> +
> +    cfg_path = sys.argv[1]
> +    out_path = sys.argv[2]
> +
> +    CC = os.getenv("CC")

Maybe provide a fall-back/default value if someone is calling it directly?

> +    TEMPLATE = os.path.join(os.path.dirname(__file__), "target.c")

No harm to double-check this exists, but also I would suggest that the
string "target.c" should be defined as a global value.

> +
> +    with open(cfg_path, "r") as f:
> +        configs = yaml.load(f)["configs"]
> +    for cfg in configs:
> +        assert "name" in cfg
> +        assert "args" in cfg
> +        assert "objects" in cfg
> +        compile_wrapper(cfg, out_path + cfg["name"])
> +
> +
> +if __name__ == '__main__':
> +    main()
> -- 
> 2.27.0

Thanks,

Darren.