Allow config of all experiment params, average across runs

benchmark/benchmark.py

@@ -1,161 +1,197 @@
 #!/usr/bin/env python3
-# Requires Python 3.6+.
+# Part of the aflpluslpus project, requires Python 3.7+.
-# Author: Chris Ball <chris@printf.net>
+# Author: Chris Ball <chris@printf.net>, ported from Marc "van Hauser" Heuse's "benchmark.sh".
-# Ported from Marc "van Hauser" Heuse's "benchmark.sh".
 import argparse
 import asyncio
-import glob
 import json
 import multiprocessing
 import os
+import platform
 import shutil
 import sys
-from collections import defaultdict
+import time
+from dataclasses import dataclass
 from decimal import Decimal
+from enum import Enum, auto
+from pathlib import Path
 
-reset = "\033[0m"
+blue   = lambda text: f"\033[1;94m{text}\033[0m"; gray = lambda text: f"\033[1;90m{text}\033[0m"
-blue  = lambda text: f"\033[1;94m{text}{reset}"
+green  = lambda text: f"\033[0;32m{text}\033[0m"; red  = lambda text: f"\033[0;31m{text}\033[0m"
-gray  = lambda text: f"\033[1;90m{text}{reset}"
+yellow = lambda text: f"\033[0;33m{text}\033[0m"
-green = lambda text: f"\033[0;32m{text}{reset}"
-red   = lambda text: f"\033[0;31m{text}{reset}"
 
-targets = [
+class Mode(Enum):
-    {"source": "../test-instr.c", "binary": "test-instr"},
+    multicore  = auto()
-    {"source": "../utils/persistent_mode/test-instr.c", "binary": "test-instr-persistent-shmem"},
+    singlecore = auto()
+
+@dataclass
+class Target:
+    source: Path
+    binary: str
+
+all_modes = [Mode.singlecore, Mode.multicore]
+all_targets = [
+    Target(source=Path("../utils/persistent_mode/test-instr.c").resolve(), binary="test-instr-persist-shmem"),
+    Target(source=Path("../test-instr.c").resolve(), binary="test-instr")
 ]
-modes = ["single-core", "multi-core"]
+mode_names = [mode.name for mode in all_modes]
-tree = lambda: defaultdict(tree) # recursive (arbitrary-depth) defaultdict!
+target_names = [target.binary for target in all_targets]
-results = tree()
+cpu_count = multiprocessing.cpu_count()
-between_tests = False
+
-parser = argparse.ArgumentParser()
+parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter)
-parser.add_argument("-d", "--debug", action="store_true")
+parser.add_argument("-b", "--basedir", help="directory to use for temp files", type=str, default="/tmp/aflpp-benchmark")
+parser.add_argument("-d", "--debug", help="show verbose debugging output", action="store_true")
+parser.add_argument("-r", "--runs", help="how many runs to average results over", type=int, default=5)
+parser.add_argument("-f", "--fuzzers", help="how many afl-fuzz workers to use", type=int, default=cpu_count)
+parser.add_argument("-m", "--mode", help="pick modes", action="append", default=["multicore"], choices=mode_names)
+parser.add_argument(
+    "-t", "--target", help="pick targets", action="append", default=["test-instr-persist-shmem"], choices=target_names
+)
 args = parser.parse_args()
 
-async def clean_up() -> None:
+# Really unsatisfying argparse behavior: we want a default and to allow multiple choices, but if there's a manual choice
-    """Remove temporary files."""
+# it should override the default.  Seems like we have to remove the default to get that and have correct help text?
-    shutil.rmtree("in")
+if len(args.target) > 1: args.target = args.target[1:]
+if len(args.mode) > 1: args.mode = args.mode[1:]
+
+targets = [target for target in all_targets if target.binary in args.target]
+modes = [mode for mode in all_modes if mode.name in args.mode]
+results = {"config": {}, "hardware": {}, "targets": {t.binary: {m.name: {} for m in modes} for t in targets}}
+debug = lambda text: args.debug and print(blue(text))
+if Mode.multicore in modes:
+    print(blue(f" [*] Using {args.fuzzers} fuzzers for multicore fuzzing "), end="")
+    print(blue("(use --fuzzers to override)" if args.fuzzers == cpu_count else f"(the default is {cpu_count})"))
+
+async def clean_up_tempfiles() -> None:
+    shutil.rmtree(f"{args.basedir}/in")
     for target in targets:
-        os.remove(target["binary"])
+        Path(target.binary).unlink()
         for mode in modes:
-            for outdir in glob.glob(f"/tmp/out-{mode}-{target['binary']}*"):
+            shutil.rmtree(f"{args.basedir}/out-{mode.name}-{target.binary}")
-                shutil.rmtree(outdir)
+
+async def check_afl_persistent() -> bool:
+    with open("/proc/cmdline", "r") as cpuinfo:
+        return "mitigations=off" in cpuinfo.read().split(" ")
+
+async def check_afl_system() -> bool:
+    sysctl = next((s for s in ["sysctl", "/sbin/sysctl"] if shutil.which(s)), None)
+    if sysctl:
+        (returncode, stdout, _) = await run_command([sysctl, "kernel.randomize_va_space"], None)
+        return returncode == 0 and stdout.decode().rstrip().split(" = ")[1] == "0"
+    return False
 
 async def check_deps() -> None:
-    """Check if the necessary files exist and are executable."""
+    """Checks for dependencies, platform, performance."""
-    if not (os.access("../afl-fuzz", os.X_OK) and os.access("../afl-cc", os.X_OK) and os.path.exists("../SanitizerCoveragePCGUARD.so")):
+    plat = platform.system()
-        sys.exit(f'{red(" [*] Error: you need to compile AFL++ first, we need afl-fuzz, afl-clang-fast and SanitizerCoveragePCGUARD.so built.")}')
+    if not plat == "Linux": sys.exit(red(f" [*] Error: Your platform '{plat}' is not supported by this script yet."))
+    if not os.access(Path("../afl-fuzz").resolve(), os.X_OK) and os.access(Path("../afl-cc").resolve(), os.X_OK) and (
+        os.path.exists(Path("../SanitizerCoveragePCGUARD.so").resolve()
+    )):
+        sys.exit(red(" [*] Compile AFL++: we need afl-fuzz, afl-clang-fast and SanitizerCoveragePCGUARD.so built."))
+
+    # Pick some sample settings from afl-{persistent,system}-config to try to see whether they were run.
+    cmd_checks = {"afl-persistent-config": check_afl_persistent, "afl-system-config": check_afl_system}
+    for cmd, checker in cmd_checks.items():
+        results["config"][cmd] = await checker()
+        if not results["config"][cmd]:
+            print(yellow(f" [*] {cmd} was not run. You can run it to improve performance (and decrease security)."))
 
 async def prep_env() -> dict:
-    # Unset AFL_* environment variables
+    """Unset AFL_* environment variables, create corpus dir and file, provide env vars for fuzzing."""
-    for e in list(os.environ.keys()):
+    Path(args.basedir).mkdir(exist_ok=True)
-        if e.startswith("AFL_"):
+    Path(f"{args.basedir}/in").mkdir(exist_ok=True)
-            os.environ.pop(e)
+    with open(f"{args.basedir}/in/in.txt", "wb") as seed: seed.write(b"\x00" * 10240)
-    # Create input directory and file
-    os.makedirs("in", exist_ok=True)
-    with open("in/in.txt", "wb") as f:
-        f.write(b"\x00" * 10240)
-    # Rest of env
-    AFL_PATH = os.path.abspath("../")
-    os.environ["PATH"] = AFL_PATH + ":" + os.environ["PATH"]
     return {
-        "AFL_BENCH_JUST_ONE": "1",
+        "AFL_BENCH_JUST_ONE": "1", "AFL_DISABLE_TRIM": "1", "AFL_I_DONT_CARE_ABOUT_MISSING_CRASHES": "1",
-        "AFL_DISABLE_TRIM": "1",
+        "AFL_NO_UI": "1", "AFL_TRY_AFFINITY": "1", "PATH": str(Path("../").resolve()),
-        "AFL_I_DONT_CARE_ABOUT_MISSING_CRASHES": "1",
-        "AFL_NO_UI": "1",
-        "AFL_TRY_AFFINITY": "1",
-        "PATH": f"{AFL_PATH}:{os.environ['PATH']}",
     }
 
 async def compile_target(source: str, binary: str) -> None:
     (returncode, stdout, stderr) = await run_command(
-        ["afl-cc", "-o", binary, source],
+        [Path("../afl-cc").resolve(), "-o", binary, source], env={"AFL_INSTRUMENT": "PCGUARD"}
-        env={"AFL_INSTRUMENT": "PCGUARD", "PATH": os.environ["PATH"]},
     )
-    if returncode != 0:
+    if returncode != 0: sys.exit(red(f" [*] Error: afl-cc is unable to compile: {stderr} {stdout}"))
-        sys.exit(f'{red(f" [*] Error: afl-cc is unable to compile: {stderr} {stdout}")}')
 
-async def cool_down() -> None:
+async def run_command(cmd: str, env: dict) -> (int | None, bytes, bytes):
-    """Avoid the next test run's results being contaminated by e.g. thermal limits hit on this one."""
+    debug(f"Launching command: {cmd} with env {env}")
-    global between_tests
+    p = await asyncio.create_subprocess_exec(
-    if between_tests:
+        *cmd, stdout=asyncio.subprocess.PIPE, stderr=asyncio.subprocess.PIPE, env=env
-        print(f'{blue("Taking a five second break to stay cool between tests.")}')
+    )
-        await asyncio.sleep(10)
-    else:
-        between_tests = True
-
-async def run_command(cmd, env) -> (int | None, bytes, bytes):
-    if args.debug:
-        print(blue(f"Launching command: {cmd} with env {env}"))
-    p = await asyncio.create_subprocess_exec(*cmd, stdout=asyncio.subprocess.PIPE, stderr=asyncio.subprocess.PIPE, env=env)
     stdout, stderr = await p.communicate()
+    debug(f"Output: {stdout.decode()} {stderr.decode()}")
     return (p.returncode, stdout, stderr)
 
 async def colon_value_or_none(filename: str, searchKey: str) -> str | None:
-    """Read a value (e.g. 'cpu MHz         : 4976.109') given its filename and key."""
+    """Return a colon-separated value given a key in a file, e.g. 'cpu MHz         : 4976.109')"""
     with open(filename, "r") as fh:
-        for line in fh:
+        kv_pairs = (line.split(": ", 1) for line in fh if ": " in line)
-            kv = line.split(": ", 1)
+        return next((v.rstrip() for k, v in kv_pairs if k.rstrip() == searchKey), None)
-            if kv and len(kv) == 2:
-                (key, value) = kv
-                key = key.strip()
-                value = value.strip()
-                if key == searchKey:
-                    return value
-        return None
 
 async def save_benchmark_results() -> None:
-    """We want a consistent JSON file, so read in the existing one, append, and replace."""
+    """Append a single row to the benchmark results in JSON Lines format (simple to write and to diff)."""
-    with open("benchmark-results.json", "r+") as jsonfile:
+    with open("benchmark-results.jsonl", "a") as jsonfile:
-        current_benchmarks = json.load(jsonfile)
+        json.dump(results, jsonfile, sort_keys=True)
-        current_benchmarks.append(results)
+        jsonfile.write("\n")
-        jsonfile.seek(0)
+        print(blue(f" [*] Results have been written to {jsonfile.name}"))
-        jsonfile.write(json.dumps(current_benchmarks, indent=2))
-        jsonfile.truncate()
-        print(json.dumps(results, indent=2))
 
 
 async def main() -> None:
-    print(f'{gray(" [*] Preparing environment")}')
+    print(" [*] Preparing environment")
-    # Remove stale files, if necessary.
     try:
-        await clean_up()
+        await clean_up_tempfiles()
     except FileNotFoundError:
         pass
     await check_deps()
+    # Only record the first core's speed for now, even though it can vary between cores.
+    results["hardware"]["cpu_mhz"]     = float(await colon_value_or_none("/proc/cpuinfo", "cpu MHz"))
+    results["hardware"]["cpu_model"]   = await colon_value_or_none("/proc/cpuinfo", "model name")
+    results["hardware"]["cpu_threads"] = cpu_count
     env_vars = await prep_env()
-    cpu_count = multiprocessing.cpu_count()
+    print(f" [*] Ready, starting benchmark...")
-    results["cpu_model"] = await colon_value_or_none("/proc/cpuinfo", "model name")
-    results["cpu_mhz"]   = await colon_value_or_none("/proc/cpuinfo", "cpu MHz")
-
-    print(f'{gray(" [*] Ready, starting benchmark - this will take approx 1-2 minutes...")}')
     for target in targets:
-        await compile_target(target["source"], target["binary"])
+        (source, binary) = [target.source, target.binary]
+        await compile_target(source, binary)
         for mode in modes:
-            await cool_down()
+            execs_per_sec, execs_total, run_time_total = ([] for _ in range(3))
-            print(f" [*] {mode} {target['binary']} benchmark starting, execs/s: ", end="", flush=True)
+            for run in range(0, args.runs):
-            if mode == "single-core":
+                print(gray(f" [*] {mode.name} {binary} run {run+1} of {args.runs}, execs/s: "), end="", flush=True)
-                cpus = [0]
+                fuzzers = range(0, args.fuzzers if mode == Mode.multicore else 1)
-            elif mode == "multi-core":
+                outdir = f"{args.basedir}/out-{mode.name}-{binary}"
-                cpus = range(0, cpu_count)
+                cmds = []
-            basedir = f"/tmp/out-{mode}-{target['binary']}-"
+                for (idx, afl) in enumerate(fuzzers):
-            cmd = [["afl-fuzz", "-i", "in", "-o", f"{basedir}{cpu}", "-M", f"{cpu}", "-s", "123", "-D", f"./{target['binary']}"] for cpu in cpus]
+                    name = ["-o", outdir, "-M" if idx == 0 else "-S", str(afl)]
+                    cmds.append(["afl-fuzz", "-i", f"{args.basedir}/in"] + name + ["-s", "123", "-D", f"./{binary}"])
 
-            # Here's where we schedule the tasks, and then block waiting for them to finish.
+                # Prepare the afl-fuzz tasks, and then block here while waiting for them to finish.
-            tasks = [run_command(cmd[cpu], env_vars) for cpu in cpus]
+                tasks = [run_command(cmds[cpu], env_vars) for cpu in fuzzers]
-            output = await asyncio.gather(*tasks)
+                start = time.time()
+                await asyncio.gather(*tasks)
+                end = time.time()
 
-            if args.debug:
+                # Our score is the sum of all execs_per_sec entries in fuzzer_stats files for the run.
-                for (_, stdout, stderr) in output:
+                tasks = [colon_value_or_none(f"{outdir}/{afl}/fuzzer_stats", "execs_per_sec") for afl in fuzzers]
-                    print(blue(f"Output: {stdout.decode()} {stderr.decode()}"))
+                all_execs_per_sec = await asyncio.gather(*tasks)
-            execs = sum([Decimal(await colon_value_or_none(f"{basedir}{cpu}/{cpu}/fuzzer_stats", "execs_per_sec")) for cpu in cpus])
+                execs = sum([Decimal(count) for count in all_execs_per_sec if count is not None])
                 print(green(execs))
-            results["targets"][target["binary"]][mode]["execs_per_second"] = str(execs)
+                execs_per_sec.append(execs)
-            results["targets"][target["binary"]][mode]["cores_used"] = len(cpus)
 
-    print("\nComparison: (note that values can change by 10-20% per run)")
+                # Also gather execs_total and total_run_time for this run.
-    with open("COMPARISON", "r") as f:
+                tasks = [colon_value_or_none(f"{outdir}/{afl}/fuzzer_stats", "execs_done") for afl in fuzzers]
-        print(f.read())
+                all_execs_total = await asyncio.gather(*tasks)
-    await clean_up()
+                execs_total.append(sum([Decimal(count) for count in all_execs_total if count is not None]))
+                run_time_total.append(Decimal(end - start))
+
+            total_run_time = round(Decimal(sum(run_time_total)), 2)
+            avg_score      = round(Decimal(sum(execs_per_sec) / len(execs_per_sec)), 2)
+            results["targets"][binary][mode.name] = {
+                "execs_per_second": float(avg_score),
+                "execs_total":      int(sum([Decimal(execs) for execs in execs_total])),
+                "fuzzers_used":     len(fuzzers),
+                "total_run_time":   float(total_run_time),
+            }
+            print(f" [*] Average score for this test across all runs was: {green(avg_score)}")
+            if (((max(execs_per_sec) - min(execs_per_sec)) / avg_score) * 100) > 15:
+                print(yellow(" [*] The difference between your slowest and fastest runs was >15%, maybe try again?"))
+    await clean_up_tempfiles()
     await save_benchmark_results()
 
 if __name__ == "__main__":