Datasets:

JetBrains-Research

/

commit-chronicle

like 7

Follow

JetBrains Research 50

[ { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-06-17\n+ fuzzers:\n+ - aflplusplus\n+ - aflplusplus_optimal\n+ - aflplusplus_optimal_shmem\n+ - aflplusplus_qemu\n+ - aflplusplus_shmem\n+\n- experiment: 2020-06-12\nfuzzers:\n- aflcc\n" } ]

[ { "change_type": "MODIFY", "old_path": "experiment/conftest.py", "new_path": "experiment/conftest.py", "diff": "@@ -27,3 +27,14 @@ def experiment_config():\nwith open(config_filepath) as file_handle:\nreturn yaml.load(file_handle, yaml.SafeLoader)\n+\n+\[email protected]\n+def local_experiment_config():\n+ \"\"\"Fixture that returns the loaded yaml configuration\n+ test_data/local_experiment-config.yaml.\"\"\"\n+ config_filepath = os.path.join(os.path.dirname(__file__), 'test_data',\n+ 'local-experiment-config.yaml')\n+\n+ with open(config_filepath) as file_handle:\n+ return yaml.load(file_handle, yaml.SafeLoader)\n" }, { "change_type": "MODIFY", "old_path": "experiment/measurer.py", "new_path": "experiment/measurer.py", "diff": "@@ -572,14 +572,14 @@ def measure_snapshot_coverage(fuzzer: str, benchmark: str, trial_num: int,\ndef set_up_coverage_binaries(pool, experiment):\n\"\"\"Set up coverage binaries for all benchmarks in |experiment|.\"\"\"\n- benchmarks = [\n+ # Use set comprehension to select distinct benchmarks.\n+ benchmarks = {\ntrial.benchmark for trial in db_utils.query(models.Trial).distinct(\nmodels.Trial.benchmark).filter(\nmodels.Trial.experiment == experiment)\n- ]\n+ }\ncoverage_binaries_dir = build_utils.get_coverage_binaries_dir()\n- if not os.path.exists(coverage_binaries_dir):\n- os.makedirs(coverage_binaries_dir)\n+ filesystem.create_directory(coverage_binaries_dir)\npool.map(set_up_coverage_binary, benchmarks)\n@@ -588,8 +588,7 @@ def set_up_coverage_binary(benchmark):\ninitialize_logs()\ncoverage_binaries_dir = build_utils.get_coverage_binaries_dir()\nbenchmark_coverage_binary_dir = coverage_binaries_dir / benchmark\n- if not os.path.exists(benchmark_coverage_binary_dir):\n- os.mkdir(benchmark_coverage_binary_dir)\n+ filesystem.create_directory(benchmark_coverage_binary_dir)\narchive_name = 'coverage-build-%s.tar.gz' % benchmark\narchive_filestore_path = exp_path.filestore(coverage_binaries_dir /\narchive_name)\n" }, { "change_type": "MODIFY", "old_path": "experiment/resources/runner-startup-script-template.sh", "new_path": "experiment/resources/runner-startup-script-template.sh", "diff": "@@ -28,7 +28,7 @@ do\necho 'Error pulling image, retrying...'\ndone{% endif %}\n-docker run {% if local_experiment %}-v {{host_gcloud_config}}:/root/.config/gcloud {% endif %}\\\n+docker run \\\n--privileged --cpus=1 --rm \\\n-e INSTANCE_NAME={{instance_name}} \\\n-e FUZZER={{fuzzer}} \\\n@@ -36,11 +36,11 @@ docker run {% if local_experiment %}-v {{host_gcloud_config}}:/root/.config/gclo\n-e EXPERIMENT={{experiment}} \\\n-e TRIAL_ID={{trial_id}} \\\n-e MAX_TOTAL_TIME={{max_total_time}} \\\n--e CLOUD_PROJECT={{cloud_project}} \\\n--e CLOUD_COMPUTE_ZONE={{cloud_compute_zone}} \\\n--e EXPERIMENT_FILESTORE={{experiment_filestore}} \\\n--e REPORT_FILESTORE={{report_filestore}} \\\n+-e CLOUD_PROJECT={{cloud_project}} {% if not local_experiment %}-e CLOUD_COMPUTE_ZONE={{cloud_compute_zone}} {% endif %}\\\n+-e EXPERIMENT_FILESTORE={{experiment_filestore}} {% if local_experiment %}-v {{experiment_filestore}}:{{experiment_filestore}} {% endif %}\\\n+-e REPORT_FILESTORE={{report_filestore}} {% if local_experiment %}-v {{report_filestore}}:{{report_filestore}} {% endif %}\\\n-e FUZZ_TARGET={{fuzz_target}} \\\n+-e LOCAL_EXPERIMENT={{local_experiment}} \\\n{{additional_env}} {% if not local_experiment %}--name=runner-container {% endif %}\\\n--cap-add SYS_NICE --cap-add SYS_PTRACE \\\n{{docker_image_url}} 2>&1 | tee /tmp/runner-log.txt\n" }, { "change_type": "MODIFY", "old_path": "experiment/run_experiment.py", "new_path": "experiment/run_experiment.py", "diff": "@@ -350,9 +350,16 @@ class LocalDispatcher:\nexperiment=self.config['experiment'])\nset_cloud_project_arg = 'CLOUD_PROJECT={cloud_project}'.format(\ncloud_project=self.config['cloud_project'])\n+ shared_experiment_filestore_arg = '{0}:{0}'.format(\n+ self.config['experiment_filestore'])\nset_experiment_filestore_arg = (\n'EXPERIMENT_FILESTORE={experiment_filestore}'.format(\nexperiment_filestore=self.config['experiment_filestore']))\n+ shared_report_filestore_arg = '{0}:{0}'.format(\n+ self.config['report_filestore'])\n+ set_report_filestore_arg = (\n+ 'REPORT_FILESTORE={report_filestore}'.format(\n+ report_filestore=self.config['report_filestore']))\ndocker_image_url = '{base_docker_tag}/dispatcher-image'.format(\nbase_docker_tag=base_docker_tag)\ncommand = [\n@@ -364,6 +371,10 @@ class LocalDispatcher:\n'/var/run/docker.sock:/var/run/docker.sock',\n'-v',\nshared_volume_volume_arg,\n+ '-v',\n+ shared_experiment_filestore_arg,\n+ '-v',\n+ shared_report_filestore_arg,\n'-e',\nshared_volume_env_arg,\n'-e',\n@@ -377,6 +388,8 @@ class LocalDispatcher:\n'-e',\nset_experiment_filestore_arg,\n'-e',\n+ set_report_filestore_arg,\n+ '-e',\n'LOCAL_EXPERIMENT=True',\n'--cap-add=SYS_PTRACE',\n'--cap-add=SYS_NICE',\n" }, { "change_type": "MODIFY", "old_path": "experiment/scheduler.py", "new_path": "experiment/scheduler.py", "diff": "@@ -130,7 +130,8 @@ def end_expired_trials(experiment_config: dict):\nif not expired_instances:\nreturn\n- if not delete_instances(expired_instances, experiment_config):\n+ if not experiment_utils.is_local_experiment() and not delete_instances(\n+ expired_instances, experiment_config):\n# If we failed to delete some instances, then don't update the status\n# of expired trials in database as we don't know which instances were\n# successfully deleted. Wait for next iteration of end_expired_trials.\n@@ -585,21 +586,26 @@ def schedule_loop(experiment_config: dict):\n# Create the thread pool once and reuse it to avoid leaking threads and\n# other issues.\nlogger.info('Starting scheduler.')\n- gce.initialize()\nnum_trials = len(\nget_experiment_trials(experiment_config['experiment']).all())\n- trial_instance_manager = TrialInstanceManager(num_trials, experiment_config)\n+ local_experiment = experiment_utils.is_local_experiment()\n+ if not local_experiment:\n+ gce.initialize()\n+ trial_instance_manager = TrialInstanceManager(num_trials,\n+ experiment_config)\nexperiment = experiment_config['experiment']\nwith multiprocessing.Pool() as pool:\nhandle_preempted = False\nwhile not all_trials_ended(experiment):\ntry:\n- if not handle_preempted and not any_pending_trials(experiment):\n- # Only start handling preempted instances once every initial\n- # trial was started. This ensures that .\n+ schedule(experiment_config, pool)\n+ if not local_experiment:\n+ if not handle_preempted and not any_pending_trials(\n+ experiment):\n+ # Only start handling preempted instances once every\n+ # initial trial was started. This ensures that.\nhandle_preempted = True\n- schedule(experiment_config, pool)\nif handle_preempted:\ntrial_instance_manager.handle_preempted_trials()\nexcept Exception: # pylint: disable=broad-except\n@@ -732,7 +738,6 @@ def render_startup_script_template(instance_name: str, fuzzer: str,\n'trial_id': trial_id,\n'max_total_time': experiment_config['max_total_time'],\n'cloud_project': experiment_config['cloud_project'],\n- 'cloud_compute_zone': experiment_config['cloud_compute_zone'],\n'experiment_filestore': experiment_config['experiment_filestore'],\n'report_filestore': experiment_config['report_filestore'],\n'fuzz_target': fuzz_target,\n@@ -740,8 +745,9 @@ def render_startup_script_template(instance_name: str, fuzzer: str,\n'additional_env': additional_env,\n'local_experiment': local_experiment\n}\n- if local_experiment:\n- kwargs['host_gcloud_config'] = os.environ['HOST_GCLOUD_CONFIG']\n+\n+ if not local_experiment:\n+ kwargs['cloud_compute_zone'] = experiment_config['cloud_compute_zone']\nreturn template.render(**kwargs)\n" }, { "change_type": "MODIFY", "old_path": "experiment/stop_experiment.py", "new_path": "experiment/stop_experiment.py", "diff": "@@ -26,9 +26,14 @@ logger = logs.Logger('stop_experiment') # pylint: disable=invalid-name\ndef stop_experiment(experiment_name, experiment_config_filename):\n\"\"\"Stop the experiment specified by |experiment_config_filename|.\"\"\"\n+ experiment_config = yaml_utils.read(experiment_config_filename)\n+\n+ if experiment_config.get('local_experiment', False):\n+ raise NotImplementedError(\n+ 'Local experiment stop logic is not implemented.')\n+\ninstances = gcloud.list_instances()\n- experiment_config = yaml_utils.read(experiment_config_filename)\ncloud_compute_zone = experiment_config['cloud_compute_zone']\ntrial_prefix = 'r-' + experiment_name\nexperiment_instances = [\n" }, { "change_type": "ADD", "old_path": null, "new_path": "experiment/test_data/local-experiment-config.yaml", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+#\n+# TODO: remove `cloud_project` by `docker_registry`.\n+\n+experiment: test-experiment\n+trials: 4\n+max_total_time: 86400\n+cloud_project: fuzzbench\n+experiment_filestore: /tmp/experiment-data\n+report_filestore: /tmp/web-reports\n+local_experiment: true\n+benchmarks: \"benchmark-1,benchmark-2\"\n+git_hash: \"git-hash\"\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_scheduler.py", "new_path": "experiment/test_scheduler.py", "diff": "@@ -106,11 +106,11 @@ docker run \\\\\n-e EXPERIMENT=test-experiment \\\\\n-e TRIAL_ID=9 \\\\\n-e MAX_TOTAL_TIME=86400 \\\\\n--e CLOUD_PROJECT=fuzzbench \\\\\n--e CLOUD_COMPUTE_ZONE=us-central1-a \\\\\n+-e CLOUD_PROJECT=fuzzbench -e CLOUD_COMPUTE_ZONE=us-central1-a \\\\\n-e EXPERIMENT_FILESTORE=gs://experiment-data \\\\\n-e REPORT_FILESTORE=gs://web-reports \\\\\n-e FUZZ_TARGET={oss_fuzz_target} \\\\\n+-e LOCAL_EXPERIMENT=False \\\\\n-e C1=custom -e C2=custom2 --name=runner-container \\\\\n--cap-add SYS_NICE --cap-add SYS_PTRACE \\\\\n{docker_image_url} 2>&1 | tee /tmp/runner-log.txt'''\n@@ -127,16 +127,16 @@ docker run \\\\\n'gcr.io/fuzzbench/runners/fuzzer-a/bloaty_fuzz_target', 'fuzz_target')])\ndef test_create_trial_instance_local_experiment(benchmark, expected_image,\nexpected_target,\n- experiment_config, environ):\n+ local_experiment_config,\n+ environ):\n\"\"\"Test that create_trial_instance invokes create_instance and creates a\nstartup script for the instance, as we expect it to when running a\nlocal_experiment.\"\"\"\nos.environ['LOCAL_EXPERIMENT'] = str(True)\n- os.environ['HOST_GCLOUD_CONFIG'] = '~/.config/gcloud'\nexpected_startup_script = '''## Start docker.\n-docker run -v ~/.config/gcloud:/root/.config/gcloud \\\\\n+docker run \\\\\n--privileged --cpus=1 --rm \\\\\n-e INSTANCE_NAME=r-test-experiment-9 \\\\\n-e FUZZER=variant \\\\\n@@ -145,16 +145,16 @@ docker run -v ~/.config/gcloud:/root/.config/gcloud \\\\\n-e TRIAL_ID=9 \\\\\n-e MAX_TOTAL_TIME=86400 \\\\\n-e CLOUD_PROJECT=fuzzbench \\\\\n--e CLOUD_COMPUTE_ZONE=us-central1-a \\\\\n--e EXPERIMENT_FILESTORE=gs://experiment-data \\\\\n--e REPORT_FILESTORE=gs://web-reports \\\\\n+-e EXPERIMENT_FILESTORE=/tmp/experiment-data -v /tmp/experiment-data:/tmp/experiment-data \\\\\n+-e REPORT_FILESTORE=/tmp/web-reports -v /tmp/web-reports:/tmp/web-reports \\\\\n-e FUZZ_TARGET={oss_fuzz_target} \\\\\n+-e LOCAL_EXPERIMENT=True \\\\\n-e C1=custom -e C2=custom2 \\\\\n--cap-add SYS_NICE --cap-add SYS_PTRACE \\\\\n{docker_image_url} 2>&1 | tee /tmp/runner-log.txt'''\n_test_create_trial_instance(benchmark, expected_image, expected_target,\n- expected_startup_script, experiment_config,\n- False)\n+ expected_startup_script,\n+ local_experiment_config, False)\[email protected]('common.gcloud.create_instance')\n" } ]

[ { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-06-18\n+ fuzzers:\n+ - aflplusplus\n+ - aflplusplus_optimal\n+ - aflplusplus_optimal_shmem\n+ - aflplusplus_qemu\n+ - aflplusplus_shmem\n+\n- experiment: 2020-06-17\nfuzzers:\n- aflplusplus\n" } ]

[ { "change_type": "MODIFY", "old_path": "experiment/scheduler.py", "new_path": "experiment/scheduler.py", "diff": "@@ -598,14 +598,16 @@ def schedule_loop(experiment_config: dict):\nhandle_preempted = False\nwhile not all_trials_ended(experiment):\ntry:\n- schedule(experiment_config, pool)\n- if not local_experiment:\n- if not handle_preempted and not any_pending_trials(\n- experiment):\n- # Only start handling preempted instances once every\n- # initial trial was started. This ensures that.\n+ if (not local_experiment and not handle_preempted and\n+ not any_pending_trials(experiment)):\n+ # This ensures that:\n+ # 1. handle_preempted will not becomes True when running\n+ # locally.\n+ # 2. Only start handling preempted instances once every\n+ # initial trial was started.\nhandle_preempted = True\n+ schedule(experiment_config, pool)\nif handle_preempted:\ntrial_instance_manager.handle_preempted_trials()\nexcept Exception: # pylint: disable=broad-except\n" } ]

[ { "change_type": "MODIFY", "old_path": "common/filestore_utils.py", "new_path": "common/filestore_utils.py", "diff": "@@ -37,12 +37,15 @@ def get_impl():\nreturn local_filestore\n-def cp(source, destination, recursive=False, parallel=False): # pylint: disable=invalid-name\n- \"\"\"Copies |source| to |destination|.\"\"\"\n+def cp(source, destination, recursive=False, parallel=False, expect_zero=True): # pylint: disable=invalid-name\n+ \"\"\"Copies |source| to |destination|. If |expect_zero| is True then it can\n+ raise subprocess.CalledProcessError. |parallel| is only effective for\n+ gsutil.\"\"\"\nreturn get_impl().cp(source,\ndestination,\nrecursive=recursive,\n- parallel=parallel)\n+ parallel=parallel,\n+ expect_zero=expect_zero)\ndef ls(path, must_exist=True): # pylint: disable=invalid-name\n@@ -67,7 +70,8 @@ def rsync( # pylint: disable=too-many-arguments\ngsutil_options=None,\noptions=None,\nparallel=False):\n- \"\"\"Syncs |source| and |destination| folders.\"\"\"\n+ \"\"\"Syncs |source| and |destination| folders. |gsutil_options| and |parallel|\n+ are only used in gsutil.\"\"\"\nreturn get_impl().rsync(source,\ndestination,\ndelete,\n" }, { "change_type": "MODIFY", "old_path": "common/gsutil.py", "new_path": "common/gsutil.py", "diff": "@@ -28,25 +28,28 @@ def gsutil_command(arguments, expect_zero=True, parallel=False):\nreturn new_process.execute(command + arguments, expect_zero=expect_zero)\n-def cp(source, destination, recursive=False, parallel=False): # pylint: disable=invalid-name\n- \"\"\"Executes gsutil's \"cp\" command with |cp_arguments|.\"\"\"\n+def cp(source, destination, recursive=False, parallel=False, expect_zero=True): # pylint: disable=invalid-name\n+ \"\"\"Executes gsutil's \"cp\" command to copy |source| to |destination|. Uses -r\n+ if |recursive|. If |expect_zero| is True and the command fails then this\n+ function will raise a subprocess.CalledError.\"\"\"\ncommand = ['cp']\nif recursive:\ncommand.append('-r')\ncommand.extend([source, destination])\n- return gsutil_command(command, parallel=parallel)\n+ return gsutil_command(command, parallel=parallel, expect_zero=expect_zero)\ndef ls(path, must_exist=True): # pylint: disable=invalid-name\n- \"\"\"Executes gsutil's \"ls\" command on |path|.\"\"\"\n+ \"\"\"Executes gsutil's \"ls\" command on |path|. If |must_exist| is True and the\n+ command fails then this function will raise a subprocess.CalledError.\"\"\"\ncommand = ['ls', path]\nprocess_result = gsutil_command(command, expect_zero=must_exist)\nreturn process_result\ndef rm(path, recursive=True, force=False, parallel=False): # pylint: disable=invalid-name\n- \"\"\"Executes gsutil's rm command with |rm_arguments| and returns the result.\n+ \"\"\"Executes gsutil's rm command on |path| and returns the result.\nUses -r if |recursive|. If |force|, then uses -f and will not except if\nreturn code is nonzero.\"\"\"\ncommand = ['rm', path]\n" }, { "change_type": "MODIFY", "old_path": "common/local_filestore.py", "new_path": "common/local_filestore.py", "diff": "@@ -23,7 +23,8 @@ def cp( # pylint: disable=invalid-name\nsource,\ndestination,\nrecursive=False,\n- parallel=False): # pylint: disable=unused-argument\n+ parallel=False, # pylint: disable=unused-argument\n+ expect_zero=True):\n\"\"\"Executes \"cp\" command from |source| to |destination|.\"\"\"\n# Create intermediate folders for `cp` command to behave like `gsutil.cp`.\nfilesystem.create_directory(os.path.dirname(destination))\n@@ -32,7 +33,7 @@ def cp( # pylint: disable=invalid-name\nif recursive:\ncommand.append('-r')\ncommand.extend([source, destination])\n- return new_process.execute(command, expect_zero=True)\n+ return new_process.execute(command, expect_zero=expect_zero)\ndef ls(path, must_exist=True): # pylint: disable=invalid-name\n" }, { "change_type": "MODIFY", "old_path": "experiment/measurer.py", "new_path": "experiment/measurer.py", "diff": "@@ -21,7 +21,6 @@ import os\nimport pathlib\nimport posixpath\nimport sys\n-import subprocess\nimport tarfile\nimport time\nfrom typing import List, Set\n@@ -400,12 +399,10 @@ class SnapshotMeasurer: # pylint: disable=too-many-instance-attributes\ndef copy_unchanged_cycles_file():\nunchanged_cycles_filestore_path = exp_path.filestore(\nself.unchanged_cycles_path)\n- try:\n- filestore_utils.cp(unchanged_cycles_filestore_path,\n- self.unchanged_cycles_path)\n- return True\n- except subprocess.CalledProcessError:\n- return False\n+ result = filestore_utils.cp(unchanged_cycles_filestore_path,\n+ self.unchanged_cycles_path,\n+ expect_zero=False)\n+ return result.retcode == 0\nif not os.path.exists(self.unchanged_cycles_path):\nif not copy_unchanged_cycles_file():\n@@ -540,9 +537,9 @@ def measure_snapshot_coverage(fuzzer: str, benchmark: str, trial_num: int,\nif not os.path.exists(corpus_archive_dir):\nos.makedirs(corpus_archive_dir)\n- try:\n- filestore_utils.cp(corpus_archive_src, corpus_archive_dst)\n- except subprocess.CalledProcessError:\n+ if filestore_utils.cp(corpus_archive_src,\n+ corpus_archive_dst,\n+ expect_zero=False).retcode:\nsnapshot_logger.warning('Corpus not found for cycle: %d.', cycle)\nreturn None\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_measurer.py", "new_path": "experiment/test_measurer.py", "diff": "import os\nimport shutil\n-import subprocess\nfrom unittest import mock\nimport queue\n@@ -200,7 +199,7 @@ def test_is_cycle_unchanged_no_file(mocked_cp, fs, experiment):\n# Make sure we log if there is no unchanged-cycles file.\nsnapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\nSNAPSHOT_LOGGER)\n- mocked_cp.side_effect = subprocess.CalledProcessError(1, ['fakecommand'])\n+ mocked_cp.return_value = new_process.ProcessResult(1, '', False)\nassert not snapshot_measurer.is_cycle_unchanged(0)\n" } ]

[ { "change_type": "MODIFY", "old_path": "experiment/build/local_build.py", "new_path": "experiment/build/local_build.py", "diff": "\"\"\"Module for building things on Google Cloud Build for use in trials.\"\"\"\nimport os\n-import posixpath\nfrom typing import Tuple\nfrom common import benchmark_utils\nfrom common import environment\n-from common import experiment_path as exp_path\n-from common import filestore_utils\n+from common import experiment_utils\nfrom common import logs\nfrom common import new_process\nfrom common import utils\n-from experiment.build import build_utils\nlogger = logs.Logger('builder') # pylint: disable=invalid-name\n@@ -43,8 +40,8 @@ def build_base_images() -> Tuple[int, str]:\ndef get_shared_coverage_binaries_dir():\n\"\"\"Returns the shared coverage binaries directory.\"\"\"\n- shared_volume = os.environ['SHARED_VOLUME']\n- return os.path.join(shared_volume, 'coverage-binaries')\n+ experiment_filestore_path = experiment_utils.get_experiment_filestore_path()\n+ return os.path.join(experiment_filestore_path, 'coverage-binaries')\ndef make_shared_coverage_binaries_dir():\n@@ -77,16 +74,10 @@ def copy_coverage_binaries(benchmark):\nshared_coverage_binaries_dir, coverage_build_archive)\ncommand = 'cd /out; tar -czvf {} *'.format(\ncoverage_build_archive_shared_dir_path)\n- new_process.execute([\n+ return new_process.execute([\n'docker', 'run', '-v', mount_arg, builder_image_url, '/bin/bash', '-c',\ncommand\n])\n- coverage_binaries_dir = build_utils.get_coverage_binaries_dir()\n- coverage_build_archive_gcs_path = posixpath.join(\n- exp_path.filestore(coverage_binaries_dir), coverage_build_archive)\n-\n- return filestore_utils.cp(coverage_build_archive_shared_dir_path,\n- coverage_build_archive_gcs_path)\ndef build_fuzzer_benchmark(fuzzer: str, benchmark: str) -> bool:\n" }, { "change_type": "MODIFY", "old_path": "experiment/run_experiment.py", "new_path": "experiment/run_experiment.py", "diff": "@@ -334,13 +334,11 @@ class LocalDispatcher:\ndef start(self):\n\"\"\"Start the experiment on the dispatcher.\"\"\"\n- shared_volume_dir = os.path.abspath('shared-volume')\n- if not os.path.exists(shared_volume_dir):\n- os.mkdir(shared_volume_dir)\n- shared_volume_volume_arg = '{0}:{0}'.format(shared_volume_dir)\n- shared_volume_env_arg = 'SHARED_VOLUME={}'.format(shared_volume_dir)\n+ experiment_filestore_path = os.path.abspath(\n+ self.config['experiment_filestore'])\n+ filesystem.create_directory(experiment_filestore_path)\nsql_database_arg = 'SQL_DATABASE_URL=sqlite:///{}'.format(\n- os.path.join(shared_volume_dir, 'local.db'))\n+ os.path.join(experiment_filestore_path, 'local.db'))\nbase_docker_tag = experiment_utils.get_base_docker_tag(\nself.config['cloud_project'])\n@@ -370,14 +368,10 @@ class LocalDispatcher:\n'-v',\n'/var/run/docker.sock:/var/run/docker.sock',\n'-v',\n- shared_volume_volume_arg,\n- '-v',\nshared_experiment_filestore_arg,\n'-v',\nshared_report_filestore_arg,\n'-e',\n- shared_volume_env_arg,\n- '-e',\nset_instance_name_arg,\n'-e',\nset_experiment_arg,\n" } ]

[ { "change_type": "MODIFY", "old_path": "common/filestore_utils.py", "new_path": "common/filestore_utils.py", "diff": "@@ -39,8 +39,8 @@ def get_impl():\ndef cp(source, destination, recursive=False, parallel=False, expect_zero=True): # pylint: disable=invalid-name\n\"\"\"Copies |source| to |destination|. If |expect_zero| is True then it can\n- raise subprocess.CalledProcessError. |parallel| is only effective for\n- gsutil.\"\"\"\n+ raise subprocess.CalledProcessError. |parallel| is only used by the gsutil\n+ implementation.\"\"\"\nreturn get_impl().cp(source,\ndestination,\nrecursive=recursive,\n@@ -71,7 +71,7 @@ def rsync( # pylint: disable=too-many-arguments\noptions=None,\nparallel=False):\n\"\"\"Syncs |source| and |destination| folders. |gsutil_options| and |parallel|\n- are only used in gsutil.\"\"\"\n+ are only used by the gsutil implementation.\"\"\"\nreturn get_impl().rsync(source,\ndestination,\ndelete,\n" } ]

[ { "change_type": "MODIFY", "old_path": "experiment/measurer.py", "new_path": "experiment/measurer.py", "diff": "@@ -570,11 +570,12 @@ def measure_snapshot_coverage(fuzzer: str, benchmark: str, trial_num: int,\ndef set_up_coverage_binaries(pool, experiment):\n\"\"\"Set up coverage binaries for all benchmarks in |experiment|.\"\"\"\n# Use set comprehension to select distinct benchmarks.\n- benchmarks = {\n- trial.benchmark for trial in db_utils.query(models.Trial).distinct(\n- models.Trial.benchmark).filter(\n- models.Trial.experiment == experiment)\n- }\n+ benchmarks = [\n+ benchmark_tuple[0]\n+ for benchmark_tuple in db_utils.query(models.Trial.benchmark).distinct(\n+ ).filter(models.Trial.experiment == experiment)\n+ ]\n+\ncoverage_binaries_dir = build_utils.get_coverage_binaries_dir()\nfilesystem.create_directory(coverage_binaries_dir)\npool.map(set_up_coverage_binary, benchmarks)\n" } ]

[ { "change_type": "MODIFY", "old_path": "benchmarks/woff2-2016-05-06/build.sh", "new_path": "benchmarks/woff2-2016-05-06/build.sh", "diff": "@@ -22,10 +22,13 @@ apt-get update && \\\nautoconf \\\nlibtool\n+# Get seeds.\n+get_git_revision https://github.com/google/oss-fuzz.git e8ffee4077b59e35824a2e97aa214ee95d39ed13 oss-fuzz\n+mkdir -p $OUT/seeds\n+cp oss-fuzz/projects/woff2/corpus/* $OUT/seeds\n+\nget_git_revision https://github.com/google/woff2.git 9476664fd6931ea6ec532c94b816d8fbbe3aed90 SRC\nget_git_revision https://github.com/google/brotli.git 3a9032ba8733532a6cd6727970bade7f7c0e2f52 BROTLI\n-get_git_revision https://github.com/FontFaceKit/roboto.git 0e41bf923e2599d651084eece345701e55a8bfde $OUT/seeds\n-rm -rf $OUT/seeds/.git # Remove unneeded .git folder.\nrm -f *.o\nfor f in font.cc normalize.cc transform.cc woff2_common.cc woff2_dec.cc woff2_enc.cc glyph.cc table_tags.cc variable_length.cc woff2_out.cc; do\n" } ]

[ { "change_type": "MODIFY", "old_path": "common/benchmark_utils.py", "new_path": "common/benchmark_utils.py", "diff": "import os\nimport re\n-from common import experiment_utils\nfrom common import fuzzer_utils\nfrom common import logs\nfrom common import oss_fuzz\n@@ -45,21 +44,18 @@ def get_fuzz_target(benchmark):\nreturn fuzzer_utils.DEFAULT_FUZZ_TARGET_NAME\n-def get_runner_image_url(benchmark, fuzzer, cloud_project):\n+def get_runner_image_url(benchmark, fuzzer, docker_registry):\n\"\"\"Get the URL of the docker runner image for fuzzing the benchmark with\nfuzzer.\"\"\"\n- base_tag = experiment_utils.get_base_docker_tag(cloud_project)\n- return '{base_tag}/runners/{fuzzer}/{benchmark}'.format(base_tag=base_tag,\n- fuzzer=fuzzer,\n- benchmark=benchmark)\n+ return '{docker_registry}/runners/{fuzzer}/{benchmark}'.format(\n+ docker_registry=docker_registry, fuzzer=fuzzer, benchmark=benchmark)\n-def get_builder_image_url(benchmark, fuzzer, cloud_project):\n+def get_builder_image_url(benchmark, fuzzer, docker_registry):\n\"\"\"Get the URL of the docker builder image for fuzzing the benchmark with\nfuzzer.\"\"\"\n- base_tag = experiment_utils.get_base_docker_tag(cloud_project)\n- return '{base_tag}/builders/{fuzzer}/{benchmark}'.format(\n- base_tag=base_tag, fuzzer=fuzzer, benchmark=benchmark)\n+ return '{docker_registry}/builders/{fuzzer}/{benchmark}'.format(\n+ docker_registry=docker_registry, fuzzer=fuzzer, benchmark=benchmark)\ndef get_oss_fuzz_builder_hash(benchmark):\n" }, { "change_type": "MODIFY", "old_path": "common/test_benchmark_utils.py", "new_path": "common/test_benchmark_utils.py", "diff": "@@ -19,7 +19,7 @@ from common import conftest\n# pylint: disable=invalid-name,unused-argument\n-CLOUD_PROJECT = 'fuzzbench'\n+DOCKER_REGISTRY = 'gcr.io/fuzzbench'\nOTHER_BENCHMARK = 'benchmark'\n@@ -62,7 +62,7 @@ def test_get_fuzz_target(benchmark, expected_fuzz_target, oss_fuzz_benchmark):\ndef test_get_runner_image_url(benchmark, expected_url, oss_fuzz_benchmark):\n\"\"\"Test that we can get the runner image url of a benchmark.\"\"\"\nassert benchmark_utils.get_runner_image_url(benchmark, 'fuzzer',\n- CLOUD_PROJECT) == expected_url\n+ DOCKER_REGISTRY) == expected_url\ndef test_get_builder_hash_oss_fuzz_benchmark(oss_fuzz_benchmark):\n" }, { "change_type": "MODIFY", "old_path": "conftest.py", "new_path": "conftest.py", "diff": "@@ -87,6 +87,8 @@ def use_local_filestore(experiment): # pylint: disable=redefined-outer-name,unu\n\"\"\"Mock a local filestore usage experiment.\"\"\"\nos.environ['EXPERIMENT_FILESTORE'] = '/experiment-data'\nos.environ['REPORT_FILESTORE'] = '/experiment-report'\n+ os.environ['LOCAL_EXPERIMENT'] = 'true'\n+ os.environ['DOCKER_REGISTRY'] = 'gcr.io/fuzzbench'\[email protected]\n" }, { "change_type": "MODIFY", "old_path": "docs/running-your-own-experiment/running_an_experiment.md", "new_path": "docs/running-your-own-experiment/running_an_experiment.md", "diff": "@@ -56,6 +56,10 @@ trials: 5\n# 1 day = 24 * 60 * 60 = 86400\nmax_total_time: 86400\n+# The docker registry for your fuzzbench experiments.\n+# If you use Google Cloud, this can be gcr.io/$PROJECT_NAME.\n+docker_registry: gcr.io/$PROJECT_NAME\n+\n# The name of your Google Cloud project.\ncloud_project: $PROJECT_NAME\n" }, { "change_type": "MODIFY", "old_path": "experiment/build/local_build.py", "new_path": "experiment/build/local_build.py", "diff": "@@ -68,7 +68,7 @@ def copy_coverage_binaries(benchmark):\nshared_coverage_binaries_dir = get_shared_coverage_binaries_dir()\nmount_arg = '{0}:{0}'.format(shared_coverage_binaries_dir)\nbuilder_image_url = benchmark_utils.get_builder_image_url(\n- benchmark, 'coverage', environment.get('CLOUD_PROJECT'))\n+ benchmark, 'coverage', environment.get('DOCKER_REGISTRY'))\ncoverage_build_archive = 'coverage-build-{}.tar.gz'.format(benchmark)\ncoverage_build_archive_shared_dir_path = os.path.join(\nshared_coverage_binaries_dir, coverage_build_archive)\n" }, { "change_type": "MODIFY", "old_path": "experiment/resources/runner-startup-script-template.sh", "new_path": "experiment/resources/runner-startup-script-template.sh", "diff": "@@ -36,7 +36,7 @@ docker run \\\n-e EXPERIMENT={{experiment}} \\\n-e TRIAL_ID={{trial_id}} \\\n-e MAX_TOTAL_TIME={{max_total_time}} \\\n--e CLOUD_PROJECT={{cloud_project}} {% if not local_experiment %}-e CLOUD_COMPUTE_ZONE={{cloud_compute_zone}} {% endif %}\\\n+-e DOCKER_REGISTRY={{docker_registry}} {% if not local_experiment %}-e CLOUD_PROJECT={{cloud_project}} -e CLOUD_COMPUTE_ZONE={{cloud_compute_zone}} {% endif %}\\\n-e EXPERIMENT_FILESTORE={{experiment_filestore}} {% if local_experiment %}-v {{experiment_filestore}}:{{experiment_filestore}} {% endif %}\\\n-e REPORT_FILESTORE={{report_filestore}} {% if local_experiment %}-v {{report_filestore}}:{{report_filestore}} {% endif %}\\\n-e FUZZ_TARGET={{fuzz_target}} \\\n" }, { "change_type": "MODIFY", "old_path": "experiment/run_experiment.py", "new_path": "experiment/run_experiment.py", "diff": "@@ -66,10 +66,11 @@ def read_and_validate_experiment_config(config_filename: str) -> Dict:\nand returns it.\"\"\"\nconfig = yaml_utils.read(config_filename)\nfilestore_params = {'experiment_filestore', 'report_filestore'}\n- cloud_config = {'cloud_compute_zone'}\n- string_params = cloud_config.union(filestore_params)\n+ cloud_config = {'cloud_compute_zone', 'cloud_project'}\n+ docker_config = {'docker_registry'}\n+ string_params = cloud_config.union(filestore_params).union(docker_config)\nint_params = {'trials', 'max_total_time'}\n- required_params = int_params.union(filestore_params)\n+ required_params = int_params.union(filestore_params).union(docker_config)\nlocal_experiment = config.get('local_experiment', False)\nif not local_experiment:\n@@ -120,6 +121,8 @@ def read_and_validate_experiment_config(config_filename: str) -> Dict:\nif not valid:\nraise ValidationError('Config: %s is invalid.' % config_filename)\n+\n+ config['local_experiment'] = local_experiment\nreturn config\n@@ -340,16 +343,16 @@ class LocalDispatcher:\nsql_database_arg = 'SQL_DATABASE_URL=sqlite:///{}'.format(\nos.path.join(experiment_filestore_path, 'local.db'))\n- base_docker_tag = experiment_utils.get_base_docker_tag(\n- self.config['cloud_project'])\n+ docker_registry = self.config['docker_registry']\nset_instance_name_arg = 'INSTANCE_NAME={instance_name}'.format(\ninstance_name=self.instance_name)\nset_experiment_arg = 'EXPERIMENT={experiment}'.format(\nexperiment=self.config['experiment'])\n- set_cloud_project_arg = 'CLOUD_PROJECT={cloud_project}'.format(\n- cloud_project=self.config['cloud_project'])\nshared_experiment_filestore_arg = '{0}:{0}'.format(\nself.config['experiment_filestore'])\n+ # TODO: (#484) Use config in function args or set as environment\n+ # variables.\n+ set_docker_registry_arg = 'DOCKER_REGISTRY={}'.format(docker_registry)\nset_experiment_filestore_arg = (\n'EXPERIMENT_FILESTORE={experiment_filestore}'.format(\nexperiment_filestore=self.config['experiment_filestore']))\n@@ -358,8 +361,8 @@ class LocalDispatcher:\nset_report_filestore_arg = (\n'REPORT_FILESTORE={report_filestore}'.format(\nreport_filestore=self.config['report_filestore']))\n- docker_image_url = '{base_docker_tag}/dispatcher-image'.format(\n- base_docker_tag=base_docker_tag)\n+ docker_image_url = '{docker_registry}/dispatcher-image'.format(\n+ docker_registry=docker_registry)\ncommand = [\n'docker',\n'run',\n@@ -376,14 +379,14 @@ class LocalDispatcher:\n'-e',\nset_experiment_arg,\n'-e',\n- set_cloud_project_arg,\n- '-e',\nsql_database_arg,\n'-e',\nset_experiment_filestore_arg,\n'-e',\nset_report_filestore_arg,\n'-e',\n+ set_docker_registry_arg,\n+ '-e',\n'LOCAL_EXPERIMENT=True',\n'--cap-add=SYS_PTRACE',\n'--cap-add=SYS_NICE',\n@@ -424,8 +427,7 @@ class GoogleCloudDispatcher(BaseDispatcher):\ngcloud.robust_begin_gcloud_ssh(self.instance_name,\nself.config['cloud_compute_zone'])\n- base_docker_tag = experiment_utils.get_base_docker_tag(\n- self.config['cloud_project'])\n+ docker_registry = self.config['docker_registry']\ncloud_sql_instance_connection_name = (\nself.config['cloud_sql_instance_connection_name'])\n@@ -435,6 +437,7 @@ class GoogleCloudDispatcher(BaseDispatcher):\n'-e INSTANCE_NAME=\"{instance_name}\" '\n'-e EXPERIMENT=\"{experiment}\" '\n'-e CLOUD_PROJECT=\"{cloud_project}\" '\n+ '-e DOCKER_REGISTRY=\"{docker_registry}\" '\n'-e EXPERIMENT_FILESTORE=\"{experiment_filestore}\" '\n'-e POSTGRES_PASSWORD=\"{postgres_password}\" '\n'-e CLOUD_SQL_INSTANCE_CONNECTION_NAME='\n@@ -442,7 +445,7 @@ class GoogleCloudDispatcher(BaseDispatcher):\n'--cap-add=SYS_PTRACE --cap-add=SYS_NICE '\n'-v /var/run/docker.sock:/var/run/docker.sock '\n'--name=dispatcher-container '\n- '{base_docker_tag}/dispatcher-image '\n+ '{docker_registry}/dispatcher-image '\n'/work/startup-dispatcher.sh'\n).format(\ninstance_name=self.instance_name,\n@@ -455,7 +458,7 @@ class GoogleCloudDispatcher(BaseDispatcher):\nexperiment_filestore=self.config['experiment_filestore'],\ncloud_sql_instance_connection_name=(\ncloud_sql_instance_connection_name),\n- base_docker_tag=base_docker_tag,\n+ docker_registry=docker_registry,\n)\nreturn gcloud.ssh(self.instance_name,\ncommand=command,\n" }, { "change_type": "MODIFY", "old_path": "experiment/scheduler.py", "new_path": "experiment/scheduler.py", "diff": "@@ -718,7 +718,8 @@ def render_startup_script_template(instance_name: str, fuzzer: str,\nprovided and return the result.\"\"\"\nfuzzer_config = fuzzer_config_utils.get_by_variant_name(fuzzer)\ndocker_image_url = benchmark_utils.get_runner_image_url(\n- benchmark, fuzzer_config['fuzzer'], experiment_config['cloud_project'])\n+ benchmark, fuzzer_config['fuzzer'],\n+ experiment_config['docker_registry'])\nfuzz_target = benchmark_utils.get_fuzz_target(benchmark)\n# Convert additional environment variables from configuration to arguments\n@@ -739,17 +740,18 @@ def render_startup_script_template(instance_name: str, fuzzer: str,\n'fuzzer': fuzzer,\n'trial_id': trial_id,\n'max_total_time': experiment_config['max_total_time'],\n- 'cloud_project': experiment_config['cloud_project'],\n'experiment_filestore': experiment_config['experiment_filestore'],\n'report_filestore': experiment_config['report_filestore'],\n'fuzz_target': fuzz_target,\n'docker_image_url': docker_image_url,\n'additional_env': additional_env,\n+ 'docker_registry': experiment_config['docker_registry'],\n'local_experiment': local_experiment\n}\nif not local_experiment:\nkwargs['cloud_compute_zone'] = experiment_config['cloud_compute_zone']\n+ kwargs['cloud_project'] = experiment_config['cloud_project']\nreturn template.render(**kwargs)\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_data/experiment-config.yaml", "new_path": "experiment/test_data/experiment-config.yaml", "diff": "@@ -16,6 +16,7 @@ experiment: test-experiment\ntrials: 4\nmax_total_time: 86400\ncloud_project: fuzzbench\n+docker_registry: gcr.io/fuzzbench\ncloud_compute_zone: us-central1-a\nexperiment_filestore: gs://experiment-data\nreport_filestore: gs://web-reports\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_data/local-experiment-config.yaml", "new_path": "experiment/test_data/local-experiment-config.yaml", "diff": "# See the License for the specific language governing permissions and\n# limitations under the License.\n#\n-# TODO: remove `cloud_project` by `docker_registry`.\nexperiment: test-experiment\ntrials: 4\nmax_total_time: 86400\n-cloud_project: fuzzbench\n+docker_registry: gcr.io/fuzzbench\nexperiment_filestore: /tmp/experiment-data\nreport_filestore: /tmp/web-reports\nlocal_experiment: true\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_run_experiment.py", "new_path": "experiment/test_run_experiment.py", "diff": "@@ -52,6 +52,8 @@ class TestReadAndValdiateExperimentConfig(unittest.TestCase):\n'experiment_filestore': 'gs://bucket',\n'report_filestore': 'gs://web-bucket',\n'experiment': 'experiment-name',\n+ 'docker_registry': 'gcr.io/fuzzbench',\n+ 'cloud_project': 'fuzzbench',\n'cloud_compute_zone': 'us-central1-a',\n'trials': 10,\n'max_total_time': 1000,\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_scheduler.py", "new_path": "experiment/test_scheduler.py", "diff": "@@ -106,7 +106,7 @@ docker run \\\\\n-e EXPERIMENT=test-experiment \\\\\n-e TRIAL_ID=9 \\\\\n-e MAX_TOTAL_TIME=86400 \\\\\n--e CLOUD_PROJECT=fuzzbench -e CLOUD_COMPUTE_ZONE=us-central1-a \\\\\n+-e DOCKER_REGISTRY=gcr.io/fuzzbench -e CLOUD_PROJECT=fuzzbench -e CLOUD_COMPUTE_ZONE=us-central1-a \\\\\n-e EXPERIMENT_FILESTORE=gs://experiment-data \\\\\n-e REPORT_FILESTORE=gs://web-reports \\\\\n-e FUZZ_TARGET={oss_fuzz_target} \\\\\n@@ -144,7 +144,7 @@ docker run \\\\\n-e EXPERIMENT=test-experiment \\\\\n-e TRIAL_ID=9 \\\\\n-e MAX_TOTAL_TIME=86400 \\\\\n--e CLOUD_PROJECT=fuzzbench \\\\\n+-e DOCKER_REGISTRY=gcr.io/fuzzbench \\\\\n-e EXPERIMENT_FILESTORE=/tmp/experiment-data -v /tmp/experiment-data:/tmp/experiment-data \\\\\n-e REPORT_FILESTORE=/tmp/web-reports -v /tmp/web-reports:/tmp/web-reports \\\\\n-e FUZZ_TARGET={oss_fuzz_target} \\\\\n" } ]

[ { "change_type": "MODIFY", "old_path": "service/experiment-config.yaml", "new_path": "service/experiment-config.yaml", "diff": "trials: 20\nmax_total_time: 82800 # 23 hours, the default time for preemptible experiments.\ncloud_project: fuzzbench\n+docker_registry: gcr.io/fuzzbench\ncloud_compute_zone: us-central1-a\nexperiment_filestore: gs://fuzzbench-data\nreport_filestore: gs://www.fuzzbench.com/reports\n" } ]

[ { "change_type": "MODIFY", "old_path": "Makefile", "new_path": "Makefile", "diff": "@@ -42,7 +42,10 @@ lint: install-dependencies\ntypecheck: install-dependencies\nsource ${VENV_ACTIVATE} && python3 presubmit.py typecheck\n-docs-serve:\n+install-docs-dependencies: docs/Gemfile.lock\n+ cd docs && bundle install\n+\n+docs-serve: install-docs-dependencies\ncd docs && bundle exec jekyll serve --livereload\nclear-cache:\n" } ]

[ { "change_type": "MODIFY", "old_path": "presubmit.py", "new_path": "presubmit.py", "diff": "@@ -260,8 +260,9 @@ def validate_experiment_requests(paths: List[Path]):\nautomatic_run_experiment.REQUESTED_EXPERIMENTS_PATH)\nreturn False\n+ # Only validate the latest request.\nresult = automatic_run_experiment.validate_experiment_requests(\n- experiment_requests)\n+ experiment_requests[:1])\nif not result:\nprint('%s is not valid.' %\n" }, { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-06-26\n+ fuzzers:\n+ - aflplusplus\n+ - aflplusplus_optimal\n+ - aflplusplus_optimal_shmem\n+ - aflplusplus_shmem\n+ - entropic\n+\n- experiment: 2020-06-18\nfuzzers:\n- aflplusplus\n" } ]

[ { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-06-30\n+ fuzzers:\n+ - aflplusplus_qemu\n+ - afl_qemu\n+ - honggfuzz_qemu\n+ - aflplusplus_optimal_shmem\n+\n- experiment: 2020-06-26\nfuzzers:\n- aflplusplus\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/fuzzers.yml", "new_path": ".github/workflows/fuzzers.yml", "diff": "@@ -32,7 +32,7 @@ jobs:\n- libfuzzer_nocmp\n- manul\n- mopt\n- # Greybox fuzzers\n+ # Binary-only (greybox) fuzzers.\n- eclipser\n- afl_qemu\n- aflplusplus_qemu\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/afl/fuzzer.py", "new_path": "fuzzers/afl/fuzzer.py", "diff": "@@ -19,8 +19,6 @@ import os\nfrom fuzzers import utils\n-# OUT environment variable is the location of build directory (default is /out).\n-\ndef prepare_build_environment():\n\"\"\"Set environment variables used to build targets for AFL-based\n@@ -74,7 +72,7 @@ def run_afl_fuzz(input_corpus,\n# FIXME: Currently AFL will exit if it encounters a crashing input in seed\n# corpus (usually timeouts). Add a way to skip/delete such inputs and\n# re-run AFL.\n- print('[run_fuzzer] Running target with afl-fuzz')\n+ print('[run_afl_fuzz] Running target with afl-fuzz')\ncommand = [\n'./afl-fuzz',\n'-i',\n@@ -100,7 +98,7 @@ def run_afl_fuzz(input_corpus,\n# performs.\n'2147483647'\n]\n- print('[run_fuzzer] Running command: ' + ' '.join(command))\n+ print('[run_afl_fuzz] Running command: ' + ' '.join(command))\noutput_stream = subprocess.DEVNULL if hide_output else None\nsubprocess.check_call(command, stdout=output_stream, stderr=output_stream)\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/afl_qemu/fuzzer.py", "new_path": "fuzzers/afl_qemu/fuzzer.py", "diff": "# limitations under the License.\n\"\"\"Integration code for AFL qemu fuzzer.\"\"\"\n-# as aflplusplus has the build for qemu already in there we include this\n+# As aflplusplus has the build for qemu already in there we include this.\nfrom fuzzers.aflplusplus import fuzzer as aflplusplus_fuzzer\n@@ -24,7 +24,7 @@ def build():\ndef fuzz(input_corpus, output_corpus, target_binary):\n\"\"\"Run fuzzer.\"\"\"\n- # necessary fuzzer options\n+ # Necessary fuzzer options.\nflags = ['-Q']\naflplusplus_fuzzer.fuzz(input_corpus,\noutput_corpus,\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/aflcc/fuzzer.py", "new_path": "fuzzers/aflcc/fuzzer.py", "diff": "@@ -303,7 +303,7 @@ def build():\npost_build(fuzz_target)\n-def run_fuzz(input_corpus,\n+def run_fuzzer(input_corpus,\noutput_corpus,\ntarget_binary,\nadditional_flags=None,\n@@ -345,10 +345,10 @@ def fuzz(input_corpus, output_corpus, target_binary):\n\"\"\"Run fuzzer.\"\"\"\nprepare_fuzz_environment(input_corpus)\n- # Note: dictionary automatically added by run_fuzz().\n+ # Note: dictionary automatically added by run_fuzzer().\n# Use a dictionary for original afl as well.\n- print('[run_fuzzer] Running AFL for original binary')\n+ print('[fuzz] Running AFL for original binary')\nsrc_file = '{target}-normalized-none-nopt.dict'.format(target=target_binary)\ndst_file = '{target}-original.dict'.format(target=target_binary)\nshutil.copy(src_file, dst_file)\n@@ -356,7 +356,7 @@ def fuzz(input_corpus, output_corpus, target_binary):\n# to be non-optimized to prevent AFL from aborting.\nos.system('sed -i \\'s/OPTIMIZED/NORMAL/g\\' {dict}'.format(dict=dst_file))\nafl_fuzz_thread1 = threading.Thread(\n- target=run_fuzz,\n+ target=run_fuzzer,\nargs=(input_corpus, output_corpus,\n'{target}-original'.format(target=target_binary),\n['-S', 'slave-original']))\n@@ -364,14 +364,14 @@ def fuzz(input_corpus, output_corpus, target_binary):\nprint('[run_fuzzer] Running AFL for normalized and optimized dictionary')\nafl_fuzz_thread2 = threading.Thread(\n- target=run_fuzz,\n+ target=run_fuzzer,\nargs=(input_corpus, output_corpus,\n'{target}-normalized-none-nopt'.format(target=target_binary),\n['-S', 'slave-normalized-nopt']))\nafl_fuzz_thread2.start()\nprint('[run_fuzzer] Running AFL for FBSP and optimized dictionary')\n- run_fuzz(input_corpus,\n+ run_fuzzer(input_corpus,\noutput_corpus,\n'{target}-no-collision-all-opt'.format(target=target_binary),\n['-S', 'slave-no-collision-all-opt'],\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/aflplusplus/fuzzer.py", "new_path": "fuzzers/aflplusplus/fuzzer.py", "diff": "@@ -20,8 +20,6 @@ import shutil\nfrom fuzzers.afl import fuzzer as afl_fuzzer\nfrom fuzzers import utils\n-# OUT environment variable is the location of build directory (default is /out).\n-\ndef get_cmplog_build_directory(target_directory):\n\"\"\"Return path to CmpLog target directory.\"\"\"\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/aflplusplus_optimal/fuzzer.py", "new_path": "fuzzers/aflplusplus_optimal/fuzzer.py", "diff": "@@ -26,8 +26,6 @@ import glob\nfrom fuzzers.aflplusplus import fuzzer as aflplusplus_fuzzer\n-# OUT environment variable is the location of build directory (default is /out).\n-\ndef build(): # pylint: disable=too-many-branches,too-many-statements\n\"\"\"Build benchmark.\"\"\"\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/aflplusplus_optimal_shmem/fuzzer.py", "new_path": "fuzzers/aflplusplus_optimal_shmem/fuzzer.py", "diff": "@@ -26,8 +26,6 @@ import glob\nfrom fuzzers.aflplusplus import fuzzer as aflplusplus_fuzzer\n-# OUT environment variable is the location of build directory (default is /out).\n-\ndef build(): # pylint: disable=too-many-branches,too-many-statements\n\"\"\"Build benchmark.\"\"\"\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/aflplusplus_qemu/fuzzer.py", "new_path": "fuzzers/aflplusplus_qemu/fuzzer.py", "diff": "@@ -18,8 +18,6 @@ import subprocess\nfrom fuzzers.aflplusplus import fuzzer as aflplusplus_fuzzer\n-# OUT environment variable is the location of build directory (default is /out).\n-\ndef build():\n\"\"\"Build benchmark.\"\"\"\n@@ -28,7 +26,7 @@ def build():\ndef fuzz(input_corpus, output_corpus, target_binary):\n\"\"\"Run fuzzer.\"\"\"\n- # get LLVMFuzzerTestOneInput address\n+ # Get LLVMFuzzerTestOneInput address.\nnm_proc = subprocess.run([\n'sh', '-c',\n'nm \\'' + target_binary + '\\' | grep \\'T afl_qemu_driver_stdin_input\\''\n@@ -36,8 +34,9 @@ def fuzz(input_corpus, output_corpus, target_binary):\nstdout=subprocess.PIPE,\ncheck=True)\ntarget_func = \"0x\" + nm_proc.stdout.split()[0].decode(\"utf-8\")\n- print('[run_fuzzer] afl_qemu_driver_stdin_input() address =', target_func)\n- # fuzzer options\n+ print('[fuzz] afl_qemu_driver_stdin_input() address =', target_func)\n+\n+ # Set fuzzer options.\nflags = ['-Q', '-L', '0'] # MOpt flags\nos.environ['AFL_COMPCOV_LEVEL'] = '3' # float compcov\nos.environ['AFL_QEMU_PERSISTENT_ADDR'] = target_func\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/aflplusplus_shmem/fuzzer.py", "new_path": "fuzzers/aflplusplus_shmem/fuzzer.py", "diff": "from fuzzers.aflplusplus import fuzzer as aflplusplus_fuzzer\n-# OUT environment variable is the location of build directory (default is /out).\n-\ndef build():\n\"\"\"Build benchmark.\"\"\"\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/aflsmart/fuzzer.py", "new_path": "fuzzers/aflsmart/fuzzer.py", "diff": "@@ -19,8 +19,6 @@ import glob\nfrom fuzzers.afl import fuzzer as afl_fuzzer\n-# OUT environment variable is the location of build directory (default is /out).\n-\ndef build():\n\"\"\"Build benchmark.\"\"\"\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/ankou/fuzzer.py", "new_path": "fuzzers/ankou/fuzzer.py", "diff": "@@ -20,8 +20,6 @@ import os\nfrom fuzzers import utils\nfrom fuzzers.afl import fuzzer as afl_fuzzer\n-# OUT environment variable is the location of build directory (default is /out).\n-\ndef build():\n\"\"\"Build benchmark.\"\"\"\n@@ -37,12 +35,12 @@ def fuzz(input_corpus, output_corpus, target_binary):\n\"\"\"Run Ankou on target.\"\"\"\nafl_fuzzer.prepare_fuzz_environment(input_corpus)\n- print('[run_fuzzer] Running target with Ankou')\n+ print('[fuzz] Running target with Ankou')\ncommand = [\n'./Ankou', '-app', target_binary, '-i', input_corpus, '-o',\noutput_corpus\n]\n# \"-dict\" option may not work for format mismatching.\n- print('[run_fuzzer] Running command: ' + ' '.join(command))\n+ print('[fuzz] Running command: ' + ' '.join(command))\nsubprocess.check_call(command)\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/eclipser/fuzzer.py", "new_path": "fuzzers/eclipser/fuzzer.py", "diff": "@@ -41,7 +41,7 @@ def fuzz(input_corpus, output_corpus, target_binary):\nif not os.path.exists(encoded_temp_corpus):\nos.mkdir(encoded_temp_corpus)\n- print('[run_fuzzer] Running target with Eclipser')\n+ print('[fuzz] Running target with Eclipser')\ncommand = [\n'dotnet',\n'/Eclipser/build/Eclipser.dll',\n@@ -69,7 +69,7 @@ def fuzz(input_corpus, output_corpus, target_binary):\n]\nif os.listdir(input_corpus): # Important, otherwise Eclipser crashes.\ncommand += ['-i', input_corpus]\n- print('[run_fuzzer] Running command: ' + ' '.join(command))\n+ print('[fuzz] Running command: ' + ' '.join(command))\nsubprocess.Popen(command)\nprocess = Process(target=copy_corpus_directory,\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/honggfuzz/fuzzer.py", "new_path": "fuzzers/honggfuzz/fuzzer.py", "diff": "@@ -19,8 +19,6 @@ import subprocess\nfrom fuzzers import utils\n-# OUT environment variable is the location of build directory (default is /out).\n-\ndef build():\n\"\"\"Build benchmark.\"\"\"\n@@ -43,7 +41,7 @@ def fuzz(input_corpus, output_corpus, target_binary):\nif not os.path.exists(output_corpus):\nos.makedirs(output_corpus)\n- print('[run_fuzzer] Running target with honggfuzz')\n+ print('[fuzz] Running target with honggfuzz')\ncommand = [\n'./honggfuzz',\n'--persistent',\n@@ -60,5 +58,5 @@ def fuzz(input_corpus, output_corpus, target_binary):\ncommand.extend(['--dict', dictionary_path])\ncommand.extend(['--', target_binary])\n- print('[run_fuzzer] Running command: ' + ' '.join(command))\n+ print('[fuzz] Running command: ' + ' '.join(command))\nsubprocess.check_call(command)\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/honggfuzz_qemu/builder.Dockerfile", "new_path": "fuzzers/honggfuzz_qemu/builder.Dockerfile", "diff": "ARG parent_image=gcr.io/fuzzbench/base-builder\nFROM $parent_image\n-# honggfuzz requires libfd and libunwid.\n+# Honggfuzz requires libbfd and libunwid.\nRUN apt-get update -y && \\\napt-get install -y \\\nlibbfd-dev \\\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/honggfuzz_qemu/fuzzer.py", "new_path": "fuzzers/honggfuzz_qemu/fuzzer.py", "diff": "@@ -19,8 +19,6 @@ import subprocess\nfrom fuzzers import utils\n-# OUT environment variable is the location of build directory (default is /out).\n-\ndef build():\n\"\"\"Build benchmark.\"\"\"\n@@ -46,7 +44,7 @@ def fuzz(input_corpus, output_corpus, target_binary):\nif not os.path.exists(output_corpus):\nos.makedirs(output_corpus)\n- print('[run_fuzzer] Running target with honggfuzz')\n+ print('[fuzz] Running target with honggfuzz')\ncommand = [\n'./honggfuzz',\n'--rlimit_rss',\n@@ -62,5 +60,5 @@ def fuzz(input_corpus, output_corpus, target_binary):\ncommand.extend(['--dict', dictionary_path])\ncommand.extend(['--', './qemu-x86_64', target_binary])\n- print('[run_fuzzer] Running command: ' + ' '.join(command))\n+ print('[fuzz] Running command: ' + ' '.join(command))\nsubprocess.check_call(command)\n" } ]

[ { "change_type": "MODIFY", "old_path": "common/gcloud.py", "new_path": "common/gcloud.py", "diff": "import enum\nimport subprocess\n-import time\nfrom typing import List\nfrom common import experiment_utils\n-from common import logs\nfrom common import new_process\n# Constants for dispatcher specs.\n@@ -35,33 +33,6 @@ RUNNER_BOOT_DISK_SIZE = '30GB'\nINSTANCE_BATCH_SIZE = 100\n-def ssh(instance: str, *args, **kwargs):\n- \"\"\"SSH into |instance|.\"\"\"\n- zone = kwargs.pop('zone', None)\n- command = kwargs.pop('command', None)\n- ssh_command = ['gcloud', 'beta', 'compute', 'ssh', instance]\n- if command:\n- ssh_command.append('--command=%s' % command)\n- if zone:\n- ssh_command.append('--zone=%s' % zone)\n- return new_process.execute(ssh_command, *args, **kwargs)\n-\n-\n-def robust_begin_gcloud_ssh(instance_name: str, zone: str):\n- \"\"\"Try to SSH into an instance, |instance_name| in |zone| that might not be\n- ready.\"\"\"\n- for _ in range(10):\n- result = ssh(instance_name,\n- zone=zone,\n- command='echo ping',\n- expect_zero=False)\n- if result.retcode == 0:\n- return\n- logs.info('GCP instance isn\\'t ready yet. Rerunning SSH momentarily.')\n- time.sleep(5)\n- raise Exception('Couldn\\'t SSH to instance.')\n-\n-\nclass InstanceType(enum.Enum):\n\"\"\"Types of instances we need for the experiment.\"\"\"\nDISPATCHER = 0\n" }, { "change_type": "MODIFY", "old_path": "common/test_gcloud.py", "new_path": "common/test_gcloud.py", "diff": "from unittest import mock\n-import pytest\n-\nfrom common import gcloud\nfrom common import new_process\nfrom test_libs import utils as test_utils\n@@ -26,49 +24,6 @@ ZONE = 'zone-a'\nCONFIG = {'cloud_compute_zone': ZONE, 'service_account': 'blah'}\n-def test_ssh():\n- \"\"\"Tests that ssh works as expected.\"\"\"\n- with test_utils.mock_popen_ctx_mgr() as mocked_popen:\n- gcloud.ssh(INSTANCE_NAME)\n- assert mocked_popen.commands == [[\n- 'gcloud', 'beta', 'compute', 'ssh', INSTANCE_NAME\n- ]]\n-\n-\[email protected](('kwargs_for_ssh', 'expected_argument'),\n- [\n- ({'command': 'ls'}, '--command=ls'),\n- ({'zone': ZONE}, '--zone=' + ZONE),\n- ]) # yapf: disable\n-def test_ssh_optional_arg(kwargs_for_ssh, expected_argument):\n- \"\"\"Tests that ssh works as expected when given an optional argument.\"\"\"\n- with test_utils.mock_popen_ctx_mgr() as mocked_popen:\n- gcloud.ssh(INSTANCE_NAME, **kwargs_for_ssh)\n- assert expected_argument in mocked_popen.commands[0]\n-\n-\[email protected]('time.sleep')\[email protected]('common.gcloud.ssh')\n-def test_robust_begin_gcloud_ssh_fail(_, mocked_ssh):\n- \"\"\"Tests that ssh works as expected.\"\"\"\n- with pytest.raises(Exception) as exception:\n- gcloud.robust_begin_gcloud_ssh(INSTANCE_NAME, ZONE)\n- assert mocked_ssh.call_count == 10\n- assert exception.value == 'Couldn\\'t SSH to instance.'\n-\n-\[email protected]('time.sleep')\[email protected]('common.gcloud.ssh')\n-def test_robust_begin_gcloud_ssh_pass(mocked_ssh, _):\n- \"\"\"Tests robust_begin_gcloud_ssh works as intended on google cloud.\"\"\"\n- mocked_ssh.return_value = new_process.ProcessResult(0, None, False)\n- gcloud.robust_begin_gcloud_ssh(INSTANCE_NAME, ZONE)\n- mocked_ssh.assert_called_with('instance-a',\n- command='echo ping',\n- expect_zero=False,\n- zone='zone-a')\n-\n-\ndef test_create_instance():\n\"\"\"Tests create_instance creates an instance.\"\"\"\nwith test_utils.mock_popen_ctx_mgr(returncode=1) as mocked_popen:\n" }, { "change_type": "MODIFY", "old_path": "experiment/dispatcher.py", "new_path": "experiment/dispatcher.py", "diff": "@@ -34,6 +34,7 @@ from experiment.build import builder\nfrom experiment import measurer\nfrom experiment import reporter\nfrom experiment import scheduler\n+from experiment import stop_experiment\nLOOP_WAIT_SECONDS = 5 * 60\n@@ -116,7 +117,7 @@ def dispatcher_main():\n# reason.\nmultiprocessing.set_start_method('spawn')\ndb_utils.initialize()\n- if os.getenv('LOCAL_EXPERIMENT'):\n+ if experiment_utils.is_local_experiment():\nmodels.Base.metadata.create_all(db_utils.engine)\nexperiment_config_file_path = os.path.join(fuzzer_config_utils.get_dir(),\n@@ -172,6 +173,17 @@ def main():\nexcept Exception as error:\nlogs.error('Error conducting experiment.')\nraise error\n+ experiment_config_file_path = os.path.join(fuzzer_config_utils.get_dir(),\n+ 'experiment.yaml')\n+\n+ if experiment_utils.is_local_experiment():\n+ return 0\n+\n+ if stop_experiment.stop_experiment(experiment_utils.get_experiment_name(),\n+ experiment_config_file_path):\n+ return 0\n+\n+ return 1\nif __name__ == '__main__':\n" }, { "change_type": "ADD", "old_path": null, "new_path": "experiment/resources/dispatcher-startup-script-template.sh", "diff": "+#!/bin/bash\n+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+echo 0 | sudo tee /proc/sys/kernel/yama/ptrace_scope\n+docker run --rm \\\n+ -e INSTANCE_NAME={{instance_name}} -e EXPERIMENT={{experiment}} \\\n+ -e CLOUD_PROJECT={{cloud_project}} \\\n+ -e EXPERIMENT_FILESTORE={{experiment_filestore}} \\\n+ -e POSTGRES_PASSWORD={{postgres_password}} \\\n+ -e CLOUD_SQL_INSTANCE_CONNECTION_NAME={{cloud_sql_instance_connection_name}} \\\n+ --cap-add=SYS_PTRACE --cap-add=SYS_NICE \\\n+ -v /var/run/docker.sock:/var/run/docker.sock --name=dispatcher-container \\\n+ {{docker_registry}}/dispatcher-image /work/startup-dispatcher.sh\n" }, { "change_type": "MODIFY", "old_path": "experiment/resources/runner-startup-script-template.sh", "new_path": "experiment/resources/runner-startup-script-template.sh", "diff": "# See the License for the specific language governing permissions and\n# limitations under the License.\n-## Configure the host.\n+# Configure the host.\n# Make everything ptrace-able.\necho 0 > /proc/sys/kernel/yama/ptrace_scope\n@@ -21,7 +21,7 @@ echo 0 > /proc/sys/kernel/yama/ptrace_scope\n# Do not notify external programs about core dumps.\necho core >/proc/sys/kernel/core_pattern\n-## Start docker.\n+# Start docker.\n{% if not local_experiment %}\nwhile ! docker pull {{docker_image_url}}\ndo\n" }, { "change_type": "MODIFY", "old_path": "experiment/run_experiment.py", "new_path": "experiment/run_experiment.py", "diff": "it needs to begin an experiment.\"\"\"\nimport argparse\n-import multiprocessing\nimport os\nimport re\nimport subprocess\nimport sys\nimport tarfile\n+import tempfile\nfrom typing import Dict, List\n+\n+import jinja2\nimport yaml\nfrom common import benchmark_utils\n@@ -35,7 +37,6 @@ from common import logs\nfrom common import new_process\nfrom common import utils\nfrom common import yaml_utils\n-from experiment import stop_experiment\nfrom src_analysis import experiment_changes\nBENCHMARKS_DIR = os.path.join(utils.ROOT_DIR, 'benchmarks')\n@@ -60,6 +61,8 @@ FILTER_SOURCE_REGEX = re.compile(r'('\nCONFIG_DIR = 'config'\n+RESOURCES_DIR = os.path.join(utils.ROOT_DIR, 'experiment', 'resources')\n+\ndef read_and_validate_experiment_config(config_filename: str) -> Dict:\n\"\"\"Reads |config_filename|, validates it, finds as many errors as possible,\n@@ -267,11 +270,8 @@ def start_dispatcher(config: Dict, config_dir: str):\n\"\"\"Start the dispatcher instance and run the dispatcher code on it.\"\"\"\ndispatcher = get_dispatcher(config)\n# Is dispatcher code being run manually (useful for debugging)?\n- manual_experiment = os.getenv('MANUAL_EXPERIMENT')\n- if not manual_experiment:\n- dispatcher.create_async()\ncopy_resources_to_bucket(config_dir, config)\n- if not manual_experiment:\n+ if not os.getenv('MANUAL_EXPERIMENT'):\ndispatcher.start()\n@@ -312,11 +312,6 @@ class BaseDispatcher:\nself.config = config\nself.instance_name = experiment_utils.get_dispatcher_instance_name(\nconfig['experiment'])\n- self.process = None\n-\n- def create_async(self):\n- \"\"\"Creates the dispatcher asynchronously.\"\"\"\n- raise NotImplementedError\ndef start(self):\n\"\"\"Start the experiment on the dispatcher.\"\"\"\n@@ -332,11 +327,10 @@ class LocalDispatcher:\nconfig['experiment'])\nself.process = None\n- def create_async(self):\n- \"\"\"Noop in local experiments.\"\"\"\n-\ndef start(self):\n\"\"\"Start the experiment on the dispatcher.\"\"\"\n+ container_name = 'dispatcher-container'\n+ logs.info('Started dispatcher with container name: %s', container_name)\nexperiment_filestore_path = os.path.abspath(\nself.config['experiment_filestore'])\nfilesystem.create_directory(experiment_filestore_path)\n@@ -390,7 +384,7 @@ class LocalDispatcher:\n'LOCAL_EXPERIMENT=True',\n'--cap-add=SYS_PTRACE',\n'--cap-add=SYS_NICE',\n- '--name=dispatcher-container',\n+ '--name=%s' % container_name,\ndocker_image_url,\n'/bin/bash',\n'-c',\n@@ -410,59 +404,47 @@ class LocalDispatcher:\nclass GoogleCloudDispatcher(BaseDispatcher):\n\"\"\"Class representing the dispatcher instance on Google Cloud.\"\"\"\n- def create_async(self):\n- \"\"\"Creates the instance asynchronously.\"\"\"\n- self.process = multiprocessing.Process(\n- target=gcloud.create_instance,\n- args=(self.instance_name, gcloud.InstanceType.DISPATCHER,\n- self.config))\n- self.process.start()\n-\ndef start(self):\n\"\"\"Start the experiment on the dispatcher.\"\"\"\n- # TODO(metzman): Replace this workflow with a startup script so we don't\n- # need to SSH into the dispatcher.\n- self.process.join() # Wait for dispatcher instance.\n- # Check that we can SSH into the instance.\n- gcloud.robust_begin_gcloud_ssh(self.instance_name,\n- self.config['cloud_compute_zone'])\n-\n- docker_registry = self.config['docker_registry']\n+ logs.info('Started dispatcher with instance name: %s',\n+ self.instance_name)\n+ with tempfile.NamedTemporaryFile(dir=os.getcwd(),\n+ mode='w') as startup_script:\n+ self.write_startup_script(startup_script)\n+ gcloud.create_instance(self.instance_name,\n+ gcloud.InstanceType.DISPATCHER,\n+ self.config,\n+ startup_script=startup_script.name)\n+\n+ def _render_startup_script(self):\n+ \"\"\"Renders the startup script template and returns the result as a\n+ string.\"\"\"\n+ jinja_env = jinja2.Environment(\n+ undefined=jinja2.StrictUndefined,\n+ loader=jinja2.FileSystemLoader(RESOURCES_DIR),\n+ )\n+ template = jinja_env.get_template(\n+ 'dispatcher-startup-script-template.sh')\ncloud_sql_instance_connection_name = (\nself.config['cloud_sql_instance_connection_name'])\n- command = (\n- 'echo 0 | sudo tee /proc/sys/kernel/yama/ptrace_scope && '\n- 'docker run --rm '\n- '-e INSTANCE_NAME=\"{instance_name}\" '\n- '-e EXPERIMENT=\"{experiment}\" '\n- '-e CLOUD_PROJECT=\"{cloud_project}\" '\n- '-e DOCKER_REGISTRY=\"{docker_registry}\" '\n- '-e EXPERIMENT_FILESTORE=\"{experiment_filestore}\" '\n- '-e POSTGRES_PASSWORD=\"{postgres_password}\" '\n- '-e CLOUD_SQL_INSTANCE_CONNECTION_NAME='\n- '\"{cloud_sql_instance_connection_name}\" '\n- '--cap-add=SYS_PTRACE --cap-add=SYS_NICE '\n- '-v /var/run/docker.sock:/var/run/docker.sock '\n- '--name=dispatcher-container '\n- '{docker_registry}/dispatcher-image '\n- '/work/startup-dispatcher.sh'\n- ).format(\n- instance_name=self.instance_name,\n- postgres_password=os.environ['POSTGRES_PASSWORD'],\n- experiment=self.config['experiment'],\n- # TODO(metzman): Create a function that sets env vars based on\n- # the contents of a dictionary, and use it instead of hardcoding\n- # the configs we use.\n- cloud_project=self.config['cloud_project'],\n- experiment_filestore=self.config['experiment_filestore'],\n- cloud_sql_instance_connection_name=(\n- cloud_sql_instance_connection_name),\n- docker_registry=docker_registry,\n- )\n- return gcloud.ssh(self.instance_name,\n- command=command,\n- zone=self.config['cloud_compute_zone'])\n+ kwargs = {\n+ 'instance_name': self.instance_name,\n+ 'postgres_password': os.environ['POSTGRES_PASSWORD'],\n+ 'experiment': self.config['experiment'],\n+ 'cloud_project': self.config['cloud_project'],\n+ 'experiment_filestore': self.config['experiment_filestore'],\n+ 'cloud_sql_instance_connection_name':\n+ (cloud_sql_instance_connection_name),\n+ 'docker_registry': self.config['docker_registry'],\n+ }\n+ return template.render(**kwargs)\n+\n+ def write_startup_script(self, startup_script_file):\n+ \"\"\"Get the startup script to start the experiment on the dispatcher.\"\"\"\n+ startup_script = self._render_startup_script()\n+ startup_script_file.write(startup_script)\n+ startup_script_file.flush()\ndef get_dispatcher(config: Dict) -> BaseDispatcher:\n@@ -530,9 +512,6 @@ def main():\nstart_experiment(args.experiment_name, args.experiment_config,\nargs.benchmarks, fuzzer_configs)\n- if not os.getenv('MANUAL_EXPERIMENT'):\n- stop_experiment.stop_experiment(args.experiment_name,\n- args.experiment_config)\nreturn 0\n" }, { "change_type": "MODIFY", "old_path": "experiment/stop_experiment.py", "new_path": "experiment/stop_experiment.py", "diff": "@@ -48,14 +48,14 @@ def stop_experiment(experiment_name, experiment_config_filename):\nif not experiment_instances:\nlogger.warning('No experiment instances found, no work to do.')\n- return 0\n+ return True\nif not gcloud.delete_instances(experiment_instances, cloud_compute_zone):\nlogger.error('Failed to stop experiment instances.')\n- return 1\n+ return False\nlogger.info('Successfully stopped experiment.')\n- return 0\n+ return True\ndef main():\n@@ -64,7 +64,7 @@ def main():\nprint(\"Usage {0} <experiment-name> <experiment-config.yaml>\")\nreturn 1\nlogs.initialize()\n- return stop_experiment(sys.argv[1], sys.argv[2])\n+ return 0 if stop_experiment(sys.argv[1], sys.argv[2]) else 1\nif __name__ == '__main__':\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_scheduler.py", "new_path": "experiment/test_scheduler.py", "diff": "@@ -91,7 +91,7 @@ def test_create_trial_instance(benchmark, expected_image, expected_target,\nexperiment_config):\n\"\"\"Test that create_trial_instance invokes create_instance\nand creates a startup script for the instance, as we expect it to.\"\"\"\n- expected_startup_script = '''## Start docker.\n+ expected_startup_script = '''# Start docker.\nwhile ! docker pull {docker_image_url}\ndo\n@@ -133,7 +133,7 @@ def test_create_trial_instance_local_experiment(benchmark, expected_image,\nstartup script for the instance, as we expect it to when running a\nlocal_experiment.\"\"\"\nos.environ['LOCAL_EXPERIMENT'] = str(True)\n- expected_startup_script = '''## Start docker.\n+ expected_startup_script = '''# Start docker.\ndocker run \\\\\n@@ -191,7 +191,7 @@ def _test_create_trial_instance( # pylint: disable=too-many-locals\nwith open(expected_startup_script_path) as file_handle:\ncontent = file_handle.read()\n- check_from = '## Start docker.'\n+ check_from = '# Start docker.'\nassert check_from in content\nscript_for_docker = content[content.find(check_from):]\nassert script_for_docker == expected_startup_script.format(\n" } ]

[ { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-07-09\n+ fuzzers:\n+ - aflplusplus_lto_dict\n+ - aflplusplus_lto\n+ - aflplusplus_ltoinstrim\n+ - aflplusplus_ctx\n+ - aflplusplus_ngram2\n+ - aflplusplus_optimal\n+ - aflplusplus_qemu\n+ - aflplusplus\n+\n- experiment: 2020-06-30\nfuzzers:\n- aflplusplus_qemu\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/presubmit.yml", "new_path": ".github/workflows/presubmit.yml", "diff": "@@ -10,10 +10,16 @@ jobs:\n- run: | # Needed for presubmit to work.\ngit fetch origin master --depth 1\ngit symbolic-ref refs/remotes/origin/HEAD refs/remotes/origin/master\n+\n- name: Setup Python environment\nuses: actions/[email protected]\nwith:\npython-version: 3.7\n+\n+ - name: Install dependencies\n+ run: |\n+ make install-dependencies\n+\n- name: Run presubmit checks\nrun: |\nFUZZBENCH_TEST_INTEGRATION=1 make presubmit\n" } ]

[ { "change_type": "MODIFY", "old_path": "service/automatic_run_experiment.py", "new_path": "service/automatic_run_experiment.py", "diff": "@@ -29,7 +29,6 @@ from common import yaml_utils\nfrom database import models\nfrom database import utils as db_utils\nfrom experiment import run_experiment\n-from experiment import stop_experiment\nlogger = logs.Logger('automatic_run_experiment') # pylint: disable=invalid-name\n@@ -215,7 +214,6 @@ def _run_experiment(experiment_name, fuzzer_configs, dry_run=False):\nreturn\nrun_experiment.start_experiment(experiment_name, EXPERIMENT_CONFIG_FILE,\nBENCHMARKS, fuzzer_configs)\n- stop_experiment.stop_experiment(experiment_name, EXPERIMENT_CONFIG_FILE)\ndef main():\n" }, { "change_type": "MODIFY", "old_path": "service/test_automatic_run_experiment.py", "new_path": "service/test_automatic_run_experiment.py", "diff": "@@ -57,10 +57,8 @@ def test_run_requested_experiment_pause_service(\[email protected]('experiment.run_experiment.start_experiment')\[email protected]('experiment.stop_experiment.stop_experiment')\[email protected]('service.automatic_run_experiment._get_requested_experiments')\ndef test_run_requested_experiment(mocked_get_requested_experiments,\n- mocked_stop_experiment,\nmocked_start_experiment, db):\n\"\"\"Tests that run_requested_experiment starts and stops the experiment\nproperly.\"\"\"\n@@ -114,9 +112,6 @@ def test_run_requested_experiment(mocked_get_requested_experiments,\nstart_experiment_call_args[0][0][3].sort(key=sort_key)\nassert start_experiment_call_args == expected_calls\n- mocked_stop_experiment.assert_called_with(expected_experiment_name,\n- expected_config_file)\n-\[email protected](\n('name', 'expected_result'), [('02000-1-1', False), ('2020-1-1', False),\n" } ]

[ { "change_type": "MODIFY", "old_path": "experiment/build/test_builder.py", "new_path": "experiment/build/test_builder.py", "diff": "import itertools\nimport os\n+import sys\nfrom unittest import mock\nimport pytest\n@@ -62,7 +63,8 @@ def get_benchmarks_or_fuzzers(benchmarks_or_fuzzers_directory, filename,\n]\n-# This test seems to hang on Python3.8+\[email protected](sys.version_info.minor > 7,\n+ reason='Test can hang on versions greater than 3.7')\[email protected]('experiment.build.builder.build_measurer')\[email protected]('time.sleep')\[email protected]('build_measurer_return_value', [True, False])\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_run_experiment.py", "new_path": "experiment/test_run_experiment.py", "diff": "@@ -226,7 +226,7 @@ def test_validate_experiment_name_invalid(experiment_name):\n# This test takes up to a minute to complete.\[email protected]\[email protected]\ndef test_copy_resources_to_bucket(tmp_path):\n\"\"\"Tests that copy_resources_to_bucket copies the correct resources.\"\"\"\n# Do this so that Ctrl-C doesn't pollute the repo.\n" }, { "change_type": "MODIFY", "old_path": "presubmit.py", "new_path": "presubmit.py", "diff": "@@ -210,6 +210,15 @@ def lint(paths: List[Path]) -> bool:\ndef pytype(paths: List[Path]) -> bool:\n\"\"\"Run pytype on |path| if it is a python file. Return False if it fails\ntype checking.\"\"\"\n+ # Pytype isn't supported on Python3.8+. See\n+ # https://github.com/google/pytype/issues/440.\n+ assert sys.version_info.major == 3, \"Need Python3.\"\n+ if sys.version_info.minor > 7:\n+ logs.error(\n+ 'Python version is: \"%s\". You should be using 3.7. '\n+ 'Not running pytype.', sys.version)\n+ return True\n+\npaths = [path for path in paths if is_python(path)]\nif not paths:\nreturn True\n" }, { "change_type": "MODIFY", "old_path": "pytest.ini", "new_path": "pytest.ini", "diff": "[pytest]\nnorecursedirs = docs/_site/* docs/vendor/* third_party/* .venv/*\n+\n+markers =\n+ slow: marks tests as slow (deselect with '-m \"not slow\"')\n\\ No newline at end of file\n" } ]

[ { "change_type": "MODIFY", "old_path": "service/experiment-config.yaml", "new_path": "service/experiment-config.yaml", "diff": "@@ -11,3 +11,10 @@ experiment_filestore: gs://fuzzbench-data\nreport_filestore: gs://www.fuzzbench.com/reports\ncloud_sql_instance_connection_name: \"fuzzbench:us-central1:postgres-experiment-db=tcp:5432\"\npreemptible_runners: true\n+\n+# This experiment should generate a report that is combined with other public\n+# \"production\" experiments.\n+merge_with_nonprivate: true\n+\n+# This experiment should be merged with other reports in later experiments.\n+private: false\n" } ]

[ { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-07-13\n+ fuzzers:\n+ - aflplusplus_ctx_default\n+ - aflplusplus_ctx_nosingle\n+ - aflplusplus_ctx_nozero\n+ - aflplusplus_ctx_nozerosingle\n+ - aflplusplus_ngram4\n+ - aflplusplus_ngram6\n+ - aflplusplus_ngram8\n+\n- experiment: 2020-07-09\nfuzzers:\n- aflplusplus_lto_dict\n" } ]

[ { "change_type": "MODIFY", "old_path": "service/run.bash", "new_path": "service/run.bash", "diff": "@@ -22,6 +22,8 @@ expriment_working_dir=/tmp/fuzzbench-automatic-experiment-working-dir\nrepo_path=/tmp/fuzzbench-automatic-experiment-repo\nrm -rf $repo_path $expriment_working_dir\n+mkdir $expriment_working_dir\n+\ngit clone https://github.com/google/fuzzbench.git $repo_path\ncd $repo_path\n@@ -30,6 +32,6 @@ source .venv/bin/activate\nexport PYTHONPATH=$repo_path\ncd $expriment_working_dir\n-python3 service/automatic_run_experiment.py diff\n+python3 service/automatic_run_experiment.py\nrm -rf $repo_path\n" } ]

[ { "change_type": "MODIFY", "old_path": "benchmarks/libxml2-v2.9.2/build.sh", "new_path": "benchmarks/libxml2-v2.9.2/build.sh", "diff": "@@ -33,7 +33,15 @@ build_lib() {\n)\n}\n-get_git_tag https://gitlab.gnome.org/GNOME/libxml2.git v2.9.2 SRC\n+git clone https://gitlab.gnome.org/GNOME/libxml2.git SRC\n+cd SRC\n+\n+# Git is converting CRLF to LF automatically and causing issues when checking\n+# out the branch. So use -f to ignore the complaint about lost changes that we\n+# don't even want.\n+git checkout -f v2.9.2\n+cd -\n+\nbuild_lib\n$CXX $CXXFLAGS -std=c++11 $SCRIPT_DIR/target.cc -I BUILD/include BUILD/.libs/libxml2.a $FUZZER_LIB -o $FUZZ_TARGET\n" } ]

[ { "change_type": "MODIFY", "old_path": "service/run.bash", "new_path": "service/run.bash", "diff": "# See the License for the specific language governing permissions and\n# limitations under the License.\n-# Use this script to clone a new copy of fuzzbench and run a diff experiment.\n+# Use this script to clone a new copy of fuzzbench and run a requested\n+# experiment.\n# Use a seperate working directory to run the experiment so we don't pollute\n# the source code with the config directory created by run_experiment.py\n@@ -30,8 +31,7 @@ cd $repo_path\nmake install-dependencies\nsource .venv/bin/activate\nexport PYTHONPATH=$repo_path\n-cd $expriment_working_dir\n+cd $expriment_working_dir\npython3 $repo_path/service/automatic_run_experiment.py\nrm -rf $repo_path\n-\n" }, { "change_type": "ADD", "old_path": null, "new_path": "service/setup.bash", "diff": "+#!/bin/bash -ex\n+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+# Use this script once to setup a machine for running the fuzzbench service.\n+\n+# Install a supported python version.\n+export PYTHON_VERSION=3.7.6\n+\n+sudo apt-get update -y && sudo apt-get install -y \\\n+ build-essential \\\n+ rsync \\\n+ curl \\\n+ zlib1g-dev \\\n+ libncurses5-dev \\\n+ libgdbm-dev \\\n+ libnss3-dev \\\n+ libssl-dev \\\n+ libreadline-dev \\\n+ libffi-dev \\\n+ virtualenv \\\n+ libbz2-dev \\\n+ liblzma-dev \\\n+ libsqlite3-dev\n+\n+cd /tmp/ && \\\n+ curl -O https://www.python.org/ftp/python/$PYTHON_VERSION/Python-$PYTHON_VERSION.tar.xz && \\\n+ tar -xvf Python-$PYTHON_VERSION.tar.xz && \\\n+ cd Python-$PYTHON_VERSION && \\\n+ ./configure --enable-loadable-sqlite-extensions --enable-optimizations && \\\n+ sudo make -j install && \\\n+ sudo rm -r /tmp/Python-$PYTHON_VERSION.tar.xz /tmp/Python-$PYTHON_VERSION\n+\n+# Download and run the cloud_sql_proxy\n+export cloud_sql_proxy_path=/tmp/cloud_sql_proxy\n+wget https://dl.google.com/cloudsql/cloud_sql_proxy.linux.amd64 -O \\\n+ $cloud_sql_proxy_path\n+chmod +x $cloud_sql_proxy_path\n+# This is a hardcoded value that only works for the official fuzzbench service.\n+$cloud_sql_proxy_path -instances=fuzzbench:us-central1:postgres-experiment-db=tcp:5432 &\n" } ]

[ { "change_type": "MODIFY", "old_path": "service/run.bash", "new_path": "service/run.bash", "diff": "# Use a seperate working directory to run the experiment so we don't pollute\n# the source code with the config directory created by run_experiment.py\n-expriment_working_dir=/tmp/fuzzbench-automatic-experiment-working-dir\n+experiment_working_dir=/tmp/fuzzbench-automatic-experiment-working-dir\nrepo_path=/tmp/fuzzbench-automatic-experiment-repo\n-rm -rf $repo_path $expriment_working_dir\n+rm -rf $repo_path $experiment_working_dir\n-mkdir $expriment_working_dir\n+mkdir $experiment_working_dir\ngit clone https://github.com/google/fuzzbench.git $repo_path\ncd $repo_path\n@@ -31,7 +31,7 @@ cd $repo_path\nmake install-dependencies\nsource .venv/bin/activate\nexport PYTHONPATH=$repo_path\n+cd $experiment_working_dir\n-cd $expriment_working_dir\npython3 $repo_path/service/automatic_run_experiment.py\nrm -rf $repo_path\n" }, { "change_type": "MODIFY", "old_path": "service/setup.bash", "new_path": "service/setup.bash", "diff": "@@ -42,10 +42,11 @@ cd /tmp/ && \\\nsudo make -j install && \\\nsudo rm -r /tmp/Python-$PYTHON_VERSION.tar.xz /tmp/Python-$PYTHON_VERSION\n-# Download and run the cloud_sql_proxy\n+# Download and run the cloud_sql_proxy.\nexport cloud_sql_proxy_path=/tmp/cloud_sql_proxy\nwget https://dl.google.com/cloudsql/cloud_sql_proxy.linux.amd64 -O \\\n$cloud_sql_proxy_path\nchmod +x $cloud_sql_proxy_path\n+\n# This is a hardcoded value that only works for the official fuzzbench service.\n$cloud_sql_proxy_path -instances=fuzzbench:us-central1:postgres-experiment-db=tcp:5432 &\n" } ]

[ { "change_type": "MODIFY", "old_path": "fuzzers/libfuzzer/builder.Dockerfile", "new_path": "fuzzers/libfuzzer/builder.Dockerfile", "diff": "@@ -17,7 +17,7 @@ FROM $parent_image\nRUN git clone https://github.com/llvm/llvm-project.git /llvm-project && \\\ncd /llvm-project/ && \\\n- git checkout d8981ce5b9f8caa567613b2bf5aa3095e0156130 && \\\n+ git checkout 12d1124c49beec0fb79d36944960e5bf0f236d4c && \\\ncd compiler-rt/lib/fuzzer && \\\n(for f in *.cpp; do \\\nclang++ -stdlib=libc++ -fPIC -O2 -std=c++11 $f -c & \\\n" } ]

[ { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-07-20\n+ fuzzers:\n+ - libfuzzer\n+\n- experiment: 2020-07-13\nfuzzers:\n- aflplusplus_ctx_default\n" } ]

[ { "change_type": "MODIFY", "old_path": "experiment/dispatcher.py", "new_path": "experiment/dispatcher.py", "diff": "@@ -59,7 +59,7 @@ def _initialize_experiment_in_db(experiment_config: dict,\ndb_utils.get_or_create(models.Experiment,\nname=experiment_config['experiment'],\ngit_hash=experiment_config['git_hash'],\n- private=experiment_config.get('private', False))\n+ private=experiment_config.get('private', True))\n])\n# TODO(metzman): Consider doing this without sqlalchemy. This can get\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/faq.md", "new_path": "docs/faq.md", "diff": "---\nlayout: default\ntitle: FAQ\n-has_children: true\n+has_children: false\nnav_order: 6\npermalink: /faq/\n---\n" }, { "change_type": "MODIFY", "old_path": "docs/getting-started/adding_a_new_fuzzer.md", "new_path": "docs/getting-started/adding_a_new_fuzzer.md", "diff": "@@ -280,10 +280,32 @@ make build-$FUZZER_NAME-all\n* Run `make presubmit` to lint your code and ensure all tests are passing.\n-* Run `make clear-cache` to clear docker containers' caches. Next time you build a project, the container will be built from scratch.\n-\n-* Add your fuzzer to the list in `.github/workflows/ci.yml` to enable building\n- it on continous integration.\n+* Run `make clear-cache` to clear docker containers' caches. Next time you build\n+ a project, the container will be built from scratch.\n+\n+## Requesting an experiment\n+\n+The FuzzBench service automatically runs experiments that are requested by users\n+twice a day at 6:00 AM PT (13:00 UTC) and 6:00 PM PT (01:00 UTC). If you want\n+the FuzzBench service to run an experiment on specific fuzzers (such as the one\n+you are adding): add an experiment request to\n+[service/experiment-requests.yaml](https://github.com/google/fuzzbench/blob/master/service/experiment-requests.yaml).\n+`service/experiment-requests.yaml` explains how to do this. At the end of the\n+experiment, FuzzBench will generate a report comparing your fuzzer to the latest\n+versions of other fuzzers, so you only need to include fuzzers that you've\n+modified in a meaningful way (i.e. fuzzers whose results are likely affected by\n+your\n+change). This report, and a real-time report of your experiment can be viewed at\n+`https://www.fuzzbench.com/reports/$YOUR_EXPERIMENT_NAME`. Note that real-time\n+reports may not appear until a few hours after the experiment starts since every\n+fuzzer-benchmark pair in the experiment must build in order for fuzzing to\n+start.\n+\n+## Submitting your integration\n+\n+* Add your fuzzer to the list in `.github/workflows/ci.yml` so that our\n+ continuous integration will test that your fuzzer can build and briefly run on\n+ all benchmarks once you've submitted a pull request.\n* Submit the integration in a\n[GitHub pull request](https://help.github.com/en/github/collaborating-with-issues-and-pull-requests/creating-a-pull-request).\n" }, { "change_type": "ADD", "old_path": null, "new_path": "docs/internal-documentation/internal_documentation.md", "diff": "+---\n+layout: default\n+title: Internal Documentation\n+has_children: true\n+nav_order: 7\n+permalink: /internal-documentation/\n+---\n+\n+# Internal documentation\n+\n+This section contains internal documentation that is meant for FuzzBench\n+maintainers. It is not useful for end users of FuzzBench.\n" }, { "change_type": "ADD", "old_path": null, "new_path": "docs/internal-documentation/service.md", "diff": "+---\n+layout: default\n+title: The FuzzBench Service\n+parent: Internal Documentation\n+nav_order: 1\n+permalink: /internal-documentation/service/\n+---\n+\n+# # The FuzzBench service\n+{: .no_toc}\n+\n+**Note:** This document and most of the `service/` directory is only intended\n+for use by FuzzBench maintainers. It will contain hardcoded values and\n+references to things that don't make sense for other users.\n+\n+- TOC\n+{:toc}\n+\n+## Overview\n+\n+This document discusses the FuzzBench service. The service works as follows:\n+When a user wants a new experiment they add the experiment to\n+`experiment-requests.yaml`. Twice a day at 6 AM PT (13:00 UTC) and 6 PM PT\n+(01:00 UTC) a cron job on the `service` instance will execute the script\n+`run.bash`. `run.bash` will clone FuzzBench and then execute\n+`automatic_run_experiment.py` which starts newly requested experiments.\n+\n+## Setting up an instance to run an experiment\n+\n+This shouldn't be necessary, but here are instructions in case the current\n+instance is lost.\n+1. Run `setup.bash`. This will build and install a supported Python version,\n+ download the `cloud_sql_proxy` and run it so that we have a connection to the\n+ db.\n+\n+1. Install the cron job. An example you can use is in the\n+ [crontab file](https://github.com/google/fuzzbench/tree/master/service/crontab).\n+ Note that you must fill in `POSTGRES_PASSWORD` and `$HOME`.\n+\n+1. Verify that the service is running. One way you can debug this is by looking\n+ at the stdout/stderr of `run.bash` which is saved in\n+ `/tmp/fuzzbench-service.log`. If something isn't working you should probably\n+ verify that `run.bash` works on its own. Note that `run.bash` is executed\n+ from a checkout of FuzzBench that isn't automatically updated. So if you need\n+ to update you must do so with `git pull --rebase`.\n+\n+## Automatic merging\n+\n+Experiments that are run using the service will be marked as nonprivate and on\n+completion automatically merge using clobbering.\n" }, { "change_type": "ADD", "old_path": null, "new_path": "service/crontab", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+# Example crontab file that can be used to run the service. Note that the path\n+# needs to be set because cron jobs will typically use different paths than\n+# users (it breaks without this). Though the SQL database can only be used by\n+# Google accounts that have permission, we don't include the password here.\n+\n+# m h dom mon dow command\n+00 1,13 * * * export PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin:/opt/puppetlabs/bin; export POSTGRES_PASSWORD=\"\"; $HOME/fuzzbench/service/run.bash\n\\ No newline at end of file\n" }, { "change_type": "MODIFY", "old_path": "service/run.bash", "new_path": "service/run.bash", "diff": "# Use this script to clone a new copy of fuzzbench and run a requested\n# experiment.\n+# Append stdout and stderr to /tmp/fuzzbench-service.logs. Redirecting the\n+# stdout/stderr of this script in the cron command didn't work for some reason.\n+exec &>> /tmp/fuzzbench-service.logs\n+\n# Use a seperate working directory to run the experiment so we don't pollute\n# the source code with the config directory created by run_experiment.py\nexperiment_working_dir=/tmp/fuzzbench-automatic-experiment-working-dir\n@@ -31,7 +35,7 @@ cd $repo_path\nmake install-dependencies\nsource .venv/bin/activate\nexport PYTHONPATH=$repo_path\n-cd $experiment_working_dir\n+cd $experiment_working_dir\npython3 $repo_path/service/automatic_run_experiment.py\nrm -rf $repo_path\n" } ]

[ { "change_type": "MODIFY", "old_path": "docker/gcb/coverage.yaml", "new_path": "docker/gcb/coverage.yaml", "diff": "@@ -31,10 +31,10 @@ steps:\n# Use two tags so that the image builds properly and we can push it to the\n# correct location.\n'--tag',\n- 'gcr.io/fuzzbench/builders/coverage:${_EXPERIMENT}:${_EXPERIMENT}',\n+ 'gcr.io/fuzzbench/builders/coverage:${_EXPERIMENT}',\n'--tag',\n- '${_REPO}/builders/coverage:${_EXPERIMENT}:${_EXPERIMENT}',\n+ '${_REPO}/builders/coverage:${_EXPERIMENT}',\n'--cache-from',\n'${_REPO}/builders/coverage',\n" } ]

[ { "change_type": "RENAME", "old_path": "fuzzbench/fake_jobs.py", "new_path": "fuzzbench/jobs.py", "diff": "# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n# See the License for the specific language governing permissions and\n# limitations under the License.\n-\"\"\"Fake jobs.\"\"\"\n+\"\"\"Defines all required jobs for one experiment.\"\"\"\n+import os\nimport subprocess\n-import time\n+BASE_TAG = 'gcr.io/fuzzbench'\n-def build_image(name):\n- \"\"\"Build a Docker image.\"\"\"\n- print('Building', name)\n- subprocess.run(['docker', '--version'], check=True)\n- time.sleep(3)\n+\n+def build_image(name: str):\n+ \"\"\"Builds a Docker image and returns whether it succeeds.\"\"\"\n+ image_tag = os.path.join(BASE_TAG, name)\n+ subprocess.run(['docker', 'pull', image_tag], check=True)\n+ subprocess.run(\n+ ['docker', 'build', '--tag', image_tag,\n+ os.path.join('docker', name)],\n+ check=True)\nreturn True\ndef run_trial():\n- \"\"\"Run a trial.\"\"\"\n+ \"\"\"Runs a trial.\"\"\"\nreturn True\ndef measure_corpus_snapshot():\n- \"\"\"Measure a corpus snapshot.\"\"\"\n+ \"\"\"Measures a corpus snapshot.\"\"\"\nreturn True\n" }, { "change_type": "MODIFY", "old_path": "fuzzbench/run_experiment.py", "new_path": "fuzzbench/run_experiment.py", "diff": "@@ -18,22 +18,37 @@ import time\nimport redis\nimport rq\n-from fuzzbench import fake_jobs\n+from fuzzbench import jobs\ndef run_experiment():\n\"\"\"Main experiment logic.\"\"\"\nprint('Initializing the job queue.')\nqueue = rq.Queue()\n- jobs = []\n- for i in range(6):\n- jobs.append(queue.enqueue(fake_jobs.build_image, 'something-%d' % i))\n+ jobs_list = []\n+ jobs_list.append(\n+ queue.enqueue(jobs.build_image,\n+ 'base-image',\n+ job_timeout=600,\n+ job_id='base-image'))\n+ jobs_list.append(\n+ queue.enqueue(jobs.build_image,\n+ 'base-builder',\n+ job_timeout=600,\n+ job_id='base-builder',\n+ depends_on='base-image'))\n+ jobs_list.append(\n+ queue.enqueue(jobs.build_image,\n+ 'base-runner',\n+ job_timeout=600,\n+ job_id='base-runner',\n+ depends_on='base-image'))\nwhile True:\nprint('Current status of jobs:')\n- for job in jobs:\n+ for job in jobs_list:\nprint(' %s%s : %s' % (job.func_name, job.args, job.get_status()))\n- if all([job.result is not None for job in jobs]):\n+ if all([job.result is not None for job in jobs_list]):\nbreak\ntime.sleep(3)\nprint('All done!')\n" }, { "change_type": "MODIFY", "old_path": "fuzzbench/test_e2e_run.py", "new_path": "fuzzbench/test_e2e_run.py", "diff": "@@ -18,17 +18,38 @@ test.\"\"\"\nimport os\nimport pytest\n+import redis\n+from rq.job import Job\n+\n+\[email protected](scope='class')\n+def redis_connection():\n+ \"\"\"Returns the default redis server connection.\"\"\"\n+ return redis.Redis(host='queue-server')\n# pylint: disable=no-self-use\[email protected]('E2E_INTEGRATION_TEST' not in os.environ,\nreason='Not running end-to-end test.')\[email protected]('redis_connection')\nclass TestEndToEndRunResults:\n\"\"\"Checks the result of a test experiment run.\"\"\"\n- def test_all_jobs_finished_sucessfully(self):\n- \"\"\"Fake test to be implemented later.\"\"\"\n- assert True\n+ def test_jobs_dependency(self, redis_connection): # pylint: disable=redefined-outer-name\n+ \"\"\"Tests that jobs dependency preserves during working.\"\"\"\n+ jobs = {\n+ name: Job.fetch(name, connection=redis_connection)\n+ for name in ['base-image', 'base-builder', 'base-runner']\n+ }\n+ assert jobs['base-image'].ended_at <= jobs['base-builder'].started_at\n+ assert jobs['base-image'].ended_at <= jobs['base-runner'].started_at\n+\n+ def test_all_jobs_finished_successfully(self, redis_connection): # pylint: disable=redefined-outer-name\n+ \"\"\"Tests all jobs finished successully.\"\"\"\n+ jobs = Job.fetch_many(['base-image', 'base-builder', 'base-runner'],\n+ connection=redis_connection)\n+ for job in jobs:\n+ assert job.get_status() == 'finished'\ndef test_measurement_jobs_were_started_before_trial_jobs_finished(self):\n\"\"\"Fake test to be implemented later.\"\"\"\n" } ]

[ { "change_type": "MODIFY", "old_path": "fuzzers/libfuzzer/fuzzer.py", "new_path": "fuzzers/libfuzzer/fuzzer.py", "diff": "@@ -25,6 +25,18 @@ def build():\n# /usr/lib/libFuzzer.a as the FUZZER_LIB for the main fuzzing binary. This\n# allows us to link against a version of LibFuzzer that we specify.\ncflags = ['-fsanitize=fuzzer-no-link']\n+\n+ # Can be removed once the patch https://reviews.llvm.org/D83987 lands\n+ # and appears in gcr.io/fuzzbench/base-builder\n+ cflags += ['-fno-builtin-memcmp']\n+ cflags += ['-fno-builtin-strncmp']\n+ cflags += ['-fno-builtin-strcmp']\n+ cflags += ['-fno-builtin-strncasecmp']\n+ cflags += ['-fno-builtin-strcasecmp']\n+ cflags += ['-fno-builtin-strstr']\n+ cflags += ['-fno-builtin-strcasestr']\n+ cflags += ['-fno-builtin-memmem']\n+\nutils.append_flags('CFLAGS', cflags)\nutils.append_flags('CXXFLAGS', cflags)\n" }, { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-07-22\n+ fuzzers:\n+ - libfuzzer\n+\n- experiment: 2020-07-20\nfuzzers:\n- libfuzzer\n" } ]

[ { "change_type": "MODIFY", "old_path": "docker/gcb/oss-fuzz-fuzzer.yaml", "new_path": "docker/gcb/oss-fuzz-fuzzer.yaml", "diff": "@@ -175,7 +175,7 @@ steps:\n'benchmark=${_BENCHMARK}',\n'--file',\n- 'docker/oss-fuzz-runner/Dockerfile',\n+ 'docker/benchmark-runner/Dockerfile',\n'.',\n]\n" }, { "change_type": "MODIFY", "old_path": "docker/generate_makefile.py", "new_path": "docker/generate_makefile.py", "diff": "@@ -125,11 +125,11 @@ endif\n\"\"\"\nOSS_FUZZER_BENCHMARK_RUN_TARGETS_TEMPLATE = \"\"\"\n-build-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-oss-fuzz-runner\n+build-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-runner\n-pull-{fuzzer}-{benchmark}: .pull-{fuzzer}-{benchmark}-oss-fuzz-runner\n+pull-{fuzzer}-{benchmark}: .pull-{fuzzer}-{benchmark}-runner\n-run-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-oss-fuzz-runner\n+run-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-runner\ndocker run \\\\\n--cpus=1 \\\\\n--cap-add SYS_NICE \\\\\n@@ -142,7 +142,7 @@ run-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-oss-fuzz-runner\n-e FUZZ_TARGET=$({benchmark}-fuzz-target) \\\\\n-it {base_tag}/runners/{fuzzer}/{benchmark}\n-test-run-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-oss-fuzz-runner\n+test-run-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-runner\ndocker run \\\\\n--cap-add SYS_NICE \\\\\n--cap-add SYS_PTRACE \\\\\n@@ -156,7 +156,7 @@ test-run-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-oss-fuzz-runner\n-e SNAPSHOT_PERIOD=10 \\\\\n{base_tag}/runners/{fuzzer}/{benchmark}\n-debug-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-oss-fuzz-runner\n+debug-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-runner\ndocker run \\\\\n--cpus=1 \\\\\n--cap-add SYS_NICE \\\\\n@@ -209,16 +209,16 @@ ifeq (,$(filter {fuzzer},coverage coverage_source_based))\n.pull-{fuzzer}-{benchmark}-oss-fuzz-intermediate-runner: pull-base-runner\ndocker pull {base_tag}/runners/{fuzzer}/{benchmark}-intermediate\n-.{fuzzer}-{benchmark}-oss-fuzz-runner: .{fuzzer}-{benchmark}-oss-fuzz-builder .{fuzzer}-{benchmark}-oss-fuzz-intermediate-runner\n+.{fuzzer}-{benchmark}-runner: .{fuzzer}-{benchmark}-oss-fuzz-builder .{fuzzer}-{benchmark}-oss-fuzz-intermediate-runner\ndocker build \\\\\n--tag {base_tag}/runners/{fuzzer}/{benchmark} \\\\\n--build-arg fuzzer={fuzzer} \\\\\n--build-arg benchmark={benchmark} \\\\\n$(call cache_from,{base_tag}/runners/{fuzzer}/{benchmark}) \\\\\n- --file docker/oss-fuzz-runner/Dockerfile \\\\\n+ --file docker/benchmark-runner/Dockerfile \\\\\n.\n-.pull-{fuzzer}-{benchmark}-oss-fuzz-runner: .pull-{fuzzer}-{benchmark}-oss-fuzz-builder .pull-{fuzzer}-{benchmark}-oss-fuzz-intermediate-runner\n+.pull-{fuzzer}-{benchmark}-runner: .pull-{fuzzer}-{benchmark}-oss-fuzz-builder .pull-{fuzzer}-{benchmark}-oss-fuzz-intermediate-runner\ndocker pull {base_tag}/runners/{fuzzer}/{benchmark}\n\"\"\" + OSS_FUZZER_BENCHMARK_RUN_TARGETS_TEMPLATE + \"\"\"\n" }, { "change_type": "DELETE", "old_path": "docker/oss-fuzz-runner/Dockerfile", "new_path": null, "diff": "-# Copyright 2020 Google LLC\n-#\n-# Licensed under the Apache License, Version 2.0 (the \"License\");\n-# you may not use this file except in compliance with the License.\n-# You may obtain a copy of the License at\n-#\n-# http://www.apache.org/licenses/LICENSE-2.0\n-#\n-# Unless required by applicable law or agreed to in writing, software\n-# distributed under the License is distributed on an \"AS IS\" BASIS,\n-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n-# See the License for the specific language governing permissions and\n-# limitations under the License.\n-#\n-# Dockerfile for running a specific OSS-Fuzz project as a benchmark for a\n-# specific fuzzer.\n-#\n-# This Dockerfile adds essential files into the runner image, regardless of\n-# whether it was built from `benchmark-runner/Dockerfile` or a custom\n-# `runner.Dockerfile`.\n-#\n-# The benchmark/fuzzer pair is defined by build arguments. To specify them, pass\n-# the following arguments to docker build:\n-#\n-# $ docker build \\\n-# --build-arg benchmark=bloaty_fuzz_target \\\n-# --build-arg fuzzer=afl \\\n-# ...\n-\n-ARG fuzzer\n-ARG benchmark\n-\n-# We use Docker's multi-stage build feature to create a minimal runner image,\n-# separate from the sometimes bulky builder images.\n-\n-# We take the already built builder image for the given fuzzer/benchmark pair\n-# and refer to it as \"builder\", so we can copy the build artifacts from it.\n-FROM gcr.io/fuzzbench/builders/$fuzzer/$benchmark AS builder\n-\n-# We base the runner image from the intermediate runner image, defined by the\n-# runner.Dockerfile of each fuzzer.\n-FROM gcr.io/fuzzbench/runners/$fuzzer/$benchmark-intermediate\n-\n-# Set up the directory for the build artifacts.\n-ENV WORKDIR /out\n-RUN mkdir -p $WORKDIR\n-WORKDIR $WORKDIR\n-\n-# Copy over all the build artifacts (without * to preserve directory structure).\n-# This also copies seed and dictionary files if they are available.\n-COPY --from=builder /out/ ./\n-# Copy the fuzzer.py file.\n-COPY --from=builder /src/fuzzer.py .\n-# Copy the fuzzers directory.\n-COPY --from=builder /src/fuzzers fuzzers\n-# Create empty __init__.py to allow python deps to work.\n-RUN touch __init__.py\n-\n-# Define environment variables used when we run the fuzzer:\n-# - Directory to get starting seeds from.\n-ENV SEED_CORPUS_DIR=$WORKDIR/seeds\n-# - Where to place new test cases generated by the fuzzer.\n-ENV OUTPUT_CORPUS_DIR=$WORKDIR/corpus\n-\n-# Create the seeds directory if it doesn't exist.\n-RUN mkdir -p $SEED_CORPUS_DIR $OUTPUT_CORPUS_DIR\n-\n-# Copy the source code into the image. We do this here instead of in base-image\n-# because it is likely to change, particularly in development. If this were done\n-# earlier, build cycles would be intolerably slow.\n-ENV ROOT_DIR=/src\n-COPY common $ROOT_DIR/common\n-COPY experiment/runner.py $ROOT_DIR/experiment/runner.py\n-COPY docker/benchmark-runner $ROOT_DIR/docker/benchmark-runner\n-\n-ENV PYTHONPATH=$ROOT_DIR\n-\n-# |VIRTUALENV_DIR| is set so that python code can know the location of current\n-# virtualenv directory and strip it if needed to execute in system python\n-# environment.\n-ENV VIRTUALENV_DIR=$ROOT_DIR/.venv\n-RUN virtualenv --python=$(which python3) $VIRTUALENV_DIR\n-RUN /bin/bash -c \"source $VIRTUALENV_DIR/bin/activate && \\\n- pip3 install -r $ROOT_DIR/docker/benchmark-runner/requirements.txt\"\n-RUN chmod +x $ROOT_DIR/docker/benchmark-runner/startup-runner.sh\n-ENTRYPOINT $ROOT_DIR/docker/benchmark-runner/startup-runner.sh\n" } ]

[ { "change_type": "MODIFY", "old_path": "fuzzers/libfuzzer/builder.Dockerfile", "new_path": "fuzzers/libfuzzer/builder.Dockerfile", "diff": "@@ -17,7 +17,7 @@ FROM $parent_image\nRUN git clone https://github.com/llvm/llvm-project.git /llvm-project && \\\ncd /llvm-project/ && \\\n- git checkout 12d1124c49beec0fb79d36944960e5bf0f236d4c && \\\n+ git checkout b52b2e1c188072e3cbc91500cfd503fb26d50ffc && \\\ncd compiler-rt/lib/fuzzer && \\\n(for f in *.cpp; do \\\nclang++ -stdlib=libc++ -fPIC -O2 -std=c++11 $f -c & \\\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/libfuzzer/fuzzer.py", "new_path": "fuzzers/libfuzzer/fuzzer.py", "diff": "@@ -26,8 +26,9 @@ def build():\n# allows us to link against a version of LibFuzzer that we specify.\ncflags = ['-fsanitize=fuzzer-no-link']\n- # Can be removed once the patch https://reviews.llvm.org/D83987 lands\n- # and appears in gcr.io/fuzzbench/base-builder\n+ # Can be removed once the patch https://reviews.llvm.org/D83987\n+ # appears in gcr.io/fuzzbench/base-builder\n+ cflags += ['-fno-builtin-bcmp']\ncflags += ['-fno-builtin-memcmp']\ncflags += ['-fno-builtin-strncmp']\ncflags += ['-fno-builtin-strcmp']\n" } ]

[ { "change_type": "MODIFY", "old_path": "fuzzbench/run_experiment.py", "new_path": "fuzzbench/run_experiment.py", "diff": "@@ -46,8 +46,14 @@ def run_experiment():\nwhile True:\nprint('Current status of jobs:')\n+ print('\\tqueued:\\t%d' % queue.count)\n+ print('\\tstarted:\\t%d' % queue.started_job_registry.count)\n+ print('\\tdeferred:\\t%d' % queue.deferred_job_registry.count)\n+ print('\\tfinished:\\t%d' % queue.finished_job_registry.count)\n+ print('\\tfailed:\\t%d' % queue.failed_job_registry.count)\nfor job in jobs_list:\n- print(' %s%s : %s' % (job.func_name, job.args, job.get_status()))\n+ print(' %s : %s\\t(%s)' % (job.func_name, job.get_status(), job.id))\n+\nif all([job.result is not None for job in jobs_list]):\nbreak\ntime.sleep(3)\n" } ]

[ { "change_type": "MODIFY", "old_path": "Makefile", "new_path": "Makefile", "diff": "# Running experiments locally.\n-run-experiment: export COMPOSE_PROJECT_NAME := fuzzbench\n-run-experiment: export COMPOSE_FILE := compose/fuzzbench.yaml\n+run-experiment stop-experiment: export COMPOSE_PROJECT_NAME := fuzzbench\n+run-experiment stop-experiment: export COMPOSE_FILE := compose/fuzzbench.yaml\nrun-experiment:\ndocker-compose up --build --scale worker=2 --detach\ndocker-compose logs --follow run-experiment\ndocker-compose down\n+# Running this is only necessary if `run-experiment` was interrupted and\n+# containers were not cleaned up.\n+stop-experiment:\n+ docker-compose down\n+\n# Development.\n-run-end-to-end-test: export COMPOSE_PROJECT_NAME := e2e-test\n-run-end-to-end-test: export COMPOSE_FILE := compose/fuzzbench.yaml:compose/e2e-test.yaml\n+run-end-to-end-test stop-end-to-end-test: export COMPOSE_PROJECT_NAME := e2e-test\n+run-end-to-end-test stop-end-to-end-test: export COMPOSE_FILE := compose/fuzzbench.yaml:compose/e2e-test.yaml\nrun-end-to-end-test:\ndocker-compose build\ndocker-compose up --detach queue-server\n@@ -32,6 +37,11 @@ run-end-to-end-test:\ndocker-compose run run-tests; STATUS=$$?; \\\ndocker-compose down; exit $$STATUS\n+# Running this is only necessary if `run-end-to-end-test` was interrupted and\n+# containers were not cleaned up.\n+stop-end-to-end-test:\n+ docker-compose down\n+\ninclude docker/build.mk\ninclude docker/generated.mk\n" } ]

[ { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-07-28\n+ fuzzers:\n+ - afl\n+ - honggfuzz\n+ - libfuzzer\n+ - entropic\n+\n- experiment: 2020-07-27\nfuzzers:\n- afl\n" } ]

[ { "change_type": "MODIFY", "old_path": "experiment/dispatcher.py", "new_path": "experiment/dispatcher.py", "diff": "@@ -133,18 +133,16 @@ def dispatcher_main():\nargs=(experiment.config,))\nscheduler_loop_thread.start()\n- max_total_time = experiment.config['max_total_time']\n- measurer_loop_process = multiprocessing.Process(\n- target=measurer.measure_loop,\n- args=(experiment.experiment_name, max_total_time))\n+ measurer_main_process = multiprocessing.Process(\n+ target=measurer.measure_main, args=(experiment.config,))\n- measurer_loop_process.start()\n+ measurer_main_process.start()\nis_complete = False\nwhile True:\ntime.sleep(LOOP_WAIT_SECONDS)\nif not scheduler_loop_thread.is_alive():\n- is_complete = not measurer_loop_process.is_alive()\n+ is_complete = not measurer_main_process.is_alive()\n# Generate periodic output reports.\nreporter.output_report(experiment.config, in_progress=not is_complete)\n@@ -155,7 +153,7 @@ def dispatcher_main():\nlogs.info('Dispatcher finished.')\nscheduler_loop_thread.join()\n- measurer_loop_process.join()\n+ measurer_main_process.join()\ndef main():\n" }, { "change_type": "MODIFY", "old_path": "experiment/measurer.py", "new_path": "experiment/measurer.py", "diff": "@@ -53,6 +53,7 @@ NUM_RETRIES = 3\nRETRY_DELAY = 3\nFAIL_WAIT_SECONDS = 30\nSNAPSHOT_QUEUE_GET_TIMEOUT = 1\n+COV_DIFF_QUEUE_GET_TIMEOUT = 1\nSNAPSHOTS_BATCH_SAVE_SIZE = 100\n@@ -67,13 +68,126 @@ def exists_in_experiment_filestore(path: pathlib.Path) -> bool:\nmust_exist=False).retcode == 0\n+def get_fuzzer_benchmark_key(fuzzer: str, benchmark: str):\n+ \"\"\"Return the key in coverage dict for a pair of fuzzer-benchmark.\"\"\"\n+ return fuzzer + ' ' + benchmark\n+\n+\n+def get_trial_ids(experiment: str, fuzzer: str, benchmark: str):\n+ \"\"\"Get ids of all finished trials for a pair of fuzzer and benchmark.\"\"\"\n+ trial_ids = [\n+ trial_id_tuple[0]\n+ for trial_id_tuple in db_utils.query(models.Trial.id).filter(\n+ models.Trial.experiment == experiment, models.Trial.fuzzer ==\n+ fuzzer, models.Trial.benchmark == benchmark,\n+ ~models.Trial.preempted)\n+ ]\n+ return trial_ids\n+\n+\n+def get_coverage_infomation(coverage_summary_file):\n+ \"\"\"Read the coverage information from |coverage_summary_file|\n+ and skip possible warnings in the file.\"\"\"\n+ with open(coverage_summary_file) as summary:\n+ return json.loads(summary.readlines()[-1])\n+\n+\n+def store_coverage_data(experiment_config: dict):\n+ \"\"\"Generate the specific coverage data and store in cloud bucket.\"\"\"\n+ logger.info('Start storing coverage data')\n+ with multiprocessing.Pool() as pool, multiprocessing.Manager() as manager:\n+ q = manager.Queue() # pytype: disable=attribute-error\n+ covered_regions = get_all_covered_regions(experiment_config, pool, q)\n+ json_src_dir = get_experiment_folders_dir()\n+ json_src = os.path.join(json_src_dir, 'covered_regions.json')\n+ with open(json_src, 'w') as src_file:\n+ json.dump(covered_regions, src_file)\n+ json_dst = exp_path.filestore(json_src)\n+ filestore_utils.cp(json_src, json_dst)\n+ logger.info('Finished storing coverage data')\n+\n+\n+def get_all_covered_regions(experiment_config: dict, pool, q) -> dict:\n+ \"\"\"Get regions covered for each pair for fuzzer and benchmark.\"\"\"\n+ logger.info('Measuring all fuzzer-benchmark pairs for final coverage data.')\n+\n+ benchmarks = experiment_config['benchmarks'].split(',')\n+ fuzzers = experiment_config['fuzzers'].split(',')\n+ experiment = experiment_config['experiment']\n+\n+ get_covered_region_args = [(experiment, fuzzer, benchmark, q)\n+ for fuzzer in fuzzers\n+ for benchmark in benchmarks]\n+\n+ result = pool.starmap_async(get_covered_region, get_covered_region_args)\n+\n+ # Poll the queue for covered region data and save them in a dict until the\n+ # pool is done processing each combination of fuzzers and benchmarks.\n+ all_covered_regions = {}\n+\n+ while True:\n+ try:\n+ covered_regions = q.get(timeout=COV_DIFF_QUEUE_GET_TIMEOUT)\n+ all_covered_regions.update(covered_regions)\n+ except queue.Empty:\n+ if result.ready():\n+ # If \"ready\" that means pool has finished. Since it is\n+ # finished and the queue is empty, we can stop checking\n+ # the queue for more covered regions.\n+ logger.debug(\n+ 'Finished call to map with get_all_covered_regions.')\n+ break\n+\n+ for key in all_covered_regions:\n+ all_covered_regions[key] = list(all_covered_regions[key])\n+ logger.info('Done measuring all coverage data.')\n+ return all_covered_regions\n+\n+\n+def get_covered_region(experiment: str, fuzzer: str, benchmark: str,\n+ q: multiprocessing.Queue):\n+ \"\"\"Get the final covered region for a specific pair of fuzzer-benchmark.\"\"\"\n+ initialize_logs()\n+ logger.debug('Measuring covered region: fuzzer: %s, benchmark: %s.', fuzzer,\n+ benchmark)\n+ key = get_fuzzer_benchmark_key(fuzzer, benchmark)\n+ covered_regions = {key: set()}\n+ trial_ids = get_trial_ids(experiment, fuzzer, benchmark)\n+ for trial_id in trial_ids:\n+ logger.info('Measuring covered region: trial_id = %d.', trial_id)\n+ snapshot_logger = logs.Logger('measurer',\n+ default_extras={\n+ 'fuzzer': fuzzer,\n+ 'benchmark': benchmark,\n+ 'trial_id': str(trial_id),\n+ })\n+ snapshot_measurer = SnapshotMeasurer(fuzzer, benchmark, trial_id,\n+ snapshot_logger)\n+ new_covered_regions = snapshot_measurer.get_current_covered_regions()\n+ covered_regions[key] = covered_regions[key].union(new_covered_regions)\n+ q.put(covered_regions)\n+ logger.debug('Done measuring covered region: fuzzer: %s, benchmark: %s.',\n+ fuzzer, benchmark)\n+\n+\n+def measure_main(experiment_config):\n+ \"\"\"Do the continuously measuring and the final measuring.\"\"\"\n+ initialize_logs()\n+ logger.info('Start measuring.')\n+\n+ # Start the measure loop first.\n+ experiment = experiment_config['experiment']\n+ max_total_time = experiment_config['max_total_time']\n+ measure_loop(experiment, max_total_time)\n+\n+ # Do the final measuring and store the coverage data.\n+ store_coverage_data(experiment_config)\n+ logger.info('Finished measuring.')\n+\n+\ndef measure_loop(experiment: str, max_total_time: int):\n\"\"\"Continuously measure trials for |experiment|.\"\"\"\n- logs.initialize(default_extras={\n- 'component': 'dispatcher',\n- 'subcomponent': 'measurer',\n- })\n- logs.info('Start measure_loop.')\n+ logger.info('Start measure_loop.')\nwith multiprocessing.Pool() as pool, multiprocessing.Manager() as manager:\nset_up_coverage_binaries(pool, experiment)\n@@ -98,7 +212,7 @@ def measure_loop(experiment: str, max_total_time: int):\ntime.sleep(FAIL_WAIT_SECONDS)\n- logger.info('Finished measuring.')\n+ logger.info('Finished measure loop.')\ndef measure_all_trials(experiment: str, max_total_time: int, pool, q) -> bool: # pylint: disable=invalid-name\n@@ -367,33 +481,47 @@ class SnapshotMeasurer: # pylint: disable=too-many-instance-attributes\nself.UNIT_BLACKLIST[self.benchmark] = (\nself.UNIT_BLACKLIST[self.benchmark].union(set(crashing_units)))\n+ def get_current_covered_regions(self):\n+ \"\"\"Get the covered regions for the current trial.\"\"\"\n+ covered_regions = set()\n+ try:\n+ coverage_info = get_coverage_infomation(self.cov_summary_file)\n+ functions_data = coverage_info['data'][0]['functions']\n+ # The fourth number in the region-list indicates if the region\n+ # is hit.\n+ hit_index = 4\n+ # The last number in the region-list indicates what type of the\n+ # region it is; 'code_region' is used to obtain various code\n+ # coverage statistic and is represented by number 0.\n+ type_index = -1\n+ for function_data in functions_data:\n+ for region in function_data['regions']:\n+ if region[hit_index] != 0 and region[type_index] == 0:\n+ covered_regions.add(tuple(region[:hit_index]))\n+ except Exception: # pylint: disable=broad-except\n+ self.logger.error(\n+ 'Coverage summary json file defective or missing.')\n+ return covered_regions\n+\ndef get_current_coverage(self) -> int:\n\"\"\"Get the current number of lines covered.\"\"\"\nif not os.path.exists(self.cov_summary_file):\nself.logger.warning('No coverage summary json file found.')\nreturn 0\ntry:\n- with open(self.cov_summary_file) as summary:\n- # Using the last line to skip the warning in bloaty\n- coverage_info = json.loads(summary.readlines()[-1])\n+ coverage_info = get_coverage_infomation(self.cov_summary_file)\ncoverage_data = coverage_info[\"data\"][0]\nsummary_data = coverage_data[\"totals\"]\nregions_coverage_data = summary_data[\"regions\"]\nregions_covered = regions_coverage_data[\"covered\"]\nreturn regions_covered\nexcept Exception: # pylint: disable=broad-except\n- self.logger.error('Coverage summary json file defective.')\n+ self.logger.error(\n+ 'Coverage summary json file defective or missing.')\nreturn 0\ndef generate_profdata(self, cycle: int):\n\"\"\"Generate .profdata file from .profraw file.\"\"\"\n- if not os.path.exists(self.profraw_file):\n- self.logger.error('No profraw file found for cycle: %d.', cycle)\n- return\n- if not os.path.getsize(self.profraw_file):\n- self.logger.error('Empty profraw file found for cycle: %d.', cycle)\n- return\n-\nif os.path.isfile(self.profdata_file):\n# If coverage profdata exists, then merge it with\n# existing available data.\n@@ -410,17 +538,9 @@ class SnapshotMeasurer: # pylint: disable=too-many-instance-attributes\ndef generate_summary(self, cycle: int):\n\"\"\"Transform the .profdata file into json form.\"\"\"\n- if not os.path.exists(self.profdata_file):\n- self.logger.error('No profdata file found for cycle: %d.', cycle)\n- return\n- if not os.path.getsize(self.profdata_file):\n- self.logger.error('Empty profdata file found for cycle: %d.', cycle)\n- return\n-\ncoverage_binary = get_coverage_binary(self.benchmark)\ncommand = [\n- 'llvm-cov', 'export', '-format=text', '-summary-only',\n- coverage_binary,\n+ 'llvm-cov', 'export', '-format=text', coverage_binary,\n'-instr-profile=%s' % self.profdata_file\n]\nwith open(self.cov_summary_file, 'w') as output_file:\n@@ -432,6 +552,25 @@ class SnapshotMeasurer: # pylint: disable=too-many-instance-attributes\n'Coverage summary json file generation failed for \\\ncycle: %d.', cycle)\n+ def generate_coverage_information(self, cycle: int):\n+ \"\"\"Generate the .profdata file and then transform it into\n+ json summary.\"\"\"\n+ if not os.path.exists(self.profraw_file):\n+ self.logger.error('No profraw file found for cycle: %d.', cycle)\n+ return\n+ if not os.path.getsize(self.profraw_file):\n+ self.logger.error('Empty profraw file found for cycle: %d.', cycle)\n+ return\n+ self.generate_profdata(cycle)\n+\n+ if not os.path.exists(self.profdata_file):\n+ self.logger.error('No profdata file found for cycle: %d.', cycle)\n+ return\n+ if not os.path.getsize(self.profdata_file):\n+ self.logger.error('Empty profdata file found for cycle: %d.', cycle)\n+ return\n+ self.generate_summary(cycle)\n+\ndef is_cycle_unchanged(self, cycle: int) -> bool:\n\"\"\"Returns True if |cycle| is unchanged according to the\nunchanged-cycles file. This file is written to by the trial's runner.\"\"\"\n@@ -592,8 +731,7 @@ def measure_snapshot_coverage(fuzzer: str, benchmark: str, trial_num: int,\nsnapshot_measurer.run_cov_new_units()\n# Generate profdata and transform it into json form.\n- snapshot_measurer.generate_profdata(cycle)\n- snapshot_measurer.generate_summary(cycle)\n+ snapshot_measurer.generate_coverage_information(cycle)\n# Get the coverage of the new corpus units.\nregions_covered = snapshot_measurer.get_current_coverage()\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_data/cov_summary.json", "new_path": "experiment/test_data/cov_summary.json", "diff": "-{\"version\":\"2.0.0\",\"type\":\"llvm.coverage.json.export\",\"data\":[{\"files\":[{\"filename\":\"/home/test/fuzzing/tutorial/libFuzzer/fuzz_me.cc\",\"summary\":{\"lines\":{\"count\":11,\"covered\":11,\"percent\":100},\"functions\":{\"count\":2,\"covered\":2,\"percent\":100},\"instantiations\":{\"count\":2,\"covered\":2,\"percent\":100},\"regions\":{\"count\":10,\"covered\":7,\"notcovered\":3,\"percent\":70}}}],\"totals\":{\"lines\":{\"count\":11,\"covered\":11,\"percent\":100},\"functions\":{\"count\":2,\"covered\":2,\"percent\":100},\"instantiations\":{\"count\":2,\"covered\":2,\"percent\":100},\"regions\":{\"count\":10,\"covered\":7,\"notcovered\":3,\"percent\":70}}}]}\n\\ No newline at end of file\n+{\"version\":\"2.0.0\",\"type\":\"llvm.coverage.json.export\",\"data\":[{\"files\":[{\"filename\":\"/home/test/fuzz_no_fuzzer.cc\",\"segments\":[[1,16,20,1,1],[1,17,20,1,1],[1,20,20,1,0],[1,24,2,1,1],[1,27,20,1,0],[1,28,0,0,0],[2,37,2,1,1],[3,24,22,1,1],[3,30,2,1,0],[3,32,20,1,1],[3,35,2,1,0],[3,36,20,1,0],[3,37,20,1,1],[3,39,20,1,1],[3,42,20,1,0],[3,48,2,1,0],[5,3,0,1,1],[6,2,0,0,0],[7,12,1,1,1],[11,3,0,1,1],[12,2,0,0,0]],\"expansions\":[{\"source_region\":[3,39,3,42,10,0,1,1],\"target_regions\":[[2,37,6,2,1,0,0,0],[3,24,3,30,11,0,0,0],[3,32,3,35,10,0,0,0],[3,36,3,37,10,0,0,3],[3,37,3,48,10,0,0,0],[3,39,3,42,10,0,1,1],[5,3,6,2,0,0,0,0],[1,16,1,28,10,1,0,0],[1,17,1,20,10,1,0,0],[1,24,1,27,1,1,0,0]],\"filenames\":[\"/home/test/fuzz_no_fuzzer.cc\",\"/home/test/fuzz_no_fuzzer.cc\"]},{\"source_region\":[3,39,3,42,10,0,1,1],\"target_regions\":[[2,37,6,2,1,0,0,0],[3,24,3,30,11,0,0,0],[3,32,3,35,10,0,0,0],[3,36,3,37,10,0,0,3],[3,37,3,48,10,0,0,0],[3,39,3,42,10,0,1,1],[5,3,6,2,0,0,0,0],[1,16,1,28,10,1,0,0],[1,17,1,20,10,1,0,0],[1,24,1,27,1,1,0,0]],\"filenames\":[\"/home/test/fuzz_no_fuzzer.cc\",\"/home/test/fuzz_no_fuzzer.cc\"]}],\"summary\":{\"lines\":{\"count\":11,\"covered\":9,\"percent\":81},\"functions\":{\"count\":2,\"covered\":2,\"percent\":100},\"instantiations\":{\"count\":3,\"covered\":3,\"percent\":100},\"regions\":{\"count\":10,\"covered\":8,\"notcovered\":2,\"percent\":80}}}],\"functions\":[{\"name\":\"main\",\"count\":1,\"regions\":[[7,12,12,2,1,0,0,0],[11,3,12,2,0,0,0,0]],\"filenames\":[\"/home/test/fuzz_no_fuzzer.cc\"]},{\"name\":\"_Z3fooIiEvT_\",\"count\":1,\"regions\":[[2,37,6,2,1,0,0,0],[3,24,3,30,11,0,0,0],[3,32,3,35,10,0,0,0],[3,36,3,37,10,0,0,3],[3,37,3,48,10,0,0,0],[3,39,3,42,10,0,1,1],[5,3,6,2,0,0,0,0],[1,16,1,28,10,1,0,0],[1,17,1,20,10,1,0,0],[1,24,1,27,1,1,0,0]],\"filenames\":[\"/home/test/fuzz_no_fuzzer.cc\",\"/home/test/fuzz_no_fuzzer.cc\"]},{\"name\":\"_Z3fooIfEvT_\",\"count\":1,\"regions\":[[2,37,6,2,1,0,0,0],[3,24,3,30,11,0,0,0],[3,32,3,35,10,0,0,0],[3,36,3,37,10,0,0,3],[3,37,3,48,10,0,0,0],[3,39,3,42,10,0,1,1],[5,3,6,2,0,0,0,0],[1,16,1,28,10,1,0,0],[1,17,1,20,10,1,0,0],[1,24,1,27,1,1,0,0]],\"filenames\":[\"/home/test/fuzz_no_fuzzer.cc\",\"/home/test/fuzz_no_fuzzer.cc\"]}],\"totals\":{\"lines\":{\"count\":11,\"covered\":9,\"percent\":81},\"functions\":{\"count\":2,\"covered\":2,\"percent\":100},\"instantiations\":{\"count\":3,\"covered\":3,\"percent\":100},\"regions\":{\"count\":10,\"covered\":8,\"notcovered\":2,\"percent\":80}}}]}\n\\ No newline at end of file\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_measurer.py", "new_path": "experiment/test_measurer.py", "diff": "@@ -56,6 +56,17 @@ def db_experiment(experiment_config, db):\nyield\n+def test_get_current_covered_regions(fs, experiment):\n+ \"\"\"Tests that get_current_coverage reads the correct data from json file.\"\"\"\n+ snapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\n+ SNAPSHOT_LOGGER)\n+ json_cov_summary_file = get_test_data_path('cov_summary.json')\n+ fs.add_real_file(json_cov_summary_file, read_only=False)\n+ snapshot_measurer.cov_summary_file = json_cov_summary_file\n+ covered_regions = snapshot_measurer.get_current_covered_regions()\n+ assert len(covered_regions) == 8\n+\n+\ndef test_get_current_coverage(fs, experiment):\n\"\"\"Tests that get_current_coverage reads the correct data from json file.\"\"\"\nsnapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\n@@ -64,7 +75,7 @@ def test_get_current_coverage(fs, experiment):\nfs.add_real_file(json_cov_summary_file, read_only=False)\nsnapshot_measurer.cov_summary_file = json_cov_summary_file\ncovered_regions = snapshot_measurer.get_current_coverage()\n- assert covered_regions == 7\n+ assert covered_regions == 8\ndef test_get_current_coverage_error(fs, experiment):\n@@ -150,7 +161,7 @@ def test_generate_summary(mocked_get_coverage_binary, mocked_execute,\nsnapshot_measurer.generate_summary(CYCLE)\nexpected = [\n- 'llvm-cov', 'export', '-format=text', '-summary-only',\n+ 'llvm-cov', 'export', '-format=text',\n'/work/coverage-binaries/benchmark-a/fuzz-target',\n'-instr-profile=/reports/data.profdata'\n]\n" } ]

[ { "change_type": "MODIFY", "old_path": "docker/gcb/coverage.yaml", "new_path": "docker/gcb/coverage.yaml", "diff": "@@ -101,7 +101,7 @@ steps:\n'${_REPO}/builders/coverage/${_BENCHMARK}',\n'--build-arg',\n- 'parent_image=${_REPO}/builders/coverage/${_BENCHMARK}-intermediate',\n+ 'parent_image=gcr.io/fuzzbench/builders/coverage/${_BENCHMARK}-intermediate',\n'--build-arg',\n'fuzzer=coverage',\n" }, { "change_type": "MODIFY", "old_path": "docker/gcb/fuzzer.yaml", "new_path": "docker/gcb/fuzzer.yaml", "diff": "@@ -102,7 +102,7 @@ steps:\n'${_REPO}/builders/${_FUZZER}/${_BENCHMARK}',\n'--build-arg',\n- 'parent_image=${_REPO}/builders/${_FUZZER}/${_BENCHMARK}-intermediate',\n+ 'parent_image=gcr.io/fuzzbench/builders/${_FUZZER}/${_BENCHMARK}-intermediate',\n'--build-arg',\n'fuzzer=${_FUZZER}',\n" } ]

[ { "change_type": "MODIFY", "old_path": "compose/fuzzbench.yaml", "new_path": "compose/fuzzbench.yaml", "diff": "@@ -14,7 +14,8 @@ services:\nimage: fuzzbench\nenvironment:\nRQ_REDIS_URL: redis://queue-server\n- command: rq worker --burst\n+ PYTHONPATH: .\n+ command: python3 fuzzbench/worker.py\nvolumes:\n# Allow access to the host's docker daemon.\n- /var/run/docker.sock:/var/run/docker.sock\n" }, { "change_type": "MODIFY", "old_path": "docker/image_types.yaml", "new_path": "docker/image_types.yaml", "diff": "tag: 'base-image'\ncontext: 'docker/base-image'\n-'base-builder':\n- tag: 'base-builder'\n- context: 'docker/base-builder'\n- depends_on:\n- - 'base-image'\n-\n'base-runner':\ntag: 'base-runner'\ncontext: 'docker/base-runner'\ncontext: 'fuzzers/coverage'\ndockerfile: 'fuzzers/coverage/builder.Dockerfile'\ndepends_on:\n- - 'base-builder'\n+ - 'base-image'\n'coverage-{benchmark}-builder':\ntag: 'builders/coverage/{benchmark}'\ncontext: 'fuzzers/{fuzzer}'\ndockerfile: 'fuzzers/{fuzzer}/builder.Dockerfile'\ndepends_on:\n- - 'base-builder'\n+ - 'base-image'\n'{fuzzer}-{benchmark}-builder':\ntag: 'builders/{fuzzer}/{benchmark}'\n" }, { "change_type": "MODIFY", "old_path": "experiment/build/test_docker_images.py", "new_path": "experiment/build/test_docker_images.py", "diff": "@@ -23,7 +23,7 @@ def test_images_to_build_list():\nbenchmarks = ['libxml', 'libpng']\nall_images = docker_images.get_images_to_build(fuzzers, benchmarks)\nassert set(all_images.keys()) == set([\n- 'base-image', 'base-builder', 'base-runner', 'coverage-builder',\n+ 'base-image', 'base-runner', 'coverage-builder',\n'coverage-libxml-builder', 'coverage-libpng-builder', 'afl-builder',\n'afl-libxml-builder', 'afl-libxml-intermediate-runner',\n'afl-libxml-runner', 'afl-libpng-builder',\n" }, { "change_type": "MODIFY", "old_path": "fuzzbench/run_experiment.py", "new_path": "fuzzbench/run_experiment.py", "diff": "@@ -24,19 +24,14 @@ from fuzzbench import jobs\ndef run_experiment():\n\"\"\"Main experiment logic.\"\"\"\nprint('Initializing the job queue.')\n- queue = rq.Queue()\n+ # Create the queue for scheduling build jobs and run jobs.\n+ queue = rq.Queue('build_n_run_queue')\njobs_list = []\njobs_list.append(\nqueue.enqueue(jobs.build_image,\n'base-image',\njob_timeout=600,\njob_id='base-image'))\n- jobs_list.append(\n- queue.enqueue(jobs.build_image,\n- 'base-builder',\n- job_timeout=600,\n- job_id='base-builder',\n- depends_on='base-image'))\njobs_list.append(\nqueue.enqueue(jobs.build_image,\n'base-runner',\n" }, { "change_type": "MODIFY", "old_path": "fuzzbench/test_e2e_run.py", "new_path": "fuzzbench/test_e2e_run.py", "diff": "@@ -39,14 +39,13 @@ class TestEndToEndRunResults:\n\"\"\"Tests that jobs dependency preserves during working.\"\"\"\njobs = {\nname: Job.fetch(name, connection=redis_connection)\n- for name in ['base-image', 'base-builder', 'base-runner']\n+ for name in ['base-image', 'base-runner']\n}\n- assert jobs['base-image'].ended_at <= jobs['base-builder'].started_at\nassert jobs['base-image'].ended_at <= jobs['base-runner'].started_at\ndef test_all_jobs_finished_successfully(self, redis_connection): # pylint: disable=redefined-outer-name\n\"\"\"Tests all jobs finished successully.\"\"\"\n- jobs = Job.fetch_many(['base-image', 'base-builder', 'base-runner'],\n+ jobs = Job.fetch_many(['base-image', 'base-runner'],\nconnection=redis_connection)\nfor job in jobs:\nassert job.get_status() == 'finished'\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzbench/worker.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Self-defined worker module.\"\"\"\n+import time\n+\n+import redis\n+import rq\n+\n+\n+def main():\n+ \"\"\"Sets up Redis connection and starts the worker.\"\"\"\n+ redis_connection = redis.Redis(host=\"queue-server\")\n+ with rq.Connection(redis_connection):\n+ queue = rq.Queue('build_n_run_queue')\n+ worker = rq.Worker([queue])\n+\n+ while queue.count + queue.deferred_job_registry.count > 0:\n+ worker.work(burst=True)\n+ time.sleep(5)\n+\n+\n+if __name__ == '__main__':\n+ main()\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/build_and_test_run_fuzzer_benchmarks.py", "new_path": ".github/workflows/build_and_test_run_fuzzer_benchmarks.py", "diff": "@@ -77,6 +77,9 @@ def delete_docker_images():\nimage_ids = result.stdout.splitlines()\nsubprocess.run(['docker', 'rmi', '-f'] + image_ids, check=False)\n+ # Needed for BUILDKIT to clear build cache & avoid insufficient disk space.\n+ subprocess.run(['docker', 'builder', 'prune', '-f'], check=False)\n+\ndef make_builds(benchmarks, fuzzer):\n\"\"\"Use make to build each target in |build_targets|.\"\"\"\n" }, { "change_type": "MODIFY", "old_path": "Makefile", "new_path": "Makefile", "diff": "@@ -42,7 +42,6 @@ run-end-to-end-test:\nstop-end-to-end-test:\ndocker-compose down\n-include docker/build.mk\ninclude docker/generated.mk\nSHELL := /bin/bash\n@@ -55,7 +54,7 @@ ${VENV_ACTIVATE}: requirements.txt\ninstall-dependencies: ${VENV_ACTIVATE}\ndocker/generated.mk: docker/generate_makefile.py $(wildcard fuzzers/*/variants.yaml) ${VENV_ACTIVATE}\n- source ${VENV_ACTIVATE} && python3 $< > $@\n+ source ${VENV_ACTIVATE} && PYTHONPATH=. python3 $< > $@\npresubmit: install-dependencies\nsource ${VENV_ACTIVATE} && python3 presubmit.py\n" }, { "change_type": "DELETE", "old_path": "docker/build.mk", "new_path": null, "diff": "-# Copyright 2020 Google LLC\n-#\n-# Licensed under the Apache License, Version 2.0 (the \"License\");\n-# you may not use this file except in compliance with the License.\n-# You may obtain a copy of the License at\n-#\n-# http://www.apache.org/licenses/LICENSE-2.0\n-#\n-# Unless required by applicable law or agreed to in writing, software\n-# distributed under the License is distributed on an \"AS IS\" BASIS,\n-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n-# See the License for the specific language governing permissions and\n-# limitations under the License.\n-BENCHMARKS := $(notdir $(shell find benchmarks -type f -name benchmark.yaml | xargs dirname))\n-\n-BASE_TAG ?= gcr.io/fuzzbench\n-\n-# If we're running on a CI service, cache-from a remote image. Otherwise just\n-# use the local cache.\n-cache_from = $(if ${RUNNING_ON_CI},--cache-from $(1),)\n-\n-# For base-* images (and those that depend on it), use a remote cache by\n-# default, unless the developer sets DISABLE_REMOTE_CACHE_FOR_BASE.\n-cache_from_base = $(if ${DISABLE_REMOTE_CACHE_FOR_BASE},,--cache-from $(1))\n-\n-base-image:\n- docker build \\\n- --tag $(BASE_TAG)/base-image \\\n- $(call cache_from_base,${BASE_TAG}/base-image) \\\n- docker/base-image\n-\n-pull-base-image:\n- docker pull $(BASE_TAG)/base-image\n-\n-dispatcher-image: base-image\n- docker build \\\n- --tag $(BASE_TAG)/dispatcher-image \\\n- $(call cache_from,${BASE_TAG}/dispatcher-image) \\\n- docker/dispatcher-image\n-\n-define benchmark_template\n-$(1)-fuzz-target := $(shell cat benchmarks/$(1)/benchmark.yaml | \\\n- grep fuzz_target | cut -d ':' -f2 | tr -d ' ')\n-\n-# TODO: It would be better to call this benchmark builder. But that would be\n-# confusing because this doesn't involve benchmark-builder/Dockerfile. Rename\n-# that and then rename this.\n-.$(1)-project-builder:\n- docker build \\\n- --tag $(BASE_TAG)/builders/benchmark/$(1) \\\n- --file benchmarks/$(1)/Dockerfile \\\n- $(call cache_from,${BASE_TAG}/builders/benchmarks/$(1)) \\\n- benchmarks/$(1)\n-\n-endef\n-# Instantiate the above template with all OSS-Fuzz benchmarks.\n-$(foreach benchmark,$(BENCHMARKS), \\\n- $(eval $(call benchmark_template,$(benchmark))))\n" }, { "change_type": "MODIFY", "old_path": "docker/generate_makefile.py", "new_path": "docker/generate_makefile.py", "diff": "import os\n-BASE_TAG = 'gcr.io/fuzzbench'\n-BENCHMARKS_DIR = os.path.join(os.path.dirname(__file__), os.pardir,\n- 'benchmarks')\n-FUZZERS_DIR = os.path.join(os.path.dirname(__file__), os.pardir, 'fuzzers')\n-\n-BOILERPLATE = \"\"\"\n-cache_from = $(if ${RUNNING_ON_CI},--cache-from {fuzzer},)\n-\"\"\"\n-\n-FUZZER_TEMPLATE = \"\"\"\n-.{fuzzer}-builder: base-image\n- docker build \\\\\n- --tag {base_tag}/builders/{fuzzer} \\\\\n- --file fuzzers/{fuzzer}/builder.Dockerfile \\\\\n- $(call cache_from,{base_tag}/builders/{fuzzer}) \\\\\n- fuzzers/{fuzzer}\n-\n-.pull-{fuzzer}-builder: pull-base-image\n- docker pull {base_tag}/builders/{fuzzer}\n-\"\"\"\n-\n-FUZZER_BENCHMARK_RUN_TARGETS_TEMPLATE = \"\"\"\n-build-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-runner\n-\n-pull-{fuzzer}-{benchmark}: .pull-{fuzzer}-{benchmark}-runner\n-\n-run-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-runner\n- docker run \\\\\n- --cpus=1 \\\\\n- --cap-add SYS_NICE \\\\\n- --cap-add SYS_PTRACE \\\\\n- -e FUZZ_OUTSIDE_EXPERIMENT=1 \\\\\n- -e FORCE_LOCAL=1 \\\\\n- -e TRIAL_ID=1 \\\\\n- -e FUZZER={fuzzer} \\\\\n- -e BENCHMARK={benchmark} \\\\\n- -e FUZZ_TARGET=$({benchmark}-fuzz-target) \\\\\n- -it {base_tag}/runners/{fuzzer}/{benchmark}\n-\n-test-run-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-runner\n- docker run \\\\\n- --cap-add SYS_NICE \\\\\n- --cap-add SYS_PTRACE \\\\\n- -e FUZZ_OUTSIDE_EXPERIMENT=1 \\\\\n- -e FORCE_LOCAL=1 \\\\\n- -e TRIAL_ID=1 \\\\\n- -e FUZZER={fuzzer} \\\\\n- -e BENCHMARK={benchmark} \\\\\n- -e FUZZ_TARGET=$({benchmark}-fuzz-target) \\\\\n- -e MAX_TOTAL_TIME=20 \\\\\n- -e SNAPSHOT_PERIOD=10 \\\\\n- {base_tag}/runners/{fuzzer}/{benchmark}\n-\n-debug-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-runner\n- docker run \\\\\n- --cpus=1 \\\\\n- --cap-add SYS_NICE \\\\\n- --cap-add SYS_PTRACE \\\\\n- -e FUZZ_OUTSIDE_EXPERIMENT=1 \\\\\n- -e FORCE_LOCAL=1 \\\\\n- -e TRIAL_ID=1 \\\\\n- -e FUZZER={fuzzer} \\\\\n- -e BENCHMARK={benchmark} \\\\\n- -e FUZZ_TARGET=$({benchmark}-fuzz-target) \\\\\n- --entrypoint \"/bin/bash\" \\\\\n- -it {base_tag}/runners/{fuzzer}/{benchmark}\n-\"\"\"\n-\n-FUZZER_BENCHMARK_TEMPLATE = \"\"\"\n-.{fuzzer}-{benchmark}-builder-intermediate: .{benchmark}-project-builder\n- docker build \\\\\n- --tag {base_tag}/builders/{fuzzer}/{benchmark}-intermediate \\\\\n- --file=fuzzers/{fuzzer}/builder.Dockerfile \\\\\n- --build-arg parent_image=gcr.io/fuzzbench/builders/benchmark/{benchmark} \\\\\n- $(call cache_from,{base_tag}/builders/{fuzzer}/{benchmark}-intermediate) \\\\\n- fuzzers/{fuzzer}\n-\n-.pull-{fuzzer}-{benchmark}-builder-intermediate:\n- docker pull {base_tag}/builders/{fuzzer}/{benchmark}-intermediate\n-\n-.{fuzzer}-{benchmark}-builder: .{fuzzer}-{benchmark}-builder-intermediate\n- docker build \\\\\n- --tag {base_tag}/builders/{fuzzer}/{benchmark} \\\\\n- --file=docker/benchmark-builder/Dockerfile \\\\\n- --build-arg parent_image={base_tag}/builders/{fuzzer}/{benchmark}-intermediate \\\\\n- --build-arg fuzzer={fuzzer} \\\\\n- --build-arg benchmark={benchmark} \\\\\n- $(call cache_from,{base_tag}/builders/{fuzzer}/{benchmark}) \\\\\n- .\n-\n-.pull-{fuzzer}-{benchmark}-builder: .pull-{fuzzer}-{benchmark}-builder-intermediate\n- docker pull {base_tag}/builders/{fuzzer}/{benchmark}\n-\n-ifneq ({fuzzer}, coverage)\n-\n-.{fuzzer}-{benchmark}-intermediate-runner: .{fuzzer}-{benchmark}-builder\n- docker build \\\\\n- --tag {base_tag}/runners/{fuzzer}/{benchmark}-intermediate \\\\\n- --file fuzzers/{fuzzer}/runner.Dockerfile \\\\\n- $(call cache_from,{base_tag}/runners/{fuzzer}/{benchmark}-intermediate) \\\\\n- fuzzers/{fuzzer}\n-\n-.pull-{fuzzer}-{benchmark}-intermediate-runner: pull-base-image\n- docker pull {base_tag}/runners/{fuzzer}/{benchmark}-intermediate\n-\n-.{fuzzer}-{benchmark}-runner: .{fuzzer}-{benchmark}-intermediate-runner\n- docker build \\\\\n- --tag {base_tag}/runners/{fuzzer}/{benchmark} \\\\\n- --build-arg fuzzer={fuzzer} \\\\\n- --build-arg benchmark={benchmark} \\\\\n- $(call cache_from,{base_tag}/runners/{fuzzer}/{benchmark}) \\\\\n- --file docker/benchmark-runner/Dockerfile \\\\\n- .\n-\n-.pull-{fuzzer}-{benchmark}-runner: .pull-{fuzzer}-{benchmark}-intermediate-runner\n- docker pull {base_tag}/runners/{fuzzer}/{benchmark}\n-\n-\"\"\" + FUZZER_BENCHMARK_RUN_TARGETS_TEMPLATE + \"\"\"\n-\n-else\n-\n-build-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-builder\n-pull-{fuzzer}-{benchmark}: .pull-{fuzzer}-{benchmark}-builder\n-\n-endif\n-\"\"\"\n-\n-\n-def generate_fuzzer(fuzzer, benchmarks):\n- \"\"\"Output make rules for a single fuzzer.\"\"\"\n- # Generate build rules for the fuzzer itself.\n- print(FUZZER_TEMPLATE.format(fuzzer=fuzzer, base_tag=BASE_TAG))\n-\n- # Generate rules for fuzzer-benchmark pairs.\n+from common import yaml_utils\n+from common import benchmark_utils\n+from common import fuzzer_utils\n+from experiment.build import docker_images\n+\n+BASE_TAG = \"gcr.io/fuzzbench\"\n+BENCHMARK_DIR = benchmark_utils.BENCHMARKS_DIR\n+\n+\n+def _print_benchmark_fuzz_target(benchmarks):\n+ \"\"\"Prints benchmark variables from benchmark.yaml files.\"\"\"\nfor benchmark in benchmarks:\n- print(\n- FUZZER_BENCHMARK_TEMPLATE.format(fuzzer=fuzzer,\n+ benchmark_vars = yaml_utils.read(\n+ os.path.join(BENCHMARK_DIR, benchmark, 'benchmark.yaml'))\n+ print(benchmark + '-fuzz-target=' + benchmark_vars['fuzz_target'])\n+ print()\n+\n+\n+def _print_makefile_run_template(image):\n+ fuzzer, benchmark = image['tag'].split('/')[1:]\n+\n+ for run_type in ('run', 'debug', 'test-run'):\n+ print(('{run_type}-{fuzzer}-{benchmark}: ' +\n+ '.{fuzzer}-{benchmark}-runner').format(run_type=run_type,\nbenchmark=benchmark,\n- base_tag=BASE_TAG))\n+ fuzzer=fuzzer))\n+\n+ print('\\\n+\\tdocker run \\\\\\n\\\n+\\t--cpus=1 \\\\\\n\\\n+\\t--cap-add SYS_NICE \\\\\\n\\\n+\\t--cap-add SYS_PTRACE \\\\\\n\\\n+\\t-e FUZZ_OUTSIDE_EXPERIMENT=1 \\\\\\n\\\n+\\t-e FORCE_LOCAL=1 \\\\\\n\\\n+\\t-e TRIAL_ID=1 \\\\\\n\\\n+\\t-e FUZZER={fuzzer} \\\\\\n\\\n+\\t-e BENCHMARK={benchmark} \\\\\\n\\\n+\\t-e FUZZ_TARGET=$({benchmark}-fuzz-target) \\\\\\\n+'.format(fuzzer=fuzzer, benchmark=benchmark))\n+\n+ if run_type == 'test-run':\n+ print('\\t-e MAX_TOTAL_TIME=20 \\\\\\n\\t-e SNAPSHOT_PERIOD=10 \\\\')\n+ if run_type == 'debug':\n+ print('\\t-entrypoint \"/bin/bash\" \\\\\\n\\t-it ', end='')\n+ else:\n+ print('\\t', end='')\n+\n+ print(os.path.join(BASE_TAG, image['tag']))\n+ print()\n+\n+\n+# TODO(tanq16): Add unit test.\n+def _print_rules_for_image(name, image):\n+ \"\"\"Print makefile section for given image to stdout.\"\"\"\n+ if not ('base' in name or 'dispatcher' in name):\n+ print('.', end='')\n+ print(name + ':', end='')\n+ if 'depends_on' in image:\n+ for dep in image['depends_on']:\n+ if 'base' in dep:\n+ print(' ' + dep, end='')\n+ else:\n+ print(' .' + dep, end='')\n+ print()\n+ print('\\tdocker build \\\\')\n+ print('\\t--tag ' + os.path.join(BASE_TAG, image['tag']) + ' \\\\')\n+ print('\\t--build-arg BUILDKIT_INLINE_CACHE=1 \\\\')\n+ print('\\t--cache-from ' + os.path.join(BASE_TAG, image['tag']) + ' \\\\')\n+ if 'build_arg' in image:\n+ for arg in image['build_arg']:\n+ print('\\t--build-arg ' + arg + ' \\\\')\n+ if 'dockerfile' in image:\n+ print('\\t--file ' + image['dockerfile'] + ' \\\\')\n+ print('\\t' + image['context'])\n+ print()\n+\n+ # Print run, debug, test-run rules if image is a runner.\n+ if 'runner' in name and not ('intermediate' in name or 'base' in name):\n+ _print_makefile_run_template(image)\n- # Generate rules for building/pulling all target/benchmark pairs.\n- all_build_targets = ' '.join(\n- ['build-{0}-{1}'.format(fuzzer, benchmark) for benchmark in benchmarks])\n- all_pull_targets = ' '.join(\n- ['pull-{0}-{1}'.format(fuzzer, benchmark) for benchmark in benchmarks])\n+\n+def main():\n+ \"\"\"Generates Makefile with docker image build rules.\"\"\"\n+ fuzzers = fuzzer_utils.get_fuzzer_names()\n+ benchmarks = benchmark_utils.get_all_benchmarks()\n+ buildable_images = docker_images.get_images_to_build(fuzzers, benchmarks)\n+\n+ print('export DOCKER_BUILDKIT := 1')\n+\n+ # Print oss-fuzz benchmarks property variables.\n+ _print_benchmark_fuzz_target(benchmarks)\n+\n+ for name, image in buildable_images.items():\n+ _print_rules_for_image(name, image)\n+\n+ # Print build targets for all fuzzer-benchmark pairs (including coverage).\n+ fuzzers.append('coverage')\n+ for fuzzer in fuzzers:\n+ image_type = \"runner\"\n+ if 'coverage' in fuzzer:\n+ image_type = \"builder\"\n+ for benchmark in benchmarks:\n+ print(('build-{fuzzer}-{benchmark}: ' +\n+ '.{fuzzer}-{benchmark}-{image_type}\\n').format(\n+ fuzzer=fuzzer,\n+ benchmark=benchmark,\n+ image_type=image_type))\n+ print()\n+\n+ # Print fuzzer-all benchmarks build targets.\n+ for fuzzer in fuzzers:\n+ all_build_targets = ' '.join([\n+ 'build-{0}-{1}'.format(fuzzer, benchmark)\n+ for benchmark in benchmarks\n+ ])\nprint('build-{fuzzer}-all: {all_targets}'.format(\nfuzzer=fuzzer, all_targets=all_build_targets))\n- print('pull-{fuzzer}-all: {all_targets}'.format(\n- fuzzer=fuzzer, all_targets=all_pull_targets))\n-\n-def main():\n- \"\"\"Main entry point.\"\"\"\n- # Output boilerplate used by other templates and generated rules.\n- print(BOILERPLATE)\n-\n- # Compute the list of benchmarks.\n- benchmarks = []\n- for benchmark in os.listdir(BENCHMARKS_DIR):\n- benchmark_path = os.path.join(BENCHMARKS_DIR, benchmark)\n- if not os.path.isdir(benchmark_path):\n- continue\n- if os.path.exists(os.path.join(benchmark_path, 'benchmark.yaml')):\n- benchmarks.append(benchmark)\n-\n- # Generate the build rules for fuzzer/benchmark pairs.\n- fuzzers = []\n- for fuzzer in os.listdir(FUZZERS_DIR):\n- # Skip non-directory files. These do not represent fuzzers.\n- fuzzer_dir = os.path.join(FUZZERS_DIR, fuzzer)\n- if not os.path.isdir(fuzzer_dir):\n- continue\n-\n- generate_fuzzer(fuzzer, benchmarks)\n- fuzzers.append(fuzzer)\n-\n- # Generate rules to build all known targets.\n+ # Print all targets build target.\nall_build_targets = ' '.join(\n['build-{0}-all'.format(name) for name in fuzzers])\n- all_pull_targets = ' '.join(\n- ['pull-{0}-all'.format(name) for name in fuzzers])\nprint('build-all: {all_targets}'.format(all_targets=all_build_targets))\n- print('pull-all: {all_targets}'.format(all_targets=all_pull_targets))\nif __name__ == '__main__':\n" }, { "change_type": "MODIFY", "old_path": "docker/image_types.yaml", "new_path": "docker/image_types.yaml", "diff": "tag: 'base-image'\ncontext: 'docker/base-image'\n-'coverage-builder':\n- tag: 'builders/coverage'\n+'dispatcher-image':\n+ tag: 'dispatcher-image'\n+ context: 'docker/dispatcher-image'\n+ depends_on:\n+ - 'base-image'\n+\n+'coverage-{benchmark}-builder-intermediate':\n+ tag: 'builders/coverage/{benchmark}-intermediate'\ncontext: 'fuzzers/coverage'\ndockerfile: 'fuzzers/coverage/builder.Dockerfile'\ndepends_on:\n- - 'base-image'\n+ - '{benchmark}-project-builder'\n+ build_arg:\n+ - 'parent_image=gcr.io/fuzzbench/builders/benchmark/{benchmark}'\n'coverage-{benchmark}-builder':\ntag: 'builders/coverage/{benchmark}'\ncontext: '.'\ndockerfile: 'docker/benchmark-builder/Dockerfile'\nbuild_arg:\n+ - 'parent_image=gcr.io/fuzzbench/builders/coverage/{benchmark}-intermediate'\n- 'fuzzer=coverage'\n- 'benchmark={benchmark}'\ndepends_on:\n- - 'coverage-builder'\n+ - 'coverage-{benchmark}-builder-intermediate'\n+\n+# TODO: It would be better to call this benchmark builder. But that would be\n+# confusing because this doesn't involve benchmark-builder/Dockerfile. Rename\n+# that and then rename this.\n+'{benchmark}-project-builder':\n+ tag: 'builders/benchmark/{benchmark}'\n+ context: 'benchmarks/{benchmark}'\n+ dockerfile: 'benchmarks/{benchmark}/Dockerfile'\n-'{fuzzer}-builder':\n- tag: 'builders/{fuzzer}'\n+'{fuzzer}-{benchmark}-builder-intermediate':\n+ tag: 'builders/{fuzzer}/{benchmark}-intermediate'\ncontext: 'fuzzers/{fuzzer}'\ndockerfile: 'fuzzers/{fuzzer}/builder.Dockerfile'\ndepends_on:\n- - 'base-image'\n+ - '{benchmark}-project-builder'\n+ build_arg:\n+ - 'parent_image=gcr.io/fuzzbench/builders/benchmark/{benchmark}'\n'{fuzzer}-{benchmark}-builder':\ntag: 'builders/{fuzzer}/{benchmark}'\ncontext: '.'\ndockerfile: 'docker/benchmark-builder/Dockerfile'\nbuild_arg:\n+ - 'parent_image=gcr.io/fuzzbench/builders/{fuzzer}/{benchmark}-intermediate'\n- 'fuzzer={fuzzer}'\n- 'benchmark={benchmark}'\ndepends_on:\n- - '{fuzzer}-builder'\n+ - '{fuzzer}-{benchmark}-builder-intermediate'\n'{fuzzer}-{benchmark}-intermediate-runner':\ntag: 'runners/{fuzzer}/{benchmark}-intermediate'\n" }, { "change_type": "UNKNOWN", "old_path": "docs/developing-fuzzbench/adding_a_new_benchmark.md", "new_path": "docs/developing-fuzzbench/adding_a_new_benchmark.md", "diff": "" }, { "change_type": "MODIFY", "old_path": "experiment/build/docker_images.py", "new_path": "experiment/build/docker_images.py", "diff": "# limitations under the License.\n\"\"\"Provides the set of buildable images and their dependencies.\"\"\"\n+import os\n+\nfrom common import yaml_utils\n+from common.utils import ROOT_DIR\ndef _substitute(template, fuzzer, benchmark):\n@@ -40,7 +43,9 @@ def _instantiate_image_obj(name_template, obj_template, fuzzer, benchmark):\ndef _get_image_type_templates():\n\"\"\"Loads the image types config that contains \"templates\" describing how to\nbuild them and their dependencies.\"\"\"\n- return yaml_utils.read('docker/image_types.yaml')\n+ yaml_file = os.path.join(ROOT_DIR, 'docker', 'image_types.yaml')\n+ all_templates = yaml_utils.read(yaml_file)\n+ return all_templates\ndef get_images_to_build(fuzzers, benchmarks):\n" }, { "change_type": "DELETE", "old_path": "experiment/build/generate_makefile.py", "new_path": null, "diff": "-# Copyright 2020 Google LLC\n-#\n-# Licensed under the Apache License, Version 2.0 (the \"License\");\n-# you may not use this file except in compliance with the License.\n-# You may obtain a copy of the License at\n-#\n-# http://www.apache.org/licenses/LICENSE-2.0\n-#\n-# Unless required by applicable law or agreed to in writing, software\n-# distributed under the License is distributed on an \"AS IS\" BASIS,\n-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n-# See the License for the specific language governing permissions and\n-# limitations under the License.\n-\"\"\"Generates Makefile containing docker image targets.\"\"\"\n-\n-import argparse\n-\n-from experiment.build import docker_images\n-\n-\n-# TODO(Tanq16): Add unit test for this.\n-def print_makefile(buildable_images, docker_registry):\n- \"\"\"Prints the generated makefile to stdout.\"\"\"\n- print('export DOCKER_BUILDKIT := 1')\n-\n- for name, image in buildable_images.items():\n- print(name + ':', end='')\n- if 'depends_on' in image:\n- for dep in image['depends_on']:\n- print(' ' + dep, end='')\n- print()\n- print('\\tdocker build \\\\')\n- print(' --tag ' + image['tag'] + ' \\\\')\n- print(' --cache-from ' + docker_registry + image['tag'] + ' \\\\')\n- if 'build_arg' in image:\n- for arg in image['build_arg']:\n- print(' --build-arg ' + arg + ' \\\\')\n- if 'dockerfile' in image:\n- print(' --file ' + image['dockerfile'] + ' \\\\')\n- print(' ' + image['context'])\n- print()\n-\n-\n-def main():\n- \"\"\"Generates Makefile with docker image build rules.\"\"\"\n- parser = argparse.ArgumentParser(\n- description='Makefile build rule generator.')\n- parser.add_argument('-r',\n- '--docker-registry',\n- default='gcr.io/fuzzbench/',\n- help='Docker registry to use as cache.')\n- args = parser.parse_args()\n-\n- # TODO(Tanq16): Create fuzzer/benchmark list dynamically.\n- fuzzers = ['afl', 'libfuzzer']\n- benchmarks = ['libxml', 'libpng']\n- buildable_images = docker_images.get_images_to_build(fuzzers, benchmarks)\n- print_makefile(buildable_images, args.docker_registry)\n-\n-\n-if __name__ == '__main__':\n- main()\n" }, { "change_type": "MODIFY", "old_path": "experiment/build/test_docker_images.py", "new_path": "experiment/build/test_docker_images.py", "diff": "@@ -23,14 +23,18 @@ def test_images_to_build_list():\nbenchmarks = ['libxml', 'libpng']\nall_images = docker_images.get_images_to_build(fuzzers, benchmarks)\nassert set(all_images.keys()) == set([\n- 'base-image', 'coverage-builder', 'coverage-libxml-builder',\n- 'coverage-libpng-builder', 'afl-builder', 'afl-libxml-builder',\n- 'afl-libxml-intermediate-runner', 'afl-libxml-runner',\n- 'afl-libpng-builder', 'afl-libpng-intermediate-runner',\n- 'afl-libpng-runner', 'libfuzzer-builder', 'libfuzzer-libxml-builder',\n+ 'base-image', 'dispatcher-image', 'libxml-project-builder',\n+ 'libpng-project-builder', 'afl-libxml-builder-intermediate',\n+ 'afl-libxml-intermediate-runner', 'afl-libxml-builder',\n+ 'coverage-libxml-builder', 'afl-libpng-builder',\n+ 'afl-libpng-intermediate-runner', 'afl-libpng-builder-intermediate',\n+ 'afl-libpng-runner', 'libfuzzer-libxml-builder-intermediate',\n+ 'libfuzzer-libxml-builder', 'libfuzzer-libpng-builder-intermediate',\n'libfuzzer-libxml-intermediate-runner', 'libfuzzer-libxml-runner',\n'libfuzzer-libpng-builder', 'libfuzzer-libpng-intermediate-runner',\n- 'libfuzzer-libpng-runner'\n+ 'libfuzzer-libpng-runner', 'coverage-libxml-builder-intermediate',\n+ 'coverage-libpng-builder', 'coverage-libxml-builder-intermediate',\n+ 'afl-libxml-runner', 'coverage-libpng-builder-intermediate'\n])\n" }, { "change_type": "MODIFY", "old_path": "src_analysis/test_change_utils.py", "new_path": "src_analysis/test_change_utils.py", "diff": "@@ -23,7 +23,7 @@ def test_get_changed_fuzzers_for_ci():\n\"\"\"Tests that get_changed_fuzzers_for_ci returns all fuzzers when a file\nthat affects all fuzzer build was changed.\"\"\"\nchanged_fuzzers = change_utils.get_changed_fuzzers_for_ci(\n- [os.path.join(utils.ROOT_DIR, 'docker', 'build.mk')])\n+ [os.path.join(utils.ROOT_DIR, 'docker', 'image_types.yaml')])\nassert changed_fuzzers == fuzzer_utils.get_fuzzer_names()\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/fuzzers.yml", "new_path": ".github/workflows/fuzzers.yml", "diff": "@@ -35,7 +35,9 @@ jobs:\n# temporary variants.\n- aflplusplus_havoc\n- libfuzzer_interceptors\n+ - libfuzzer_keepseed\n- entropic_interceptors\n+ - entropic_keepseed\nbenchmark_type:\n- oss-fuzz\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_keepseed/builder.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+ARG parent_image\n+FROM $parent_image\n+\n+COPY patch.diff /\n+\n+RUN git clone https://github.com/llvm/llvm-project.git /llvm-project && \\\n+ cd /llvm-project && \\\n+ git checkout b52b2e1c188072e3cbc91500cfd503fb26d50ffc && \\\n+ patch -p1 < /patch.diff && \\\n+ cd /llvm-project/compiler-rt/lib/fuzzer && \\\n+ (for f in *.cpp; do \\\n+ clang++ -stdlib=libc++ -fPIC -O2 -std=c++11 $f -c & \\\n+ done && wait) && \\\n+ ar r /libEntropic.a *.o\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_keepseed/fuzzer.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Integration code for libFuzzer fuzzer.\"\"\"\n+\n+from fuzzers.entropic_interceptors import fuzzer as entropic_fuzzer\n+from fuzzers.libfuzzer_interceptors import fuzzer as libfuzzer_fuzzer\n+\n+\n+def build():\n+ \"\"\"Build benchmark.\"\"\"\n+ entropic_fuzzer.build()\n+\n+\n+def fuzz(input_corpus, output_corpus, target_binary):\n+ \"\"\"Run fuzzer.\"\"\"\n+ libfuzzer_fuzzer.run_fuzzer(input_corpus,\n+ output_corpus,\n+ target_binary,\n+ extra_flags=['-entropic=1', '-keep_seed=1'])\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_keepseed/patch.diff", "diff": "+commit aac9771fa16e3fc00725d4bbd662d71186a09532\n+Author: Dokyung Song <[email protected]>\n+Date: Fri Jul 31 00:07:20 2020 +0000\n+\n+ [libFuzzer] Optionally keep initial seed inputs regardless of whether they discover new features or not.\n+\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerCorpus.h b/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n+index 54d1e09ec6d..80398a9d7ce 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n++++ b/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n+@@ -33,6 +33,7 @@ struct InputInfo {\n+ // Stats.\n+ size_t NumExecutedMutations = 0;\n+ size_t NumSuccessfullMutations = 0;\n++ bool SeedInput = false;\n+ bool MayDeleteFile = false;\n+ bool Reduced = false;\n+ bool HasFocusFunction = false;\n+@@ -131,9 +132,11 @@ class InputCorpus {\n+\n+ EntropicOptions Entropic;\n+\n++ bool KeepSeed = false;\n++\n+ public:\n+- InputCorpus(const std::string &OutputCorpus, EntropicOptions Entropic)\n+- : Entropic(Entropic), OutputCorpus(OutputCorpus) {\n++ InputCorpus(const std::string &OutputCorpus, EntropicOptions Entropic, bool KeepSeed)\n++ : Entropic(Entropic), OutputCorpus(OutputCorpus), KeepSeed(KeepSeed) {\n+ memset(InputSizesPerFeature, 0, sizeof(InputSizesPerFeature));\n+ memset(SmallestElementPerFeature, 0, sizeof(SmallestElementPerFeature));\n+ }\n+@@ -177,7 +180,7 @@ public:\n+ bool empty() const { return Inputs.empty(); }\n+ const Unit &operator[] (size_t Idx) const { return Inputs[Idx]->U; }\n+ InputInfo *AddToCorpus(const Unit &U, size_t NumFeatures, bool MayDeleteFile,\n+- bool HasFocusFunction,\n++ bool HasFocusFunction, bool SeedInput,\n+ const Vector<uint32_t> &FeatureSet,\n+ const DataFlowTrace &DFT, const InputInfo *BaseII) {\n+ assert(!U.empty());\n+@@ -187,6 +190,7 @@ public:\n+ InputInfo &II = *Inputs.back();\n+ II.U = U;\n+ II.NumFeatures = NumFeatures;\n++ II.SeedInput = SeedInput;\n+ II.MayDeleteFile = MayDeleteFile;\n+ II.UniqFeatureSet = FeatureSet;\n+ II.HasFocusFunction = HasFocusFunction;\n+@@ -471,7 +475,7 @@ private:\n+\n+ for (size_t i = 0; i < N; i++) {\n+\n+- if (Inputs[i]->NumFeatures == 0) {\n++ if (Inputs[i]->NumFeatures == 0 && !(Inputs[i]->SeedInput && KeepSeed)) {\n+ // If the seed doesn't represent any features, assign zero energy.\n+ Weights[i] = 0.;\n+ } else if (Inputs[i]->NumExecutedMutations / kMaxMutationFactor >\n+@@ -491,7 +495,7 @@ private:\n+\n+ if (VanillaSchedule) {\n+ for (size_t i = 0; i < N; i++)\n+- Weights[i] = Inputs[i]->NumFeatures\n++ Weights[i] = (Inputs[i]->NumFeatures || (KeepSeed && Inputs[i]->SeedInput))\n+ ? (i + 1) * (Inputs[i]->HasFocusFunction ? 1000 : 1)\n+ : 0.;\n+ }\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerDriver.cpp b/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n+index 00a33a413d2..ef7991c1e27 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n++++ b/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n+@@ -649,6 +649,7 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n+ Options.Verbosity = Flags.verbosity;\n+ Options.MaxLen = Flags.max_len;\n+ Options.LenControl = Flags.len_control;\n++ Options.KeepSeed = Flags.keep_seed;\n+ Options.UnitTimeoutSec = Flags.timeout;\n+ Options.ErrorExitCode = Flags.error_exitcode;\n+ Options.TimeoutExitCode = Flags.timeout_exitcode;\n+@@ -753,7 +754,7 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n+\n+ Random Rand(Seed);\n+ auto *MD = new MutationDispatcher(Rand, Options);\n+- auto *Corpus = new InputCorpus(Options.OutputCorpus, Entropic);\n++ auto *Corpus = new InputCorpus(Options.OutputCorpus, Entropic, Options.KeepSeed);\n+ auto *F = new Fuzzer(Callback, *Corpus, *MD, Options);\n+\n+ for (auto &U: Dictionary)\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerFlags.def b/compiler-rt/lib/fuzzer/FuzzerFlags.def\n+index 832224a705d..0dac7e705a3 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerFlags.def\n++++ b/compiler-rt/lib/fuzzer/FuzzerFlags.def\n+@@ -23,6 +23,8 @@ FUZZER_FLAG_INT(len_control, 100, \"Try generating small inputs first, \"\n+ FUZZER_FLAG_STRING(seed_inputs, \"A comma-separated list of input files \"\n+ \"to use as an additional seed corpus. Alternatively, an \\\"@\\\" followed by \"\n+ \"the name of a file containing the comma-separated list.\")\n++FUZZER_FLAG_INT(keep_seed, 0, \"If 1, keep seed inputs for mutation even if \"\n++ \"they do not produce new coverage.\")\n+ FUZZER_FLAG_INT(cross_over, 1, \"If 1, cross over inputs.\")\n+ FUZZER_FLAG_INT(mutate_depth, 5,\n+ \"Apply this number of consecutive mutations to each input.\")\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerFork.cpp b/compiler-rt/lib/fuzzer/FuzzerFork.cpp\n+index d9e6b79443e..38fb82fc12d 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerFork.cpp\n++++ b/compiler-rt/lib/fuzzer/FuzzerFork.cpp\n+@@ -309,11 +309,17 @@ void FuzzWithFork(Random &Rand, const FuzzingOptions &Options,\n+ else\n+ Env.MainCorpusDir = CorpusDirs[0];\n+\n+- auto CFPath = DirPlusFile(Env.TempDir, \"merge.txt\");\n+- CrashResistantMerge(Env.Args, {}, SeedFiles, &Env.Files, {}, &Env.Features,\n+- {}, &Env.Cov,\n+- CFPath, false);\n+- RemoveFile(CFPath);\n++ if (Options.KeepSeed) {\n++ for (auto &File : SeedFiles)\n++ Env.Files.push_back(File.File);\n++ }\n++ else {\n++ auto CFPath = DirPlusFile(Env.TempDir, \"merge.txt\");\n++ CrashResistantMerge(Env.Args, {}, SeedFiles, &Env.Files, {}, &Env.Features,\n++ {}, &Env.Cov,\n++ CFPath, false);\n++ RemoveFile(CFPath);\n++ }\n+ Printf(\"INFO: -fork=%d: %zd seed inputs, starting to fuzz in %s\\n\", NumJobs,\n+ Env.Files.size(), Env.TempDir.c_str());\n+\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerInternal.h b/compiler-rt/lib/fuzzer/FuzzerInternal.h\n+index 31096ce804b..e75807209f5 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerInternal.h\n++++ b/compiler-rt/lib/fuzzer/FuzzerInternal.h\n+@@ -119,6 +119,8 @@ private:\n+\n+ size_t LastCorpusUpdateRun = 0;\n+\n++ bool IsExecutingSeedCorpora = false;\n++\n+ bool HasMoreMallocsThanFrees = false;\n+ size_t NumberOfLeakDetectionAttempts = 0;\n+\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerLoop.cpp b/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n+index 02db6d27b0a..a9af25a3070 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n++++ b/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n+@@ -487,10 +487,11 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\n+ *FoundUniqFeatures = FoundUniqFeaturesOfII;\n+ PrintPulseAndReportSlowInput(Data, Size);\n+ size_t NumNewFeatures = Corpus.NumFeatureUpdates() - NumUpdatesBefore;\n+- if (NumNewFeatures) {\n++ if (NumNewFeatures || (Options.KeepSeed && IsExecutingSeedCorpora)) {\n+ TPC.UpdateObservedPCs();\n+ auto NewII = Corpus.AddToCorpus({Data, Data + Size}, NumNewFeatures,\n+ MayDeleteFile, TPC.ObservedFocusFunction(),\n++ IsExecutingSeedCorpora,\n+ UniqFeatureSetTmp, DFT, II);\n+ WriteFeatureSetToFile(Options.FeaturesDir, Sha1ToString(NewII->Sha1),\n+ NewII->UniqFeatureSet);\n+@@ -764,6 +765,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+ assert(CorporaFiles.front().Size <= CorporaFiles.back().Size);\n+ }\n+\n++ IsExecutingSeedCorpora = true;\n++\n+ // Load and execute inputs one by one.\n+ for (auto &SF : CorporaFiles) {\n+ auto U = FileToVector(SF.File, MaxInputLen, /*ExitOnError=*/false);\n+@@ -773,6 +776,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+ TryDetectingAMemoryLeak(U.data(), U.size(),\n+ /*DuringInitialCorpusExecution*/ true);\n+ }\n++\n++ IsExecutingSeedCorpora = false;\n+ }\n+\n+ PrintStats(\"INITED\");\n+@@ -785,6 +790,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+ Corpus.NumInputsThatTouchFocusFunction());\n+ }\n+\n++ Printf(\"INFO: corpus size = %d\\n\", Corpus.size());\n++\n+ if (Corpus.empty() && Options.MaxNumberOfRuns) {\n+ Printf(\"ERROR: no interesting inputs were found. \"\n+ \"Is the code instrumented for coverage? Exiting.\\n\");\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerOptions.h b/compiler-rt/lib/fuzzer/FuzzerOptions.h\n+index 9d975bd61fe..ccd0b3dcb56 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerOptions.h\n++++ b/compiler-rt/lib/fuzzer/FuzzerOptions.h\n+@@ -18,6 +18,7 @@ struct FuzzingOptions {\n+ int Verbosity = 1;\n+ size_t MaxLen = 0;\n+ size_t LenControl = 1000;\n++ bool KeepSeed = false;\n+ int UnitTimeoutSec = 300;\n+ int TimeoutExitCode = 70;\n+ int OOMExitCode = 71;\n+diff --git a/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp b/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp\n+index 0e9435ab8fc..dfc642ab6d0 100644\n+--- a/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp\n++++ b/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp\n+@@ -593,7 +593,8 @@ TEST(Corpus, Distribution) {\n+ DataFlowTrace DFT;\n+ Random Rand(0);\n+ struct EntropicOptions Entropic = {false, 0xFF, 100};\n+- std::unique_ptr<InputCorpus> C(new InputCorpus(\"\", Entropic));\n++ bool KeepSeed = false;\n++ std::unique_ptr<InputCorpus> C(new InputCorpus(\"\", Entropic, KeepSeed));\n+ size_t N = 10;\n+ size_t TriesPerUnit = 1<<16;\n+ for (size_t i = 0; i < N; i++)\n+@@ -1057,7 +1058,8 @@ TEST(Entropic, UpdateFrequency) {\n+ size_t Index;\n+ // Create input corpus with default entropic configuration\n+ struct EntropicOptions Entropic = {true, 0xFF, 100};\n+- std::unique_ptr<InputCorpus> C(new InputCorpus(\"\", Entropic));\n++ bool KeepSeed = false;\n++ std::unique_ptr<InputCorpus> C(new InputCorpus(\"\", Entropic, KeepSeed));\n+ std::unique_ptr<InputInfo> II(new InputInfo());\n+\n+ C->AddRareFeature(FeatIdx1);\n+@@ -1094,7 +1096,8 @@ double SubAndSquare(double X, double Y) {\n+ TEST(Entropic, ComputeEnergy) {\n+ const double Precision = 0.01;\n+ struct EntropicOptions Entropic = {true, 0xFF, 100};\n+- std::unique_ptr<InputCorpus> C(new InputCorpus(\"\", Entropic));\n++ bool KeepSeed = false;\n++ std::unique_ptr<InputCorpus> C(new InputCorpus(\"\", Entropic, KeepSeed));\n+ std::unique_ptr<InputInfo> II(new InputInfo());\n+ Vector<std::pair<uint32_t, uint16_t>> FeatureFreqs = {{1, 3}, {2, 3}, {3, 3}};\n+ II->FeatureFreqs = FeatureFreqs;\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_keepseed/runner.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+FROM gcr.io/fuzzbench/base-runner\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/libfuzzer_keepseed/builder.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+ARG parent_image\n+FROM $parent_image\n+\n+COPY patch.diff /\n+\n+RUN git clone https://github.com/llvm/llvm-project.git /llvm-project && \\\n+ cd /llvm-project/ && \\\n+ git checkout b52b2e1c188072e3cbc91500cfd503fb26d50ffc && \\\n+ patch -p1 < /patch.diff && \\\n+ cd compiler-rt/lib/fuzzer && \\\n+ (for f in *.cpp; do \\\n+ clang++ -stdlib=libc++ -fPIC -O2 -std=c++11 $f -c & \\\n+ done && wait) && \\\n+ ar r /usr/lib/libFuzzer.a *.o\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/libfuzzer_keepseed/fuzzer.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Integration code for libFuzzer fuzzer.\"\"\"\n+\n+from fuzzers.libfuzzer_interceptors import fuzzer as libfuzzer_fuzzer\n+\n+\n+def build():\n+ \"\"\"Build benchmark.\"\"\"\n+ libfuzzer_fuzzer.build()\n+\n+\n+def fuzz(input_corpus, output_corpus, target_binary):\n+ \"\"\"Run fuzzer.\"\"\"\n+ libfuzzer_fuzzer.run_fuzzer(input_corpus,\n+ output_corpus,\n+ target_binary,\n+ extra_flags=['-keep_seed=1'])\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/libfuzzer_keepseed/patch.diff", "diff": "+commit aac9771fa16e3fc00725d4bbd662d71186a09532\n+Author: Dokyung Song <[email protected]>\n+Date: Fri Jul 31 00:07:20 2020 +0000\n+\n+ [libFuzzer] Optionally keep initial seed inputs regardless of whether they discover new features or not.\n+\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerCorpus.h b/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n+index 54d1e09ec6d..80398a9d7ce 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n++++ b/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n+@@ -33,6 +33,7 @@ struct InputInfo {\n+ // Stats.\n+ size_t NumExecutedMutations = 0;\n+ size_t NumSuccessfullMutations = 0;\n++ bool SeedInput = false;\n+ bool MayDeleteFile = false;\n+ bool Reduced = false;\n+ bool HasFocusFunction = false;\n+@@ -131,9 +132,11 @@ class InputCorpus {\n+\n+ EntropicOptions Entropic;\n+\n++ bool KeepSeed = false;\n++\n+ public:\n+- InputCorpus(const std::string &OutputCorpus, EntropicOptions Entropic)\n+- : Entropic(Entropic), OutputCorpus(OutputCorpus) {\n++ InputCorpus(const std::string &OutputCorpus, EntropicOptions Entropic, bool KeepSeed)\n++ : Entropic(Entropic), OutputCorpus(OutputCorpus), KeepSeed(KeepSeed) {\n+ memset(InputSizesPerFeature, 0, sizeof(InputSizesPerFeature));\n+ memset(SmallestElementPerFeature, 0, sizeof(SmallestElementPerFeature));\n+ }\n+@@ -177,7 +180,7 @@ public:\n+ bool empty() const { return Inputs.empty(); }\n+ const Unit &operator[] (size_t Idx) const { return Inputs[Idx]->U; }\n+ InputInfo *AddToCorpus(const Unit &U, size_t NumFeatures, bool MayDeleteFile,\n+- bool HasFocusFunction,\n++ bool HasFocusFunction, bool SeedInput,\n+ const Vector<uint32_t> &FeatureSet,\n+ const DataFlowTrace &DFT, const InputInfo *BaseII) {\n+ assert(!U.empty());\n+@@ -187,6 +190,7 @@ public:\n+ InputInfo &II = *Inputs.back();\n+ II.U = U;\n+ II.NumFeatures = NumFeatures;\n++ II.SeedInput = SeedInput;\n+ II.MayDeleteFile = MayDeleteFile;\n+ II.UniqFeatureSet = FeatureSet;\n+ II.HasFocusFunction = HasFocusFunction;\n+@@ -471,7 +475,7 @@ private:\n+\n+ for (size_t i = 0; i < N; i++) {\n+\n+- if (Inputs[i]->NumFeatures == 0) {\n++ if (Inputs[i]->NumFeatures == 0 && !(Inputs[i]->SeedInput && KeepSeed)) {\n+ // If the seed doesn't represent any features, assign zero energy.\n+ Weights[i] = 0.;\n+ } else if (Inputs[i]->NumExecutedMutations / kMaxMutationFactor >\n+@@ -491,7 +495,7 @@ private:\n+\n+ if (VanillaSchedule) {\n+ for (size_t i = 0; i < N; i++)\n+- Weights[i] = Inputs[i]->NumFeatures\n++ Weights[i] = (Inputs[i]->NumFeatures || (KeepSeed && Inputs[i]->SeedInput))\n+ ? (i + 1) * (Inputs[i]->HasFocusFunction ? 1000 : 1)\n+ : 0.;\n+ }\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerDriver.cpp b/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n+index 00a33a413d2..ef7991c1e27 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n++++ b/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n+@@ -649,6 +649,7 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n+ Options.Verbosity = Flags.verbosity;\n+ Options.MaxLen = Flags.max_len;\n+ Options.LenControl = Flags.len_control;\n++ Options.KeepSeed = Flags.keep_seed;\n+ Options.UnitTimeoutSec = Flags.timeout;\n+ Options.ErrorExitCode = Flags.error_exitcode;\n+ Options.TimeoutExitCode = Flags.timeout_exitcode;\n+@@ -753,7 +754,7 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n+\n+ Random Rand(Seed);\n+ auto *MD = new MutationDispatcher(Rand, Options);\n+- auto *Corpus = new InputCorpus(Options.OutputCorpus, Entropic);\n++ auto *Corpus = new InputCorpus(Options.OutputCorpus, Entropic, Options.KeepSeed);\n+ auto *F = new Fuzzer(Callback, *Corpus, *MD, Options);\n+\n+ for (auto &U: Dictionary)\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerFlags.def b/compiler-rt/lib/fuzzer/FuzzerFlags.def\n+index 832224a705d..0dac7e705a3 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerFlags.def\n++++ b/compiler-rt/lib/fuzzer/FuzzerFlags.def\n+@@ -23,6 +23,8 @@ FUZZER_FLAG_INT(len_control, 100, \"Try generating small inputs first, \"\n+ FUZZER_FLAG_STRING(seed_inputs, \"A comma-separated list of input files \"\n+ \"to use as an additional seed corpus. Alternatively, an \\\"@\\\" followed by \"\n+ \"the name of a file containing the comma-separated list.\")\n++FUZZER_FLAG_INT(keep_seed, 0, \"If 1, keep seed inputs for mutation even if \"\n++ \"they do not produce new coverage.\")\n+ FUZZER_FLAG_INT(cross_over, 1, \"If 1, cross over inputs.\")\n+ FUZZER_FLAG_INT(mutate_depth, 5,\n+ \"Apply this number of consecutive mutations to each input.\")\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerFork.cpp b/compiler-rt/lib/fuzzer/FuzzerFork.cpp\n+index d9e6b79443e..38fb82fc12d 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerFork.cpp\n++++ b/compiler-rt/lib/fuzzer/FuzzerFork.cpp\n+@@ -309,11 +309,17 @@ void FuzzWithFork(Random &Rand, const FuzzingOptions &Options,\n+ else\n+ Env.MainCorpusDir = CorpusDirs[0];\n+\n+- auto CFPath = DirPlusFile(Env.TempDir, \"merge.txt\");\n+- CrashResistantMerge(Env.Args, {}, SeedFiles, &Env.Files, {}, &Env.Features,\n+- {}, &Env.Cov,\n+- CFPath, false);\n+- RemoveFile(CFPath);\n++ if (Options.KeepSeed) {\n++ for (auto &File : SeedFiles)\n++ Env.Files.push_back(File.File);\n++ }\n++ else {\n++ auto CFPath = DirPlusFile(Env.TempDir, \"merge.txt\");\n++ CrashResistantMerge(Env.Args, {}, SeedFiles, &Env.Files, {}, &Env.Features,\n++ {}, &Env.Cov,\n++ CFPath, false);\n++ RemoveFile(CFPath);\n++ }\n+ Printf(\"INFO: -fork=%d: %zd seed inputs, starting to fuzz in %s\\n\", NumJobs,\n+ Env.Files.size(), Env.TempDir.c_str());\n+\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerInternal.h b/compiler-rt/lib/fuzzer/FuzzerInternal.h\n+index 31096ce804b..e75807209f5 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerInternal.h\n++++ b/compiler-rt/lib/fuzzer/FuzzerInternal.h\n+@@ -119,6 +119,8 @@ private:\n+\n+ size_t LastCorpusUpdateRun = 0;\n+\n++ bool IsExecutingSeedCorpora = false;\n++\n+ bool HasMoreMallocsThanFrees = false;\n+ size_t NumberOfLeakDetectionAttempts = 0;\n+\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerLoop.cpp b/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n+index 02db6d27b0a..a9af25a3070 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n++++ b/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n+@@ -487,10 +487,11 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\n+ *FoundUniqFeatures = FoundUniqFeaturesOfII;\n+ PrintPulseAndReportSlowInput(Data, Size);\n+ size_t NumNewFeatures = Corpus.NumFeatureUpdates() - NumUpdatesBefore;\n+- if (NumNewFeatures) {\n++ if (NumNewFeatures || (Options.KeepSeed && IsExecutingSeedCorpora)) {\n+ TPC.UpdateObservedPCs();\n+ auto NewII = Corpus.AddToCorpus({Data, Data + Size}, NumNewFeatures,\n+ MayDeleteFile, TPC.ObservedFocusFunction(),\n++ IsExecutingSeedCorpora,\n+ UniqFeatureSetTmp, DFT, II);\n+ WriteFeatureSetToFile(Options.FeaturesDir, Sha1ToString(NewII->Sha1),\n+ NewII->UniqFeatureSet);\n+@@ -764,6 +765,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+ assert(CorporaFiles.front().Size <= CorporaFiles.back().Size);\n+ }\n+\n++ IsExecutingSeedCorpora = true;\n++\n+ // Load and execute inputs one by one.\n+ for (auto &SF : CorporaFiles) {\n+ auto U = FileToVector(SF.File, MaxInputLen, /*ExitOnError=*/false);\n+@@ -773,6 +776,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+ TryDetectingAMemoryLeak(U.data(), U.size(),\n+ /*DuringInitialCorpusExecution*/ true);\n+ }\n++\n++ IsExecutingSeedCorpora = false;\n+ }\n+\n+ PrintStats(\"INITED\");\n+@@ -785,6 +790,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+ Corpus.NumInputsThatTouchFocusFunction());\n+ }\n+\n++ Printf(\"INFO: corpus size = %d\\n\", Corpus.size());\n++\n+ if (Corpus.empty() && Options.MaxNumberOfRuns) {\n+ Printf(\"ERROR: no interesting inputs were found. \"\n+ \"Is the code instrumented for coverage? Exiting.\\n\");\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerOptions.h b/compiler-rt/lib/fuzzer/FuzzerOptions.h\n+index 9d975bd61fe..ccd0b3dcb56 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerOptions.h\n++++ b/compiler-rt/lib/fuzzer/FuzzerOptions.h\n+@@ -18,6 +18,7 @@ struct FuzzingOptions {\n+ int Verbosity = 1;\n+ size_t MaxLen = 0;\n+ size_t LenControl = 1000;\n++ bool KeepSeed = false;\n+ int UnitTimeoutSec = 300;\n+ int TimeoutExitCode = 70;\n+ int OOMExitCode = 71;\n+diff --git a/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp b/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp\n+index 0e9435ab8fc..dfc642ab6d0 100644\n+--- a/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp\n++++ b/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp\n+@@ -593,7 +593,8 @@ TEST(Corpus, Distribution) {\n+ DataFlowTrace DFT;\n+ Random Rand(0);\n+ struct EntropicOptions Entropic = {false, 0xFF, 100};\n+- std::unique_ptr<InputCorpus> C(new InputCorpus(\"\", Entropic));\n++ bool KeepSeed = false;\n++ std::unique_ptr<InputCorpus> C(new InputCorpus(\"\", Entropic, KeepSeed));\n+ size_t N = 10;\n+ size_t TriesPerUnit = 1<<16;\n+ for (size_t i = 0; i < N; i++)\n+@@ -1057,7 +1058,8 @@ TEST(Entropic, UpdateFrequency) {\n+ size_t Index;\n+ // Create input corpus with default entropic configuration\n+ struct EntropicOptions Entropic = {true, 0xFF, 100};\n+- std::unique_ptr<InputCorpus> C(new InputCorpus(\"\", Entropic));\n++ bool KeepSeed = false;\n++ std::unique_ptr<InputCorpus> C(new InputCorpus(\"\", Entropic, KeepSeed));\n+ std::unique_ptr<InputInfo> II(new InputInfo());\n+\n+ C->AddRareFeature(FeatIdx1);\n+@@ -1094,7 +1096,8 @@ double SubAndSquare(double X, double Y) {\n+ TEST(Entropic, ComputeEnergy) {\n+ const double Precision = 0.01;\n+ struct EntropicOptions Entropic = {true, 0xFF, 100};\n+- std::unique_ptr<InputCorpus> C(new InputCorpus(\"\", Entropic));\n++ bool KeepSeed = false;\n++ std::unique_ptr<InputCorpus> C(new InputCorpus(\"\", Entropic, KeepSeed));\n+ std::unique_ptr<InputInfo> II(new InputInfo());\n+ Vector<std::pair<uint32_t, uint16_t>> FeatureFreqs = {{1, 3}, {2, 3}, {3, 3}};\n+ II->FeatureFreqs = FeatureFreqs;\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/libfuzzer_keepseed/runner.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+FROM gcr.io/fuzzbench/base-runner\n" }, { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-08-01\n+ fuzzers:\n+ - afl\n+ - aflfast\n+ - aflplusplus\n+ - aflplusplus_optimal\n+ - aflplusplus_qemu\n+ - afl_qemu\n+ - aflsmart\n+ - eclipser\n+ - entropic\n+ - fairfuzz\n+ - fastcgs_lm\n+ - honggfuzz\n+ - honggfuzz_qemu\n+ - lafintel\n+ - libfuzzer\n+ - manul\n+ - mopt\n+ - libfuzzer_keepseed\n+ - entropic_keepseed\n+\n- experiment: 2020-07-30\nfuzzers:\n- libfuzzer\n" } ]

[ { "change_type": "MODIFY", "old_path": "experiment/measurer.py", "new_path": "experiment/measurer.py", "diff": "@@ -163,6 +163,7 @@ def get_covered_region(experiment: str, fuzzer: str, benchmark: str,\n})\nsnapshot_measurer = SnapshotMeasurer(fuzzer, benchmark, trial_id,\nsnapshot_logger)\n+ snapshot_measurer.generate_summary(0, summary_only=False)\nnew_covered_regions = snapshot_measurer.get_current_covered_regions()\ncovered_regions[key] = covered_regions[key].union(new_covered_regions)\nq.put(covered_regions)\n@@ -536,21 +537,29 @@ class SnapshotMeasurer: # pylint: disable=too-many-instance-attributes\nself.logger.error(\n'Coverage profdata generation failed for cycle: %d.', cycle)\n- def generate_summary(self, cycle: int):\n+ def generate_summary(self, cycle: int, summary_only=True):\n\"\"\"Transform the .profdata file into json form.\"\"\"\ncoverage_binary = get_coverage_binary(self.benchmark)\ncommand = [\n'llvm-cov', 'export', '-format=text', coverage_binary,\n'-instr-profile=%s' % self.profdata_file\n]\n+\n+ if summary_only:\n+ command.append('-summary-only')\n+\nwith open(self.cov_summary_file, 'w') as output_file:\nresult = new_process.execute(command,\noutput_file=output_file,\nexpect_zero=False)\nif result.retcode != 0:\n+ if cycle != 0:\nself.logger.error(\n'Coverage summary json file generation failed for \\\ncycle: %d.', cycle)\n+ else:\n+ self.logger.error(\n+ 'Coverage summary json file generation failed in the end.')\ndef generate_coverage_information(self, cycle: int):\n\"\"\"Generate the .profdata file and then transform it into\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_measurer.py", "new_path": "experiment/test_measurer.py", "diff": "@@ -163,7 +163,7 @@ def test_generate_summary(mocked_get_coverage_binary, mocked_execute,\nexpected = [\n'llvm-cov', 'export', '-format=text',\n'/work/coverage-binaries/benchmark-a/fuzz-target',\n- '-instr-profile=/reports/data.profdata'\n+ '-instr-profile=/reports/data.profdata', '-summary-only'\n]\nassert (len(mocked_execute.call_args_list)) == 1\n" } ]

[ { "change_type": "MODIFY", "old_path": "docker/generate_makefile.py", "new_path": "docker/generate_makefile.py", "diff": "@@ -58,7 +58,7 @@ def _print_makefile_run_template(image):\nif run_type == 'test-run':\nprint('\\t-e MAX_TOTAL_TIME=20 \\\\\\n\\t-e SNAPSHOT_PERIOD=10 \\\\')\nif run_type == 'debug':\n- print('\\t-entrypoint \"/bin/bash\" \\\\\\n\\t-it ', end='')\n+ print('\\t--entrypoint \"/bin/bash\" \\\\\\n\\t-it ', end='')\nelse:\nprint('\\t', end='')\n" } ]

[ { "change_type": "MODIFY", "old_path": "docker/generate_makefile.py", "new_path": "docker/generate_makefile.py", "diff": "@@ -59,6 +59,8 @@ def _print_makefile_run_template(image):\nprint('\\t-e MAX_TOTAL_TIME=20 \\\\\\n\\t-e SNAPSHOT_PERIOD=10 \\\\')\nif run_type == 'debug':\nprint('\\t--entrypoint \"/bin/bash\" \\\\\\n\\t-it ', end='')\n+ elif run_type == 'run':\n+ print('\\t-it ', end='')\nelse:\nprint('\\t', end='')\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/fuzzers.yml", "new_path": ".github/workflows/fuzzers.yml", "diff": "@@ -66,6 +66,20 @@ jobs:\nwith:\npython-version: 3.7\n+ # Copied from:\n+ # https://docs.github.com/en/actions/language-and-framework-guides/using-python-with-github-actions\n+ - name: Cache pip\n+ uses: actions/cache@v2\n+ with:\n+ # This path is specific to Ubuntu.\n+ path: ~/.cache/pip\n+ # Look to see if there is a cache hit for the corresponding requirements\n+ # file.\n+ key: ${{ runner.os }}-pip-${{ hashFiles('requirements.txt') }}\n+ restore-keys: |\n+ ${{ runner.os }}-pip-\n+ ${{ runner.os }}-\n+\n- name: Install dependencies\nrun: |\nmake install-dependencies\n" }, { "change_type": "MODIFY", "old_path": ".github/workflows/presubmit.yml", "new_path": ".github/workflows/presubmit.yml", "diff": "@@ -16,6 +16,20 @@ jobs:\nwith:\npython-version: 3.7\n+ # Copied from:\n+ # https://docs.github.com/en/actions/language-and-framework-guides/using-python-with-github-actions\n+ - name: Cache pip\n+ uses: actions/cache@v2\n+ with:\n+ # This path is specific to Ubuntu.\n+ path: ~/.cache/pip\n+ # Look to see if there is a cache hit for the corresponding requirements\n+ # file.\n+ key: ${{ runner.os }}-pip-${{ hashFiles('requirements.txt') }}\n+ restore-keys: |\n+ ${{ runner.os }}-pip-\n+ ${{ runner.os }}-\n+\n- name: Install dependencies\nrun: |\nmake install-dependencies\n" } ]

[ { "change_type": "MODIFY", "old_path": "experiment/resources/runner-startup-script-template.sh", "new_path": "experiment/resources/runner-startup-script-template.sh", "diff": "@@ -36,6 +36,8 @@ docker run \\\n-e EXPERIMENT={{experiment}} \\\n-e TRIAL_ID={{trial_id}} \\\n-e MAX_TOTAL_TIME={{max_total_time}} \\\n+-e NO_SEEDS={{no_seeds}} \\\n+-e NO_DICTIONARIES={{no_dictionaries}} \\\n-e DOCKER_REGISTRY={{docker_registry}} {% if not local_experiment %}-e CLOUD_PROJECT={{cloud_project}} -e CLOUD_COMPUTE_ZONE={{cloud_compute_zone}} {% endif %}\\\n-e EXPERIMENT_FILESTORE={{experiment_filestore}} {% if local_experiment %}-v {{experiment_filestore}}:{{experiment_filestore}} {% endif %}\\\n-e REPORT_FILESTORE={{report_filestore}} {% if local_experiment %}-v {{report_filestore}}:{{report_filestore}} {% endif %}\\\n" }, { "change_type": "MODIFY", "old_path": "experiment/run_experiment.py", "new_path": "experiment/run_experiment.py", "diff": "@@ -37,7 +37,6 @@ from common import logs\nfrom common import new_process\nfrom common import utils\nfrom common import yaml_utils\n-from src_analysis import experiment_changes\nBENCHMARKS_DIR = os.path.join(utils.ROOT_DIR, 'benchmarks')\nFUZZERS_DIR = os.path.join(utils.ROOT_DIR, 'fuzzers')\n@@ -199,8 +198,13 @@ def get_git_hash():\nreturn output.strip().decode('utf-8')\n-def start_experiment(experiment_name: str, config_filename: str,\n- benchmarks: List[str], fuzzers: List[str]):\n+def start_experiment( # pylint: disable=too-many-arguments\n+ experiment_name: str,\n+ config_filename: str,\n+ benchmarks: List[str],\n+ fuzzers: List[str],\n+ no_seeds=False,\n+ no_dictionaries=False):\n\"\"\"Start a fuzzer benchmarking experiment.\"\"\"\ncheck_no_local_changes()\n@@ -212,6 +216,8 @@ def start_experiment(experiment_name: str, config_filename: str,\nconfig['benchmarks'] = ','.join(benchmarks)\nconfig['experiment'] = experiment_name\nconfig['git_hash'] = get_git_hash()\n+ config['no_seeds'] = no_seeds\n+ config['no_dictionaries'] = no_dictionaries\nset_up_experiment_config_file(config)\n@@ -440,36 +446,35 @@ def main():\n'--experiment-name',\nhelp='Experiment name.',\nrequired=True)\n- fuzzers_group = parser.add_mutually_exclusive_group()\n- fuzzers_group.add_argument('-f',\n+ parser.add_argument('-f',\n'--fuzzers',\nhelp='Fuzzers to use.',\nnargs='+',\nrequired=False,\ndefault=None,\nchoices=all_fuzzers)\n- fuzzers_group.add_argument('-cf',\n- '--changed-fuzzers',\n- help=('Use fuzzers that have changed since the '\n- 'last experiment. The last experiment is '\n- 'determined by the database your '\n- 'experiment uses, not necessarily the '\n- 'fuzzbench service'),\n- action='store_true',\n- required=False)\n+ parser.add_argument('-ns',\n+ '--no-seeds',\n+ help='Should trials be conducted without seed corpora.',\n+ required=False,\n+ default=False,\n+ action='store_true')\n+ parser.add_argument('-nd',\n+ '--no-dictionaries',\n+ help='Should trials be conducted without dictionaries.',\n+ required=False,\n+ default=False,\n+ action='store_true')\nargs = parser.parse_args()\n-\n- if args.changed_fuzzers:\n- fuzzers = experiment_changes.get_fuzzers_changed_since_last()\n- if not fuzzers:\n- logs.error('No fuzzers changed since last experiment. Exiting.')\n- return 1\n- else:\nfuzzers = args.fuzzers or all_fuzzers\n- start_experiment(args.experiment_name, args.experiment_config,\n- args.benchmarks, fuzzers)\n+ start_experiment(args.experiment_name,\n+ args.experiment_config,\n+ args.benchmarks,\n+ fuzzers,\n+ no_seeds=args.no_seeds,\n+ no_dictionaries=args.no_dictionaries)\nreturn 0\n" }, { "change_type": "MODIFY", "old_path": "experiment/runner.py", "new_path": "experiment/runner.py", "diff": "@@ -65,11 +65,20 @@ fuzzer_errored_out = False # pylint:disable=invalid-name\ndef _clean_seed_corpus(seed_corpus_dir):\n- \"\"\"Moves seed corpus files from sub-directories into the corpus directory\n- root. Also, deletes any files that exceed the 1 MB limit.\"\"\"\n+ \"\"\"Prepares |seed_corpus_dir| for the trial. This ensures that it can be\n+ used by AFL which is picky about the seed corpus. Moves seed corpus files\n+ from sub-directories into the corpus directory root. Also, deletes any files\n+ that exceed the 1 MB limit. If the NO_SEEDS env var is specified than the\n+ seed corpus files are deleted.\"\"\"\nif not os.path.exists(seed_corpus_dir):\nreturn\n+ if environment.get('NO_SEEDS'):\n+ logs.info('NO_SEEDS specified, deleting seed corpus files.')\n+ shutil.rmtree(seed_corpus_dir)\n+ os.mkdir(seed_corpus_dir)\n+ return\n+\nfailed_to_move_files = []\nfor root, _, files in os.walk(seed_corpus_dir):\nfor filename in files:\n" }, { "change_type": "MODIFY", "old_path": "experiment/scheduler.py", "new_path": "experiment/scheduler.py", "diff": "@@ -733,7 +733,9 @@ def render_startup_script_template(instance_name: str, fuzzer: str,\n'fuzz_target': fuzz_target,\n'docker_image_url': docker_image_url,\n'docker_registry': experiment_config['docker_registry'],\n- 'local_experiment': local_experiment\n+ 'local_experiment': local_experiment,\n+ 'no_seeds': experiment_config['no_seeds'],\n+ 'no_dictionaries': experiment_config['no_dictionaries'],\n}\nif not local_experiment:\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_data/experiment-config.yaml", "new_path": "experiment/test_data/experiment-config.yaml", "diff": "@@ -24,3 +24,5 @@ cloud_sql_instance_connection_name: \"fuzzbench:us-central1:experiment-db=tcp:543\nbenchmarks: \"benchmark-1,benchmark-2\"\nfuzzers: \"fuzzer-a,fuzzer-b\"\ngit_hash: \"git-hash\"\n+no_seeds: false\n+no_dictionaries: false\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_runner.py", "new_path": "experiment/test_runner.py", "diff": "@@ -297,6 +297,18 @@ class TestIntegrationRunner:\nmocked_error.assert_not_called()\n+def test_clean_seed_corpus_no_seeds(fs):\n+ \"\"\"Test that seed corpus files are deleted if NO_SEEDS is set in the\n+ environment to 'True'.\"\"\"\n+ seed_corpus_dir = '/seeds'\n+ fs.create_dir(seed_corpus_dir)\n+ seed_file = os.path.join(seed_corpus_dir, 'a')\n+ fs.create_file(seed_file, contents='abc')\n+ runner._clean_seed_corpus(seed_corpus_dir) # pylint: disable=protected-access\n+ assert not os.path.exists(seed_file)\n+ assert os.path.exists(seed_corpus_dir)\n+\n+\ndef test_clean_seed_corpus(fs):\n\"\"\"Test that seed corpus files are moved to root directory and deletes files\nexceeding 1 MB limit.\"\"\"\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_scheduler.py", "new_path": "experiment/test_scheduler.py", "diff": "@@ -108,6 +108,8 @@ docker run \\\\\n-e EXPERIMENT=test-experiment \\\\\n-e TRIAL_ID=9 \\\\\n-e MAX_TOTAL_TIME=86400 \\\\\n+-e NO_SEEDS=False \\\\\n+-e NO_DICTIONARIES=False \\\\\n-e DOCKER_REGISTRY=gcr.io/fuzzbench -e CLOUD_PROJECT=fuzzbench -e CLOUD_COMPUTE_ZONE=us-central1-a \\\\\n-e EXPERIMENT_FILESTORE=gs://experiment-data \\\\\n-e REPORT_FILESTORE=gs://web-reports \\\\\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/test_utils.py", "new_path": "fuzzers/test_utils.py", "diff": "@@ -62,9 +62,9 @@ def test_dictionary_options_file_with_dict(fs):\nassert utils.get_dictionary_path('/fuzz-target') == '/fuzz.dict'\n-def test_dictionary_skip(fs, environ):\n- \"\"\"Test that None is return when SKIP_DICT is set.\"\"\"\n- os.environ['SKIP_DICT'] = '1'\n+def test_dictionary_no_dictionaries(fs, environ):\n+ \"\"\"Test that None is return when NO_DICTIONARIES is set.\"\"\"\n+ os.environ['NO_DICTIONARIES'] = '1'\nfs.create_file('/fuzz-target.dict', contents='A')\nassert utils.get_dictionary_path('/fuzz-target') is None\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/utils.py", "new_path": "fuzzers/utils.py", "diff": "@@ -117,7 +117,8 @@ def restore_directory(directory):\ndef get_dictionary_path(target_binary):\n\"\"\"Return dictionary path for a target binary.\"\"\"\n- if os.getenv('SKIP_DICT'):\n+ if os.getenv('NO_DICTIONARIES'):\n+ # Don't use dictionaries if experiment specifies not to.\nreturn None\ndictionary_path = target_binary + '.dict'\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/fuzzers.yml", "new_path": ".github/workflows/fuzzers.yml", "diff": "@@ -60,7 +60,7 @@ jobs:\nsteps:\n- uses: actions/checkout@v2\n- run: | # Needed for git diff to work.\n- git fetch origin master --depth 1\n+ git fetch origin master --unshallow\ngit symbolic-ref refs/remotes/origin/HEAD refs/remotes/origin/master\n- name: Clear unnecessary files\n" }, { "change_type": "MODIFY", "old_path": ".github/workflows/presubmit.yml", "new_path": ".github/workflows/presubmit.yml", "diff": "@@ -8,7 +8,7 @@ jobs:\nsteps:\n- uses: actions/checkout@v2\n- run: | # Needed for presubmit to work.\n- git fetch origin master --depth 1\n+ git fetch origin master --unshallow\ngit symbolic-ref refs/remotes/origin/HEAD refs/remotes/origin/master\n- name: Setup Python environment\n" }, { "change_type": "MODIFY", "old_path": "src_analysis/diff_utils.py", "new_path": "src_analysis/diff_utils.py", "diff": "@@ -52,7 +52,7 @@ def get_changed_files(commit_name: str = 'origin...') -> List[str]:\npass\nraise DiffError((\n'\"%s\" failed.\\n'\n- 'Please run \"git fetch origin master && '\n+ 'Please run \"git fetch origin master --unshallow && '\n'git symbolic-ref refs/remotes/origin/HEAD refs/remotes/origin/master\" '\n'and try again.\\n'\n'Please file an issue if this doesn\\'t fix things.') %\n" } ]

[ { "change_type": "MODIFY", "old_path": "docker/gcb/base-images.yaml", "new_path": "docker/gcb/base-images.yaml", "diff": "# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n# See the License for the specific language governing permissions and\n# limitations under the License.\n-\n-# The order of execution is as follows:\n-# 1. Pull base-image.\n-# 2. After pulling base-image completes: Build base-image.\n-# This is optimized to take advantage of concurrency and caching as much as\n-# possible. When nothing changes, this does build takes two minutes to complete\n-# instead of ten.'\n-# See https://cloud.google.com/cloud-build/docs/speeding-up-builds for a discussion\n-# on using caching.\n-# See https://cloud.google.com/cloud-build/docs/configuring-builds/configure-build-step-order\n-# for a discussion on paralellism/concurrency for build steps in GCB.\n-\n-steps:\n-\n-# Pull base-image to fill cache.\n-- name: 'gcr.io/cloud-builders/docker'\n- entrypoint: 'bash'\n- args:\n- - '-c'\n- - |\n- docker pull ${_REPO}/base-image || exit 0\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- args: [\n- 'build',\n-\n- # Use two tags so that the image builds properly and we can push it to the\n- # correct location.\n- '--tag',\n- 'gcr.io/fuzzbench/base-image',\n-\n- '--tag',\n- '${_REPO}/base-image:${_EXPERIMENT}',\n-\n- '--cache-from',\n- '${_REPO}/base-image',\n-\n- 'docker/base-image'\n- ]\n- id: 'base-image'\n+# This file is used by GCB to build base images everytime a commit is\n+# made to the master branch.\nimages:\n- - '${_REPO}/base-image:${_EXPERIMENT}'\n- - '${_REPO}/base-image'\n+- gcr.io/fuzzbench/base-image:test-experiment\n+- gcr.io/fuzzbench/base-image\n+steps:\n+- args:\n+ - build\n+ - --tag\n+ - gcr.io/fuzzbench/base-image\n+ - --tag\n+ - gcr.io/fuzzbench/base-image:test-experiment\n+ - --cache-from\n+ - gcr.io/fuzzbench/base-image\n+ - --build-arg\n+ - BUILDKIT_INLINE_CACHE=1\n+ - --file\n+ - docker/base-image/Dockerfile\n+ - docker/base-image\n+ env: DOCKER_BUILDKIT=1\n+ id: base-image\n+ name: gcr.io/cloud-builders/docker\n" }, { "change_type": "DELETE", "old_path": "docker/gcb/coverage.yaml", "new_path": null, "diff": "-# Copyright 2020 Google LLC\n-#\n-# Licensed under the Apache License, Version 2.0 (the \"License\");\n-# you may not use this file except in compliance with the License.\n-# You may obtain a copy of the License at\n-#\n-# http://www.apache.org/licenses/LICENSE-2.0\n-#\n-# Unless required by applicable law or agreed to in writing, software\n-# distributed under the License is distributed on an \"AS IS\" BASIS,\n-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n-# See the License for the specific language governing permissions and\n-# limitations under the License.\n-\n-steps:\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- entrypoint: 'bash'\n- args:\n- - '-c'\n- - |\n- docker pull ${_REPO}/builders/benchmark/${_BENCHMARK} || exit 0\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- args: [\n- 'build',\n-\n- # Use two tags so that the image builds properly and we can push it to the\n- # correct location.\n- '--tag',\n- 'gcr.io/fuzzbench/builders/benchmark/${_BENCHMARK}',\n-\n- '--tag',\n- '${_REPO}/builders/benchmark/${_BENCHMARK}:${_EXPERIMENT}',\n-\n- '--file=benchmarks/${_BENCHMARK}/Dockerfile',\n-\n- '--cache-from',\n- '${_REPO}/builders/benchmark/${_BENCHMARK}:${_EXPERIMENT}',\n-\n- 'benchmarks/${_BENCHMARK}',\n- ]\n- id: 'build-project-builder'\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- entrypoint: 'bash'\n- args:\n- - '-c'\n- - |\n- docker pull ${_REPO}/builders/coverage/${_BENCHMARK}-intermediate || exit 0\n- id: 'pull-coverage-benchmark-builder-intermediate'\n- wait_for: ['-']\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- args: [\n- 'build',\n-\n- # Use two tags so that the image builds properly and we can push it to the\n- # correct location.\n- '--tag',\n- 'gcr.io/fuzzbench/builders/coverage/${_BENCHMARK}-intermediate',\n-\n- '--tag',\n- '${_REPO}/builders/coverage/${_BENCHMARK}-intermediate:${_EXPERIMENT}',\n-\n- '--file=fuzzers/coverage/builder.Dockerfile',\n-\n- '--cache-from',\n- '${_REPO}/builders/coverage/${_BENCHMARK}-intermediate',\n-\n- '--build-arg',\n- 'parent_image=gcr.io/fuzzbench/builders/benchmark/${_BENCHMARK}',\n-\n- 'fuzzers/coverage',\n- ]\n- id: 'build-coverage-benchmark-builder-intermediate'\n- wait_for: ['build-project-builder', 'pull-coverage-benchmark-builder-intermediate']\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- entrypoint: 'bash'\n- args:\n- - '-c'\n- - |\n- docker pull ${_REPO}/builders/coverage/${_BENCHMARK} || exit 0\n- id: 'pull-coverage-benchmark-builder'\n- wait_for: ['-']\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- args: [\n- 'build',\n-\n- '--tag',\n- 'gcr.io/fuzzbench/builders/coverage/${_BENCHMARK}',\n-\n- '--tag',\n- '${_REPO}/builders/coverage/${_BENCHMARK}:${_EXPERIMENT}',\n-\n- '--file=docker/benchmark-builder/Dockerfile',\n-\n- '--cache-from',\n- '${_REPO}/builders/coverage/${_BENCHMARK}',\n-\n- '--build-arg',\n- 'parent_image=gcr.io/fuzzbench/builders/coverage/${_BENCHMARK}-intermediate',\n-\n- '--build-arg',\n- 'fuzzer=coverage',\n-\n- '--build-arg',\n- 'benchmark=${_BENCHMARK}',\n-\n- '.',\n- ]\n- wait_for: ['pull-coverage-benchmark-builder', 'build-coverage-benchmark-builder-intermediate']\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- args: [\n- 'run',\n- '-v',\n- '/workspace/out:/host-out',\n- '${_REPO}/builders/coverage/${_BENCHMARK}:${_EXPERIMENT}',\n- '/bin/bash',\n- '-c',\n- 'cd /out; tar -czvf /host-out/coverage-build-${_BENCHMARK}.tar.gz *'\n- ]\n-\n-- name: 'gcr.io/cloud-builders/gsutil'\n- args: [\n- '-m',\n- 'cp',\n- '/workspace/out/coverage-build-${_BENCHMARK}.tar.gz',\n- '${_GCS_COVERAGE_BINARIES_DIR}/',\n- ]\n-\n-images:\n- - '${_REPO}/builders/benchmark/${_BENCHMARK}:${_EXPERIMENT}'\n- - '${_REPO}/builders/benchmark/${_BENCHMARK}'\n- - '${_REPO}/builders/coverage/${_BENCHMARK}-intermediate:${_EXPERIMENT}'\n- - '${_REPO}/builders/coverage/${_BENCHMARK}-intermediate'\n- - '${_REPO}/builders/coverage/${_BENCHMARK}:${_EXPERIMENT}'\n- - '${_REPO}/builders/coverage/${_BENCHMARK}'\n" }, { "change_type": "DELETE", "old_path": "docker/gcb/fuzzer.yaml", "new_path": null, "diff": "-# Copyright 2020 Google LLC\n-#\n-# Licensed under the Apache License, Version 2.0 (the \"License\");\n-# you may not use this file except in compliance with the License.\n-# You may obtain a copy of the License at\n-#\n-# http://www.apache.org/licenses/LICENSE-2.0\n-#\n-# Unless required by applicable law or agreed to in writing, software\n-# distributed under the License is distributed on an \"AS IS\" BASIS,\n-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n-# See the License for the specific language governing permissions and\n-# limitations under the License.\n-\n-steps:\n-\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- entrypoint: 'bash'\n- args:\n- - '-c'\n- - |\n- docker pull ${_REPO}/builders/benchmark/${_BENCHMARK} || exit 0\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- args: [\n- 'build',\n-\n- # Use two tags so that the image builds properly and we can push it to the\n- # correct location.\n- '--tag',\n- 'gcr.io/fuzzbench/builders/benchmark/${_BENCHMARK}',\n-\n- '--tag',\n- '${_REPO}/builders/benchmark/${_BENCHMARK}:${_EXPERIMENT}',\n-\n- '--file=benchmarks/${_BENCHMARK}/Dockerfile',\n-\n- '--cache-from',\n- '${_REPO}/builders/benchmark/${_BENCHMARK}:${_EXPERIMENT}',\n-\n- 'benchmarks/${_BENCHMARK}',\n- ]\n- id: 'build-project-builder'\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- entrypoint: 'bash'\n- args:\n- - '-c'\n- - |\n- docker pull ${_REPO}/builders/${_FUZZER}/${_BENCHMARK}-intermediate || exit 0\n- id: 'pull-fuzzer-benchmark-builder-intermediate'\n- wait_for: ['-']\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- args: [\n- 'build',\n-\n- '--tag',\n- 'gcr.io/fuzzbench/builders/${_FUZZER}/${_BENCHMARK}-intermediate',\n-\n- '--tag',\n- '${_REPO}/builders/${_FUZZER}/${_BENCHMARK}-intermediate:${_EXPERIMENT}',\n-\n- '--file=fuzzers/${_FUZZER}/builder.Dockerfile',\n-\n- '--cache-from',\n- '${_REPO}/builders/${_FUZZER}/${_BENCHMARK}-intermediate',\n-\n- # Use a hardcoded repo because the parent image is pinned by SHA. Users\n- # won't have it.\n- '--build-arg',\n- 'parent_image=gcr.io/fuzzbench/builders/benchmark/${_BENCHMARK}',\n-\n- 'fuzzers/${_FUZZER}',\n- ]\n- id: 'build-fuzzer-benchmark-builder-intermediate'\n- wait_for: ['pull-fuzzer-benchmark-builder-intermediate', 'build-project-builder']\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- entrypoint: 'bash'\n- args:\n- - '-c'\n- - |\n- docker pull ${_REPO}/builders/${_FUZZER}/${_BENCHMARK} || exit 0\n- id: 'pull-fuzzer-benchmark-builder'\n- wait_for: ['-']\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- args: [\n- 'build',\n-\n- '--tag',\n- 'gcr.io/fuzzbench/builders/${_FUZZER}/${_BENCHMARK}',\n-\n- '--tag',\n- '${_REPO}/builders/${_FUZZER}/${_BENCHMARK}:${_EXPERIMENT}',\n-\n- '--file=docker/benchmark-builder/Dockerfile',\n-\n- '--cache-from',\n- '${_REPO}/builders/${_FUZZER}/${_BENCHMARK}',\n-\n- '--build-arg',\n- 'parent_image=gcr.io/fuzzbench/builders/${_FUZZER}/${_BENCHMARK}-intermediate',\n-\n- '--build-arg',\n- 'fuzzer=${_FUZZER}',\n-\n- '--build-arg',\n- 'benchmark=${_BENCHMARK}',\n-\n- '.',\n- ]\n- id: 'build-fuzzer-benchmark-builder'\n- wait_for: ['pull-fuzzer-benchmark-builder', 'build-fuzzer-benchmark-builder-intermediate']\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- entrypoint: 'bash'\n- args:\n- - '-c'\n- - |\n- docker pull ${_REPO}/builders/${_FUZZER}/${_BENCHMARK}-intermediate || exit 0\n- id: 'pull-fuzzer-benchmark-runner-intermediate'\n- wait_for: ['-']\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- args: [\n- 'build',\n-\n- '--tag',\n- 'gcr.io/fuzzbench/runners/${_FUZZER}/${_BENCHMARK}-intermediate',\n-\n- '--tag',\n- '${_REPO}/runners/${_FUZZER}/${_BENCHMARK}-intermediate:${_EXPERIMENT}',\n-\n- '--file',\n- 'fuzzers/${_FUZZER}/runner.Dockerfile',\n-\n- '--cache-from',\n- '${_REPO}/runners/${_FUZZER}/${_BENCHMARK}-intermediate',\n-\n- 'fuzzers/${_FUZZER}',\n- ]\n- id: 'build-fuzzer-benchmark-runner-intermediate'\n- wait_for: ['pull-fuzzer-benchmark-runner-intermediate', 'build-fuzzer-benchmark-builder']\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- entrypoint: 'bash'\n- args:\n- - '-c'\n- - |\n- docker pull ${_REPO}/builders/${_FUZZER}/${_BENCHMARK} || exit 0\n- id: 'pull-fuzzer-benchmark-runner'\n- wait_for: ['-']\n-\n-\n-- name: 'gcr.io/cloud-builders/docker'\n- args: [\n- 'build',\n-\n- '--tag',\n- 'gcr.io/fuzzbench/runners/${_FUZZER}/${_BENCHMARK}',\n-\n- '--tag',\n- '${_REPO}/runners/${_FUZZER}/${_BENCHMARK}:${_EXPERIMENT}',\n-\n- '--build-arg',\n- 'fuzzer=${_FUZZER}',\n-\n- '--cache-from',\n- '${_REPO}/runners/${_FUZZER}/${_BENCHMARK}',\n-\n- '--build-arg',\n- 'benchmark=${_BENCHMARK}',\n-\n- '--file',\n- 'docker/benchmark-runner/Dockerfile',\n-\n- '.',\n- ]\n- wait_for: ['pull-fuzzer-benchmark-runner', 'build-fuzzer-benchmark-runner-intermediate']\n-\n-images:\n- - '${_REPO}/builders/benchmark/${_BENCHMARK}:${_EXPERIMENT}'\n- - '${_REPO}/builders/benchmark/${_BENCHMARK}'\n- - '${_REPO}/builders/${_FUZZER}/${_BENCHMARK}-intermediate:${_EXPERIMENT}'\n- - '${_REPO}/builders/${_FUZZER}/${_BENCHMARK}-intermediate'\n- - '${_REPO}/builders/${_FUZZER}/${_BENCHMARK}:${_EXPERIMENT}'\n- - '${_REPO}/builders/${_FUZZER}/${_BENCHMARK}'\n- - '${_REPO}/runners/${_FUZZER}/${_BENCHMARK}-intermediate:${_EXPERIMENT}'\n- - '${_REPO}/runners/${_FUZZER}/${_BENCHMARK}-intermediate'\n- - '${_REPO}/runners/${_FUZZER}/${_BENCHMARK}:${_EXPERIMENT}'\n- - '${_REPO}/runners/${_FUZZER}/${_BENCHMARK}'\n" }, { "change_type": "MODIFY", "old_path": "docker/generate_makefile.py", "new_path": "docker/generate_makefile.py", "diff": "@@ -34,7 +34,8 @@ def _print_benchmark_fuzz_target(benchmarks):\ndef _print_makefile_run_template(image):\n- fuzzer, benchmark = image['tag'].split('/')[1:]\n+ fuzzer = image['fuzzer']\n+ benchmark = image['benchmark']\nfor run_type in ('run', 'debug', 'test-run'):\nprint(('{run_type}-{fuzzer}-{benchmark}: ' +\n@@ -68,8 +69,8 @@ def _print_makefile_run_template(image):\nprint()\n-# TODO(tanq16): Add unit test.\n-def _print_rules_for_image(name, image):\n+# TODO(Tanq16): Function must return a string as opposed to printing it.\n+def print_rules_for_image(name, image):\n\"\"\"Print makefile section for given image to stdout.\"\"\"\nif not ('base' in name or 'dispatcher' in name):\nprint('.', end='')\n@@ -110,7 +111,7 @@ def main():\n_print_benchmark_fuzz_target(benchmarks)\nfor name, image in buildable_images.items():\n- _print_rules_for_image(name, image)\n+ print_rules_for_image(name, image)\n# Print build targets for all fuzzer-benchmark pairs (including coverage).\nfuzzers.append('coverage')\n" }, { "change_type": "MODIFY", "old_path": "docker/image_types.yaml", "new_path": "docker/image_types.yaml", "diff": "# dependency graph.\n'base-image':\n- tag: 'base-image'\n+ dockerfile: 'docker/base-image/Dockerfile'\ncontext: 'docker/base-image'\n+ tag: 'base-image'\n+ type: 'base'\n'dispatcher-image':\n- tag: 'dispatcher-image'\n- context: 'docker/dispatcher-image'\ndepends_on:\n- 'base-image'\n+ dockerfile: 'docker/dispatcher-image/Dockerfile'\n+ context: 'docker/dispatcher-image'\n+ tag: 'dispatcher-image'\n+ type: 'dispatcher'\n'coverage-{benchmark}-builder-intermediate':\n- tag: 'builders/coverage/{benchmark}-intermediate'\n- context: 'fuzzers/coverage'\n- dockerfile: 'fuzzers/coverage/builder.Dockerfile'\n- depends_on:\n- - '{benchmark}-project-builder'\nbuild_arg:\n- 'parent_image=gcr.io/fuzzbench/builders/benchmark/{benchmark}'\n+ depends_on:\n+ - '{benchmark}-project-builder'\n+ dockerfile: 'fuzzers/coverage/builder.Dockerfile'\n+ context: 'fuzzers/coverage'\n+ tag: 'builders/coverage/{benchmark}-intermediate'\n+ type: 'coverage'\n'coverage-{benchmark}-builder':\n- tag: 'builders/coverage/{benchmark}'\n- context: '.'\n- dockerfile: 'docker/benchmark-builder/Dockerfile'\nbuild_arg:\n- - 'parent_image=gcr.io/fuzzbench/builders/coverage/{benchmark}-intermediate'\n- - 'fuzzer=coverage'\n- 'benchmark={benchmark}'\n+ - 'fuzzer=coverage'\n+ - 'parent_image=gcr.io/fuzzbench/builders/coverage/{benchmark}-intermediate'\ndepends_on:\n- 'coverage-{benchmark}-builder-intermediate'\n+ dockerfile: 'docker/benchmark-builder/Dockerfile'\n+ context: '.'\n+ tag: 'builders/coverage/{benchmark}'\n+ type: 'coverage'\n# TODO: It would be better to call this benchmark builder. But that would be\n# confusing because this doesn't involve benchmark-builder/Dockerfile. Rename\n# that and then rename this.\n'{benchmark}-project-builder':\n- tag: 'builders/benchmark/{benchmark}'\n- context: 'benchmarks/{benchmark}'\ndockerfile: 'benchmarks/{benchmark}/Dockerfile'\n+ context: 'benchmarks/{benchmark}'\n+ tag: 'builders/benchmark/{benchmark}'\n+ type: 'builder'\n'{fuzzer}-{benchmark}-builder-intermediate':\n- tag: 'builders/{fuzzer}/{benchmark}-intermediate'\n- context: 'fuzzers/{fuzzer}'\n- dockerfile: 'fuzzers/{fuzzer}/builder.Dockerfile'\n- depends_on:\n- - '{benchmark}-project-builder'\nbuild_arg:\n- 'parent_image=gcr.io/fuzzbench/builders/benchmark/{benchmark}'\n+ depends_on:\n+ - '{benchmark}-project-builder'\n+ dockerfile: 'fuzzers/{fuzzer}/builder.Dockerfile'\n+ context: 'fuzzers/{fuzzer}'\n+ tag: 'builders/{fuzzer}/{benchmark}-intermediate'\n+ type: 'builder'\n'{fuzzer}-{benchmark}-builder':\n- tag: 'builders/{fuzzer}/{benchmark}'\n- context: '.'\n- dockerfile: 'docker/benchmark-builder/Dockerfile'\nbuild_arg:\n- - 'parent_image=gcr.io/fuzzbench/builders/{fuzzer}/{benchmark}-intermediate'\n- - 'fuzzer={fuzzer}'\n- 'benchmark={benchmark}'\n+ - 'fuzzer={fuzzer}'\n+ - 'parent_image=gcr.io/fuzzbench/builders/{fuzzer}/{benchmark}-intermediate'\ndepends_on:\n- '{fuzzer}-{benchmark}-builder-intermediate'\n+ dockerfile: 'docker/benchmark-builder/Dockerfile'\n+ context: '.'\n+ tag: 'builders/{fuzzer}/{benchmark}'\n+ type: 'builder'\n'{fuzzer}-{benchmark}-intermediate-runner':\n- tag: 'runners/{fuzzer}/{benchmark}-intermediate'\n- context: 'fuzzers/{fuzzer}'\n- dockerfile: 'fuzzers/{fuzzer}/runner.Dockerfile'\ndepends_on:\n- '{fuzzer}-{benchmark}-builder'\n+ dockerfile: 'fuzzers/{fuzzer}/runner.Dockerfile'\n+ context: 'fuzzers/{fuzzer}'\n+ tag: 'runners/{fuzzer}/{benchmark}-intermediate'\n+ type: 'runner'\n'{fuzzer}-{benchmark}-runner':\n- tag: 'runners/{fuzzer}/{benchmark}'\n- context: '.'\n- dockerfile: 'docker/benchmark-runner/Dockerfile'\n+ benchmark: '{benchmark}'\nbuild_arg:\n- - 'fuzzer={fuzzer}'\n- 'benchmark={benchmark}'\n+ - 'fuzzer={fuzzer}'\ndepends_on:\n- '{fuzzer}-{benchmark}-intermediate-runner'\n+ dockerfile: 'docker/benchmark-runner/Dockerfile'\n+ fuzzer: '{fuzzer}'\n+ context: '.'\n+ tag: 'runners/{fuzzer}/{benchmark}'\n+ type: 'runner'\n" }, { "change_type": "ADD", "old_path": null, "new_path": "docker/test_generate_makefile.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Tests for generate_makefile.py.\"\"\"\n+\n+from unittest.mock import call\n+from unittest.mock import patch\n+\n+from docker import generate_makefile\n+\n+\n+@patch('builtins.print')\n+def test_print_makefile_build(mocked_print):\n+ \"\"\"Tests result of a makefile generation for an image.\"\"\"\n+\n+ name = 'afl-zlib-builder-intermediate'\n+ image = {\n+ 'tag': 'builders/afl/zlib-intermediate',\n+ 'context': 'fuzzers/afl',\n+ 'dockerfile': 'fuzzers/afl/builder.Dockerfile',\n+ 'depends_on': ['zlib-project-builder'],\n+ 'build_arg': ['parent_image=gcr.io/fuzzbench/builders/benchmark/zlib']\n+ }\n+\n+ generate_makefile.print_rules_for_image(name, image)\n+ assert mocked_print.mock_calls == [\n+ call('.', end=''),\n+ call('afl-zlib-builder-intermediate:', end=''),\n+ call(' .zlib-project-builder', end=''),\n+ call(),\n+ call('\\tdocker build \\\\'),\n+ call('\\t--tag gcr.io/fuzzbench/builders/afl/zlib-intermediate \\\\'),\n+ call('\\t--build-arg BUILDKIT_INLINE_CACHE=1 \\\\'),\n+ call('\\t--cache-from gcr.io/fuzzbench/builders/afl/zlib-intermediate \\\\'\n+ ),\n+ call('\\t--build-arg parent_image=gcr.io/' +\n+ 'fuzzbench/builders/benchmark/zlib \\\\'),\n+ call('\\t--file fuzzers/afl/builder.Dockerfile \\\\'),\n+ call('\\tfuzzers/afl'),\n+ call()\n+ ]\n+\n+\n+@patch('builtins.print')\n+def test_print_makefile_runner_image(mocked_print):\n+ \"\"\"Tests result of a makefile generation for a runner image.\"\"\"\n+\n+ name = 'afl-zlib-runner'\n+ image = {\n+ 'tag': 'runners/afl/zlib',\n+ 'fuzzer': 'afl',\n+ 'benchmark': 'zlib',\n+ 'context': '.',\n+ 'dockerfile': 'docker/benchmark-runner/Dockerfile',\n+ 'build_arg': ['fuzzer=afl', 'benchmark=zlib'],\n+ 'depends_on': ['afl-zlib-builder', 'afl-zlib-intermediate-runner']\n+ }\n+\n+ generate_makefile.print_rules_for_image(name, image)\n+\n+ assert mocked_print.mock_calls == [\n+ call('.', end=''),\n+ call('afl-zlib-runner:', end=''),\n+ call(' .afl-zlib-builder', end=''),\n+ call(' .afl-zlib-intermediate-runner', end=''),\n+ call(),\n+ call('\\tdocker build \\\\'),\n+ call('\\t--tag gcr.io/fuzzbench/runners/afl/zlib \\\\'),\n+ call('\\t--build-arg BUILDKIT_INLINE_CACHE=1 \\\\'),\n+ call('\\t--cache-from gcr.io/fuzzbench/runners/afl/zlib \\\\'),\n+ call('\\t--build-arg fuzzer=afl \\\\'),\n+ call('\\t--build-arg benchmark=zlib \\\\'),\n+ call('\\t--file docker/benchmark-runner/Dockerfile \\\\'),\n+ call('\\t.'),\n+ call(),\n+ call('run-afl-zlib: .afl-zlib-runner'),\n+ call('\\\n+\\tdocker run \\\\\\n\\\n+\\t--cpus=1 \\\\\\n\\\n+\\t--cap-add SYS_NICE \\\\\\n\\\n+\\t--cap-add SYS_PTRACE \\\\\\n\\\n+\\t-e FUZZ_OUTSIDE_EXPERIMENT=1 \\\\\\n\\\n+\\t-e FORCE_LOCAL=1 \\\\\\n\\\n+\\t-e TRIAL_ID=1 \\\\\\n\\\n+\\t-e FUZZER=afl \\\\\\n\\\n+\\t-e BENCHMARK=zlib \\\\\\n\\\n+\\t-e FUZZ_TARGET=$(zlib-fuzz-target) \\\\\\\n+'),\n+ call('\\t-it ', end=''),\n+ call('gcr.io/fuzzbench/runners/afl/zlib'),\n+ call(),\n+ call('debug-afl-zlib: .afl-zlib-runner'),\n+ call('\\\n+\\tdocker run \\\\\\n\\\n+\\t--cpus=1 \\\\\\n\\\n+\\t--cap-add SYS_NICE \\\\\\n\\\n+\\t--cap-add SYS_PTRACE \\\\\\n\\\n+\\t-e FUZZ_OUTSIDE_EXPERIMENT=1 \\\\\\n\\\n+\\t-e FORCE_LOCAL=1 \\\\\\n\\\n+\\t-e TRIAL_ID=1 \\\\\\n\\\n+\\t-e FUZZER=afl \\\\\\n\\\n+\\t-e BENCHMARK=zlib \\\\\\n\\\n+\\t-e FUZZ_TARGET=$(zlib-fuzz-target) \\\\\\\n+'),\n+ call('\\t--entrypoint \"/bin/bash\" \\\\\\n\\t-it ', end=''),\n+ call('gcr.io/fuzzbench/runners/afl/zlib'),\n+ call(),\n+ call('test-run-afl-zlib: .afl-zlib-runner'),\n+ call('\\\n+\\tdocker run \\\\\\n\\\n+\\t--cpus=1 \\\\\\n\\\n+\\t--cap-add SYS_NICE \\\\\\n\\\n+\\t--cap-add SYS_PTRACE \\\\\\n\\\n+\\t-e FUZZ_OUTSIDE_EXPERIMENT=1 \\\\\\n\\\n+\\t-e FORCE_LOCAL=1 \\\\\\n\\\n+\\t-e TRIAL_ID=1 \\\\\\n\\\n+\\t-e FUZZER=afl \\\\\\n\\\n+\\t-e BENCHMARK=zlib \\\\\\n\\\n+\\t-e FUZZ_TARGET=$(zlib-fuzz-target) \\\\\\\n+'),\n+ call('\\t-e MAX_TOTAL_TIME=20 \\\\\\n\\t-e SNAPSHOT_PERIOD=10 \\\\'),\n+ call('\\t', end=''),\n+ call('gcr.io/fuzzbench/runners/afl/zlib'),\n+ call()\n+ ]\n" }, { "change_type": "MODIFY", "old_path": "experiment/build/gcb_build.py", "new_path": "experiment/build/gcb_build.py", "diff": "# limitations under the License.\n\"\"\"Module for building things on Google Cloud Build for use in trials.\"\"\"\n-import os\n+import tempfile\nfrom typing import Dict\n-from common import experiment_path as exp_path\n-from common import experiment_utils\nfrom common import logs\nfrom common import new_process\nfrom common import utils\n+from common import yaml_utils\nfrom experiment.build import build_utils\n+from experiment.build import docker_images\n+from experiment.build import generate_cloudbuild\nBUILDER_STEP_IDS = [\n'build-fuzzer-builder',\n@@ -39,31 +40,40 @@ GCB_MACHINE_TYPE = 'n1-highcpu-8'\nlogger = logs.Logger('builder') # pylint: disable=invalid-name\n+def _get_buildable_images(fuzzer=None, benchmark=None):\n+ return docker_images.get_images_to_build([fuzzer], [benchmark])\n+\n+\ndef build_base_images():\n\"\"\"Build base images on GCB.\"\"\"\n- _build(get_build_config_file('base-images.yaml'), 'base-images')\n+ image_templates = {'base-image': _get_buildable_images()['base-image']}\n+ config = generate_cloudbuild.create_cloud_build_spec(image_templates,\n+ build_base_images=True)\n+ _build(config, 'base-images')\ndef build_coverage(benchmark):\n\"\"\"Build coverage image for benchmark on GCB.\"\"\"\n- coverage_binaries_dir = exp_path.filestore(\n- build_utils.get_coverage_binaries_dir())\n- substitutions = {\n- '_GCS_COVERAGE_BINARIES_DIR': coverage_binaries_dir,\n- '_BENCHMARK': benchmark,\n+ buildable_images = _get_buildable_images(benchmark=benchmark)\n+ image_templates = {\n+ image_name: image_specs\n+ for image_name, image_specs in buildable_images.items()\n+ if image_specs['type'] in 'coverage'\n}\n- config_file = get_build_config_file('coverage.yaml')\n+ config = generate_cloudbuild.create_cloud_build_spec(image_templates,\n+ benchmark=benchmark)\nconfig_name = 'benchmark-{benchmark}-coverage'.format(benchmark=benchmark)\n- _build(config_file, config_name, substitutions)\n+ _build(config, config_name)\n-def _build(config_file: str,\n+def _build(config: Dict,\nconfig_name: str,\n- substitutions: Dict[str, str] = None,\ntimeout_seconds: int = GCB_BUILD_TIMEOUT\n) -> new_process.ProcessResult:\n- \"\"\"Build each of |args| on gcb.\"\"\"\n- config_arg = '--config=%s' % config_file\n+ \"\"\"Submit build to GCB.\"\"\"\n+ with tempfile.NamedTemporaryFile() as config_file:\n+ yaml_utils.write(config_file.name, config)\n+ config_arg = '--config=%s' % config_file.name\nmachine_type_arg = '--machine-type=%s' % GCB_MACHINE_TYPE\n# Use \"s\" suffix to denote seconds.\n@@ -79,21 +89,6 @@ def _build(config_file: str,\nmachine_type_arg,\n]\n- if substitutions is None:\n- substitutions = {}\n-\n- assert '_REPO' not in substitutions\n- substitutions['_REPO'] = experiment_utils.get_base_docker_tag()\n-\n- assert '_EXPERIMENT' not in substitutions\n- substitutions['_EXPERIMENT'] = experiment_utils.get_experiment_name()\n-\n- substitutions = [\n- '%s=%s' % (key, value) for key, value in substitutions.items()\n- ]\n- substitutions = ','.join(substitutions)\n- command.append('--substitutions=%s' % substitutions)\n-\n# Don't write to stdout to make concurrent building faster. Otherwise\n# writing becomes the bottleneck.\nresult = new_process.execute(command,\n@@ -104,19 +99,16 @@ def _build(config_file: str,\nreturn result\n-def get_build_config_file(filename: str) -> str:\n- \"\"\"Return the path of the GCB build config file |filename|.\"\"\"\n- return os.path.join(utils.ROOT_DIR, 'docker', 'gcb', filename)\n-\n-\n-def build_fuzzer_benchmark(fuzzer: str, benchmark: str) -> bool:\n+def build_fuzzer_benchmark(fuzzer: str, benchmark: str):\n\"\"\"Builds |benchmark| for |fuzzer|.\"\"\"\n- substitutions = {\n- '_BENCHMARK': benchmark,\n- '_FUZZER': fuzzer,\n- }\n- config_file = get_build_config_file('fuzzer.yaml')\n+ image_templates = {}\n+ buildable_images = _get_buildable_images(fuzzer=fuzzer, benchmark=benchmark)\n+ for image_name, image_specs in buildable_images.items():\n+ if image_specs['type'] in ('base', 'coverage', 'dispatcher'):\n+ continue\n+ image_templates[image_name] = image_specs\n+ config = generate_cloudbuild.create_cloud_build_spec(image_templates)\nconfig_name = 'benchmark-{benchmark}-fuzzer-{fuzzer}'.format(\nbenchmark=benchmark, fuzzer=fuzzer)\n- _build(config_file, config_name, substitutions)\n+ _build(config, config_name)\n" }, { "change_type": "MODIFY", "old_path": "experiment/build/generate_cloudbuild.py", "new_path": "experiment/build/generate_cloudbuild.py", "diff": "# limitations under the License.\n\"\"\"Generates Cloud Build specification\"\"\"\n-import argparse\n-import json\n-import yaml\n-\n-from experiment.build import docker_images\n-\n-\n-# TODO(Tanq16): Add unit test for this.\n-def create_cloud_build_spec(buildable_images, docker_registry):\n- \"\"\"Returns Cloud Build specificatiion.\"\"\"\n-\n- cloud_build_spec = {}\n- cloud_build_spec['steps'] = []\n- cloud_build_spec['images'] = []\n-\n- for name, image in buildable_images.items():\n- step = {}\n- step['id'] = name\n- step['name'] = 'gcr.io/cloud-builders/docker'\n- step['args'] = []\n- step['args'] += ['--tag', image['tag']]\n- step['args'] += ['--cache-from', docker_registry + image['tag']]\n- step['args'] += ['--build-arg', 'BUILDKIT_INLINE_CACHE=1']\n- if 'build_arg' in image:\n- for build_arg in image['build_arg']:\n+import os\n+import posixpath\n+\n+from common import yaml_utils\n+from common import experiment_utils\n+from common import experiment_path as exp_path\n+from common.utils import ROOT_DIR\n+from experiment.build import build_utils\n+\n+DOCKER_REGISTRY = 'gcr.io/fuzzbench'\n+\n+\n+def get_experiment_tag_for_image(image_specs, tag_by_experiment=True):\n+ \"\"\"Returns the registry with the experiment tag for given image.\"\"\"\n+ tag = posixpath.join(experiment_utils.get_base_docker_tag(),\n+ image_specs['tag'])\n+ if tag_by_experiment:\n+ tag += ':' + experiment_utils.get_experiment_name()\n+ return tag\n+\n+\n+def coverage_steps(benchmark):\n+ \"\"\"Returns GCB run steps for coverage builds.\"\"\"\n+ coverage_binaries_dir = exp_path.filestore(\n+ build_utils.get_coverage_binaries_dir())\n+ steps = [{\n+ 'name':\n+ 'gcr.io/cloud-builders/docker',\n+ 'args': [\n+ 'run', '-v', '/workspace/out:/host-out',\n+ posixpath.join(experiment_utils.get_base_docker_tag(), 'builders',\n+ 'coverage', benchmark) + ':' +\n+ experiment_utils.get_experiment_name(), '/bin/bash', '-c',\n+ 'cd /out; tar -czvf /host-out/coverage-build-' + benchmark +\n+ '.tar.gz *'\n+ ]\n+ }]\n+ step = {'name': 'gcr.io/cloud-builders/gsutil'}\n+ step['args'] = [\n+ '-m', 'cp', '/workspace/out/coverage-build-' + benchmark + '.tar.gz',\n+ coverage_binaries_dir + '/'\n+ ]\n+ steps.append(step)\n+ return steps\n+\n+\n+def create_cloud_build_spec(image_templates,\n+ benchmark='',\n+ build_base_images=False):\n+ \"\"\"Generates Cloud Build specification.\n+\n+ Args:\n+ image_templates: Image types and their properties.\n+ benchmark: Name of benchmark (required for coverage builds only).\n+ build_base_images: True if building only base images.\n+\n+ Returns:\n+ GCB build steps.\n+ \"\"\"\n+ cloud_build_spec = {'steps': [], 'images': []}\n+\n+ for image_name, image_specs in image_templates.items():\n+ step = {\n+ 'id': image_name,\n+ 'env': 'DOCKER_BUILDKIT=1',\n+ 'name': 'gcr.io/cloud-builders/docker'\n+ }\n+ step['args'] = [\n+ 'build', '--tag',\n+ posixpath.join(DOCKER_REGISTRY, image_specs['tag']), '--tag',\n+ get_experiment_tag_for_image(image_specs), '--cache-from',\n+ get_experiment_tag_for_image(image_specs, tag_by_experiment=False),\n+ '--build-arg', 'BUILDKIT_INLINE_CACHE=1'\n+ ]\n+ for build_arg in image_specs.get('build_arg', []):\nstep['args'] += ['--build-arg', build_arg]\n- if 'dockerfile' in image:\n- step['args'] += ['--file', image['dockerfile']]\n- step['args'] += [image['context']]\n- if 'depends_on' in image:\n+\n+ step['args'] += [\n+ '--file', image_specs['dockerfile'], image_specs['context']\n+ ]\nstep['wait_for'] = []\n- for dep in image['depends_on']:\n- step['wait_for'] += [dep]\n- cloud_build_spec['images'].append(name)\n+ for dependency in image_specs.get('depends_on', []):\n+ # Base images are built before creating fuzzer benchmark builds,\n+ # so it's not required to wait for them to build.\n+ if 'base' in dependency and not build_base_images:\n+ continue\n+ step['wait_for'] += [dependency]\n+\ncloud_build_spec['steps'].append(step)\n+ cloud_build_spec['images'].append(\n+ get_experiment_tag_for_image(image_specs))\n+ cloud_build_spec['images'].append(\n+ get_experiment_tag_for_image(image_specs, tag_by_experiment=False))\n+\n+ if any(image_specs['type'] in 'coverage'\n+ for _, image_specs in image_templates.items()):\n+ cloud_build_spec['steps'] += coverage_steps(benchmark)\nreturn cloud_build_spec\ndef main():\n- \"\"\"Generates Cloud Build specification.\"\"\"\n- parser = argparse.ArgumentParser(description='GCB spec generator.')\n- parser.add_argument('-r',\n- '--docker-registry',\n- default='gcr.io/fuzzbench/',\n- help='Docker registry to use.')\n- args = parser.parse_args()\n-\n- # TODO(Tanq16): Create fuzzer/benchmark list dynamically.\n- fuzzers = ['afl', 'libfuzzer']\n- benchmarks = ['libxml', 'libpng']\n- buildable_images = docker_images.get_images_to_build(fuzzers, benchmarks)\n- cloud_build_spec = create_cloud_build_spec(buildable_images,\n- args.docker_registry)\n- # Build spec can be yaml or json, use whichever:\n- # https://cloud.google.com/cloud-build/docs/configuring-builds/create-basic-configuration\n- print(yaml.dump(cloud_build_spec))\n- print(json.dumps(cloud_build_spec))\n+ \"\"\"Write base-images build spec when run from command line.\"\"\"\n+ image_templates = yaml_utils.read(\n+ os.path.join(ROOT_DIR, 'docker', 'image_types.yaml'))\n+ base_images_spec = create_cloud_build_spec(\n+ {'base-image': image_templates['base-image']}, build_base_images=True)\n+ base_images_spec_file = os.path.join(ROOT_DIR, 'docker', 'gcb',\n+ 'base-images.yaml')\n+ yaml_utils.write(base_images_spec_file, base_images_spec)\nif __name__ == '__main__':\n" }, { "change_type": "ADD", "old_path": null, "new_path": "experiment/build/test_generate_cloudbuild.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Tests for generate_cloudbuild.py.\"\"\"\n+\n+import os\n+\n+from unittest.mock import patch\n+\n+from experiment.build import generate_cloudbuild\n+\n+\[email protected](os.environ, {\n+ 'CLOUD_PROJECT': 'fuzzbench',\n+ 'EXPERIMENT': 'test-experiment'\n+})\n+def test_generate_cloud_build_spec():\n+ \"\"\"Tests result of a makefile generation for an image.\"\"\"\n+\n+ image = {\n+ 'afl-zlib-builder-intermediate': {\n+ 'build_arg': [\n+ 'parent_image=gcr.io/fuzzbench/builders/benchmark/zlib'\n+ ],\n+ 'depends_on': ['zlib-project-builder'],\n+ 'dockerfile': 'fuzzers/afl/builder.Dockerfile',\n+ 'context': 'fuzzers/afl',\n+ 'tag': 'builders/afl/zlib-intermediate',\n+ 'type': 'builder'\n+ }\n+ }\n+\n+ generated_spec = generate_cloudbuild.create_cloud_build_spec(image)\n+\n+ expected_spec = {\n+ 'steps': [{\n+ 'id': 'afl-zlib-builder-intermediate',\n+ 'env': 'DOCKER_BUILDKIT=1',\n+ 'name': 'gcr.io/cloud-builders/docker',\n+ 'args': [\n+ 'build', '--tag',\n+ 'gcr.io/fuzzbench/builders/afl/zlib-intermediate', '--tag',\n+ 'gcr.io/fuzzbench/builders/afl/zlib-intermediate'\n+ ':test-experiment', '--cache-from',\n+ 'gcr.io/fuzzbench/builders/afl/zlib-intermediate',\n+ '--build-arg', 'BUILDKIT_INLINE_CACHE=1', '--build-arg',\n+ 'parent_image=gcr.io/fuzzbench/builders/benchmark/zlib',\n+ '--file', 'fuzzers/afl/builder.Dockerfile', 'fuzzers/afl'\n+ ],\n+ 'wait_for': ['zlib-project-builder']\n+ }],\n+ 'images': [\n+ 'gcr.io/fuzzbench/builders/afl/zlib-intermediate:test-experiment',\n+ 'gcr.io/fuzzbench/builders/afl/zlib-intermediate'\n+ ]\n+ }\n+\n+ assert generated_spec == expected_spec\n" } ]

[ { "change_type": "MODIFY", "old_path": "docker/benchmark-runner/Dockerfile", "new_path": "docker/benchmark-runner/Dockerfile", "diff": "@@ -54,13 +54,15 @@ ENV WORKDIR /out\nRUN mkdir -p $WORKDIR\nWORKDIR $WORKDIR\n+# The argument needs to be re-declared otherwise it returns an empty string.\n+ARG fuzzer\n# Copy over all the build artifacts (without * to preserve directory structure).\n# This also copies seed and dictionary files if they are available.\nCOPY --from=builder /out/ ./\n# Copy the fuzzer.py file.\n-COPY --from=builder /src/fuzzer.py .\n+COPY fuzzers/$fuzzer/fuzzer.py .\n# Copy the fuzzers directory.\n-COPY --from=builder /src/fuzzers fuzzers\n+COPY fuzzers/ ./fuzzers\n# Create empty __init__.py to allow python deps to work.\nRUN touch __init__.py\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/fuzzers.yml", "new_path": ".github/workflows/fuzzers.yml", "diff": "@@ -46,8 +46,10 @@ jobs:\n- aflplusplus_seek\n- libfuzzer_interceptors\n- libfuzzer_keepseed\n+ - libfuzzer_fixcrossover\n- entropic_interceptors\n- entropic_keepseed\n+ - entropic_fixcrossover\nbenchmark_type:\n- oss-fuzz\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_fixcrossover/builder.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+ARG parent_image\n+FROM $parent_image\n+\n+COPY patch.diff /\n+\n+RUN git clone https://github.com/llvm/llvm-project.git /llvm-project && \\\n+ cd /llvm-project && \\\n+ git checkout b52b2e1c188072e3cbc91500cfd503fb26d50ffc && \\\n+ patch -p1 < /patch.diff && \\\n+ cd /llvm-project/compiler-rt/lib/fuzzer && \\\n+ (for f in *.cpp; do \\\n+ clang++ -stdlib=libc++ -fPIC -O2 -std=c++11 $f -c & \\\n+ done && wait) && \\\n+ ar r /libEntropic.a *.o\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_fixcrossover/fuzzer.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Integration code for libFuzzer fuzzer.\"\"\"\n+\n+from fuzzers.entropic_interceptors import fuzzer as entropic_fuzzer\n+from fuzzers.libfuzzer_interceptors import fuzzer as libfuzzer_fuzzer\n+\n+\n+def build():\n+ \"\"\"Build benchmark.\"\"\"\n+ entropic_fuzzer.build()\n+\n+\n+def fuzz(input_corpus, output_corpus, target_binary):\n+ \"\"\"Run fuzzer.\"\"\"\n+ libfuzzer_fuzzer.run_fuzzer(input_corpus,\n+ output_corpus,\n+ target_binary,\n+ extra_flags=['-entropic=1'])\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_fixcrossover/patch.diff", "diff": "+commit df6a841a7c8b1c540f0e3c31c3697a11b2a00f51\n+Author: Dokyung Song <[email protected]>\n+Date: Wed Aug 5 23:12:19 2020 +0000\n+\n+ [libFuzzer] Fix arguments of InsertPartOf/CopyPartOf calls in CrossOver mutator.\n+\n+ The CrossOver mutator is meant to cross over two given buffers\n+ (referred to as the first/second buffer below). Previously\n+ InsertPartOf/CopyPartOf calls used in the CrossOver mutator\n+ incorrectly inserted/copied part of the second buffer into a \"scratch\n+ buffer\" (MutateInPlaceHere of the size CurrentMaxMutationLen), rather\n+ than the first buffer. This is not intended behavior, because the\n+ scratch buffer does not always (i) contain the content of the first\n+ buffer, and (ii) have the same size as the first buffer;\n+ CurrentMaxMutationLen is typically a lot larger than the size of the\n+ first buffer. This patch fixes the issue by using the first buffer\n+ instead of the scratch buffer in InsertPartOf/CopyPartOf calls.\n+\n+ This patch also adds two new tests, namely \"cross_over_insert\" and\n+ \"cross_over_copy\", which specifically target InsertPartOf and\n+ CopyPartOf, respectively.\n+\n+ - cross_over_insert.test checks if the fuzzer can use InsertPartOf to\n+ trigger the crash.\n+\n+ - cross_over_copy.test checks if the fuzzer can use CopyPartOf to\n+ trigger the crash.\n+\n+ These newly added tests were designed to pass with the current patch,\n+ but not without the it (with b216c80cc2496b87bf827260ce7e24dc62247d71\n+ these tests do no pass). To achieve this, -max_len was intentionally\n+ given a high value. Without this patch, InsertPartOf/CopyPartOf will\n+ generate larger inputs, possibly with unpredictable data in it,\n+ thereby failing to trigger the crash.\n+\n+ The test pass condition for these new tests is narrowed down by (i)\n+ limiting mutation depth to 1 (i.e., a single CrossOver mutation should\n+ be able to trigger the crash) and (ii) checking whether the mutation\n+ sequence of \"CrossOver-\" leads to the crash.\n+\n+ Also note that these newly added tests and an existing test\n+ (cross_over.test) all use \"-reduce_inputs=0\" flags to prevent reducing\n+ inputs; it's easier to force the fuzzer to keep original input string\n+ this way than tweaking cov-instrumented basic blocks in the source\n+ code of the fuzzer executable.\n+\n+ Differential Revision: https://reviews.llvm.org/D85554\n+\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerMutate.cpp b/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n+index 29541eac5dc..df9ada45bb0 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n++++ b/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n+@@ -425,26 +425,26 @@ size_t MutationDispatcher::Mutate_CrossOver(uint8_t *Data, size_t Size,\n+ if (!CrossOverWith) return 0;\n+ const Unit &O = *CrossOverWith;\n+ if (O.empty()) return 0;\n+- MutateInPlaceHere.resize(MaxSize);\n+- auto &U = MutateInPlaceHere;\n+ size_t NewSize = 0;\n+ switch(Rand(3)) {\n+ case 0:\n+- NewSize = CrossOver(Data, Size, O.data(), O.size(), U.data(), U.size());\n++ MutateInPlaceHere.resize(MaxSize);\n++ NewSize = CrossOver(Data, Size, O.data(), O.size(),\n++ MutateInPlaceHere.data(), MaxSize);\n++ memcpy(Data, MutateInPlaceHere.data(), NewSize);\n+ break;\n+ case 1:\n+- NewSize = InsertPartOf(O.data(), O.size(), U.data(), U.size(), MaxSize);\n++ NewSize = InsertPartOf(O.data(), O.size(), Data, Size, MaxSize);\n+ if (!NewSize)\n+- NewSize = CopyPartOf(O.data(), O.size(), U.data(), U.size());\n++ NewSize = CopyPartOf(O.data(), O.size(), Data, Size);\n+ break;\n+ case 2:\n+- NewSize = CopyPartOf(O.data(), O.size(), U.data(), U.size());\n++ NewSize = CopyPartOf(O.data(), O.size(), Data, Size);\n+ break;\n+ default: assert(0);\n+ }\n+ assert(NewSize > 0 && \"CrossOver returned empty unit\");\n+ assert(NewSize <= MaxSize && \"CrossOver returned overisized unit\");\n+- memcpy(Data, U.data(), NewSize);\n+ return NewSize;\n+ }\n+\n+diff --git a/compiler-rt/test/fuzzer/CrossOverTest.cpp b/compiler-rt/test/fuzzer/CrossOverTest.cpp\n+index a7643570a92..3ca53a8a851 100644\n+--- a/compiler-rt/test/fuzzer/CrossOverTest.cpp\n++++ b/compiler-rt/test/fuzzer/CrossOverTest.cpp\n+@@ -4,10 +4,10 @@\n+\n+ // Test for a fuzzer. The fuzzer must find the string\n+ // ABCDEFGHIJ\n+-// We use it as a test for CrossOver functionality\n+-// by passing two inputs to it:\n+-// ABCDE00000\n+-// ZZZZZFGHIJ\n++// We use it as a test for each of CrossOver functionalities\n++// by passing the following sets of two inputs to it:\n++// {ABCDEHIJ, ZFG} to test InsertPartOf\n++// {ABCDE00HIJ, ZFG} to test CopyPartOf\n+ //\n+ #include <assert.h>\n+ #include <cstddef>\n+@@ -16,6 +16,17 @@\n+ #include <iostream>\n+ #include <ostream>\n+\n++#ifndef INPUT_A\n++#define INPUT_A \"ABCDE00000\"\n++#endif\n++\n++#ifndef INPUT_B\n++#define INPUT_B \"ZZZZZFGHIJ\"\n++#endif\n++\n++const char *InputA = INPUT_A;\n++const char *InputB = INPUT_B;\n++\n+ static volatile int Sink;\n+ static volatile int *NullPtr;\n+\n+@@ -42,13 +53,11 @@ static const uint32_t ExpectedHash = 0xe1677acb;\n+\n+ extern \"C\" int LLVMFuzzerTestOneInput(const uint8_t *Data, size_t Size) {\n+ // fprintf(stderr, \"ExpectedHash: %x\\n\", ExpectedHash);\n+- if (Size != 10) return 0;\n+- if (*Data == 'A')\n++ if (Size == 10 && ExpectedHash == simple_hash(Data, Size))\n++ *NullPtr = 0;\n++ if (*Data == InputA[0])\n+ Sink++;\n+- if (*Data == 'Z')\n++ if (*Data == InputB[0])\n+ Sink--;\n+- if (ExpectedHash == simple_hash(Data, Size))\n+- *NullPtr = 0;\n+ return 0;\n+ }\n+-\n+diff --git a/compiler-rt/test/fuzzer/cross_over.test b/compiler-rt/test/fuzzer/cross_over.test\n+index 058b5eb2c85..64e06e8cd36 100644\n+--- a/compiler-rt/test/fuzzer/cross_over.test\n++++ b/compiler-rt/test/fuzzer/cross_over.test\n+@@ -12,7 +12,7 @@ RUN: echo -n ABCDE00000 > %t-corpus/A\n+ RUN: echo -n ZZZZZFGHIJ > %t-corpus/B\n+\n+\n+-RUN: not %run %t-CrossOverTest -max_len=10 -seed=1 -runs=10000000 %t-corpus\n++RUN: not %run %t-CrossOverTest -max_len=10 -reduce_inputs=0 -seed=1 -runs=10000000 %t-corpus\n+\n+ # Test the same thing but using -seed_inputs instead of passing the corpus dir.\n+-RUN: not %run %t-CrossOverTest -max_len=10 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B\n++RUN: not %run %t-CrossOverTest -max_len=10 -reduce_inputs=0 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B\n+diff --git a/compiler-rt/test/fuzzer/cross_over_copy.test b/compiler-rt/test/fuzzer/cross_over_copy.test\n+new file mode 100644\n+index 00000000000..f8f45c974e2\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/cross_over_copy.test\n+@@ -0,0 +1,20 @@\n++# Tests CrossOver CopyPartOf.\n++# We want to make sure that the test can find the input\n++# ABCDEFGHIJ when given two other inputs in the seed corpus:\n++# ABCDE00HIJ and\n++# (Z) FG\n++#\n++RUN: %cpp_compiler -DINPUT_A='\"ABCDE00HIJ\"' -DINPUT_B='\"ZFG\"' %S/CrossOverTest.cpp -o %t-CrossOverTest\n++\n++RUN: rm -rf %t-corpus\n++RUN: mkdir %t-corpus\n++RUN: echo -n ABCDE00HIJ > %t-corpus/A\n++RUN: echo -n ZFG > %t-corpus/B\n++\n++\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 %t-corpus 2>&1 | FileCheck %s\n++\n++# Test the same thing but using -seed_inputs instead of passing the corpus dir.\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B 2>&1 | FileCheck %s\n++\n++CHECK: MS: 1 CrossOver-\n+diff --git a/compiler-rt/test/fuzzer/cross_over_insert.test b/compiler-rt/test/fuzzer/cross_over_insert.test\n+new file mode 100644\n+index 00000000000..5ad2ff0a633\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/cross_over_insert.test\n+@@ -0,0 +1,20 @@\n++# Tests CrossOver InsertPartOf.\n++# We want to make sure that the test can find the input\n++# ABCDEFGHIJ when given two other inputs in the seed corpus:\n++# ABCDE HIJ and\n++# (Z) FG\n++#\n++RUN: %cpp_compiler -DINPUT_A='\"ABCDEHIJ\"' -DINPUT_B='\"ZFG\"' %S/CrossOverTest.cpp -o %t-CrossOverTest\n++\n++RUN: rm -rf %t-corpus\n++RUN: mkdir %t-corpus\n++RUN: echo -n ABCDEHIJ > %t-corpus/A\n++RUN: echo -n ZFG > %t-corpus/B\n++\n++\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 %t-corpus 2>&1 | FileCheck %s\n++\n++# Test the same thing but using -seed_inputs instead of passing the corpus dir.\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B 2>&1 | FileCheck %s\n++\n++CHECK: MS: 1 CrossOver-\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_fixcrossover/runner.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+FROM gcr.io/fuzzbench/base-runner\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/libfuzzer_fixcrossover/builder.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+ARG parent_image\n+FROM $parent_image\n+\n+COPY patch.diff /\n+\n+RUN git clone https://github.com/llvm/llvm-project.git /llvm-project && \\\n+ cd /llvm-project/ && \\\n+ git checkout b52b2e1c188072e3cbc91500cfd503fb26d50ffc && \\\n+ patch -p1 < /patch.diff && \\\n+ cd compiler-rt/lib/fuzzer && \\\n+ (for f in *.cpp; do \\\n+ clang++ -stdlib=libc++ -fPIC -O2 -std=c++11 $f -c & \\\n+ done && wait) && \\\n+ ar r /usr/lib/libFuzzer.a *.o\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/libfuzzer_fixcrossover/fuzzer.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Integration code for libFuzzer fuzzer.\"\"\"\n+\n+from fuzzers.libfuzzer_interceptors import fuzzer as libfuzzer_fuzzer\n+\n+\n+def build():\n+ \"\"\"Build benchmark.\"\"\"\n+ libfuzzer_fuzzer.build()\n+\n+\n+def fuzz(input_corpus, output_corpus, target_binary):\n+ \"\"\"Run fuzzer.\"\"\"\n+ libfuzzer_fuzzer.run_fuzzer(input_corpus, output_corpus, target_binary)\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/libfuzzer_fixcrossover/patch.diff", "diff": "+commit df6a841a7c8b1c540f0e3c31c3697a11b2a00f51\n+Author: Dokyung Song <[email protected]>\n+Date: Wed Aug 5 23:12:19 2020 +0000\n+\n+ [libFuzzer] Fix arguments of InsertPartOf/CopyPartOf calls in CrossOver mutator.\n+\n+ The CrossOver mutator is meant to cross over two given buffers\n+ (referred to as the first/second buffer below). Previously\n+ InsertPartOf/CopyPartOf calls used in the CrossOver mutator\n+ incorrectly inserted/copied part of the second buffer into a \"scratch\n+ buffer\" (MutateInPlaceHere of the size CurrentMaxMutationLen), rather\n+ than the first buffer. This is not intended behavior, because the\n+ scratch buffer does not always (i) contain the content of the first\n+ buffer, and (ii) have the same size as the first buffer;\n+ CurrentMaxMutationLen is typically a lot larger than the size of the\n+ first buffer. This patch fixes the issue by using the first buffer\n+ instead of the scratch buffer in InsertPartOf/CopyPartOf calls.\n+\n+ This patch also adds two new tests, namely \"cross_over_insert\" and\n+ \"cross_over_copy\", which specifically target InsertPartOf and\n+ CopyPartOf, respectively.\n+\n+ - cross_over_insert.test checks if the fuzzer can use InsertPartOf to\n+ trigger the crash.\n+\n+ - cross_over_copy.test checks if the fuzzer can use CopyPartOf to\n+ trigger the crash.\n+\n+ These newly added tests were designed to pass with the current patch,\n+ but not without the it (with b216c80cc2496b87bf827260ce7e24dc62247d71\n+ these tests do no pass). To achieve this, -max_len was intentionally\n+ given a high value. Without this patch, InsertPartOf/CopyPartOf will\n+ generate larger inputs, possibly with unpredictable data in it,\n+ thereby failing to trigger the crash.\n+\n+ The test pass condition for these new tests is narrowed down by (i)\n+ limiting mutation depth to 1 (i.e., a single CrossOver mutation should\n+ be able to trigger the crash) and (ii) checking whether the mutation\n+ sequence of \"CrossOver-\" leads to the crash.\n+\n+ Also note that these newly added tests and an existing test\n+ (cross_over.test) all use \"-reduce_inputs=0\" flags to prevent reducing\n+ inputs; it's easier to force the fuzzer to keep original input string\n+ this way than tweaking cov-instrumented basic blocks in the source\n+ code of the fuzzer executable.\n+\n+ Differential Revision: https://reviews.llvm.org/D85554\n+\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerMutate.cpp b/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n+index 29541eac5dc..df9ada45bb0 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n++++ b/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n+@@ -425,26 +425,26 @@ size_t MutationDispatcher::Mutate_CrossOver(uint8_t *Data, size_t Size,\n+ if (!CrossOverWith) return 0;\n+ const Unit &O = *CrossOverWith;\n+ if (O.empty()) return 0;\n+- MutateInPlaceHere.resize(MaxSize);\n+- auto &U = MutateInPlaceHere;\n+ size_t NewSize = 0;\n+ switch(Rand(3)) {\n+ case 0:\n+- NewSize = CrossOver(Data, Size, O.data(), O.size(), U.data(), U.size());\n++ MutateInPlaceHere.resize(MaxSize);\n++ NewSize = CrossOver(Data, Size, O.data(), O.size(),\n++ MutateInPlaceHere.data(), MaxSize);\n++ memcpy(Data, MutateInPlaceHere.data(), NewSize);\n+ break;\n+ case 1:\n+- NewSize = InsertPartOf(O.data(), O.size(), U.data(), U.size(), MaxSize);\n++ NewSize = InsertPartOf(O.data(), O.size(), Data, Size, MaxSize);\n+ if (!NewSize)\n+- NewSize = CopyPartOf(O.data(), O.size(), U.data(), U.size());\n++ NewSize = CopyPartOf(O.data(), O.size(), Data, Size);\n+ break;\n+ case 2:\n+- NewSize = CopyPartOf(O.data(), O.size(), U.data(), U.size());\n++ NewSize = CopyPartOf(O.data(), O.size(), Data, Size);\n+ break;\n+ default: assert(0);\n+ }\n+ assert(NewSize > 0 && \"CrossOver returned empty unit\");\n+ assert(NewSize <= MaxSize && \"CrossOver returned overisized unit\");\n+- memcpy(Data, U.data(), NewSize);\n+ return NewSize;\n+ }\n+\n+diff --git a/compiler-rt/test/fuzzer/CrossOverTest.cpp b/compiler-rt/test/fuzzer/CrossOverTest.cpp\n+index a7643570a92..3ca53a8a851 100644\n+--- a/compiler-rt/test/fuzzer/CrossOverTest.cpp\n++++ b/compiler-rt/test/fuzzer/CrossOverTest.cpp\n+@@ -4,10 +4,10 @@\n+\n+ // Test for a fuzzer. The fuzzer must find the string\n+ // ABCDEFGHIJ\n+-// We use it as a test for CrossOver functionality\n+-// by passing two inputs to it:\n+-// ABCDE00000\n+-// ZZZZZFGHIJ\n++// We use it as a test for each of CrossOver functionalities\n++// by passing the following sets of two inputs to it:\n++// {ABCDEHIJ, ZFG} to test InsertPartOf\n++// {ABCDE00HIJ, ZFG} to test CopyPartOf\n+ //\n+ #include <assert.h>\n+ #include <cstddef>\n+@@ -16,6 +16,17 @@\n+ #include <iostream>\n+ #include <ostream>\n+\n++#ifndef INPUT_A\n++#define INPUT_A \"ABCDE00000\"\n++#endif\n++\n++#ifndef INPUT_B\n++#define INPUT_B \"ZZZZZFGHIJ\"\n++#endif\n++\n++const char *InputA = INPUT_A;\n++const char *InputB = INPUT_B;\n++\n+ static volatile int Sink;\n+ static volatile int *NullPtr;\n+\n+@@ -42,13 +53,11 @@ static const uint32_t ExpectedHash = 0xe1677acb;\n+\n+ extern \"C\" int LLVMFuzzerTestOneInput(const uint8_t *Data, size_t Size) {\n+ // fprintf(stderr, \"ExpectedHash: %x\\n\", ExpectedHash);\n+- if (Size != 10) return 0;\n+- if (*Data == 'A')\n++ if (Size == 10 && ExpectedHash == simple_hash(Data, Size))\n++ *NullPtr = 0;\n++ if (*Data == InputA[0])\n+ Sink++;\n+- if (*Data == 'Z')\n++ if (*Data == InputB[0])\n+ Sink--;\n+- if (ExpectedHash == simple_hash(Data, Size))\n+- *NullPtr = 0;\n+ return 0;\n+ }\n+-\n+diff --git a/compiler-rt/test/fuzzer/cross_over.test b/compiler-rt/test/fuzzer/cross_over.test\n+index 058b5eb2c85..64e06e8cd36 100644\n+--- a/compiler-rt/test/fuzzer/cross_over.test\n++++ b/compiler-rt/test/fuzzer/cross_over.test\n+@@ -12,7 +12,7 @@ RUN: echo -n ABCDE00000 > %t-corpus/A\n+ RUN: echo -n ZZZZZFGHIJ > %t-corpus/B\n+\n+\n+-RUN: not %run %t-CrossOverTest -max_len=10 -seed=1 -runs=10000000 %t-corpus\n++RUN: not %run %t-CrossOverTest -max_len=10 -reduce_inputs=0 -seed=1 -runs=10000000 %t-corpus\n+\n+ # Test the same thing but using -seed_inputs instead of passing the corpus dir.\n+-RUN: not %run %t-CrossOverTest -max_len=10 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B\n++RUN: not %run %t-CrossOverTest -max_len=10 -reduce_inputs=0 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B\n+diff --git a/compiler-rt/test/fuzzer/cross_over_copy.test b/compiler-rt/test/fuzzer/cross_over_copy.test\n+new file mode 100644\n+index 00000000000..f8f45c974e2\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/cross_over_copy.test\n+@@ -0,0 +1,20 @@\n++# Tests CrossOver CopyPartOf.\n++# We want to make sure that the test can find the input\n++# ABCDEFGHIJ when given two other inputs in the seed corpus:\n++# ABCDE00HIJ and\n++# (Z) FG\n++#\n++RUN: %cpp_compiler -DINPUT_A='\"ABCDE00HIJ\"' -DINPUT_B='\"ZFG\"' %S/CrossOverTest.cpp -o %t-CrossOverTest\n++\n++RUN: rm -rf %t-corpus\n++RUN: mkdir %t-corpus\n++RUN: echo -n ABCDE00HIJ > %t-corpus/A\n++RUN: echo -n ZFG > %t-corpus/B\n++\n++\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 %t-corpus 2>&1 | FileCheck %s\n++\n++# Test the same thing but using -seed_inputs instead of passing the corpus dir.\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B 2>&1 | FileCheck %s\n++\n++CHECK: MS: 1 CrossOver-\n+diff --git a/compiler-rt/test/fuzzer/cross_over_insert.test b/compiler-rt/test/fuzzer/cross_over_insert.test\n+new file mode 100644\n+index 00000000000..5ad2ff0a633\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/cross_over_insert.test\n+@@ -0,0 +1,20 @@\n++# Tests CrossOver InsertPartOf.\n++# We want to make sure that the test can find the input\n++# ABCDEFGHIJ when given two other inputs in the seed corpus:\n++# ABCDE HIJ and\n++# (Z) FG\n++#\n++RUN: %cpp_compiler -DINPUT_A='\"ABCDEHIJ\"' -DINPUT_B='\"ZFG\"' %S/CrossOverTest.cpp -o %t-CrossOverTest\n++\n++RUN: rm -rf %t-corpus\n++RUN: mkdir %t-corpus\n++RUN: echo -n ABCDEHIJ > %t-corpus/A\n++RUN: echo -n ZFG > %t-corpus/B\n++\n++\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 %t-corpus 2>&1 | FileCheck %s\n++\n++# Test the same thing but using -seed_inputs instead of passing the corpus dir.\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B 2>&1 | FileCheck %s\n++\n++CHECK: MS: 1 CrossOver-\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/libfuzzer_fixcrossover/runner.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+FROM gcr.io/fuzzbench/base-runner\n" }, { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-08-08\n+ fuzzers:\n+ - libfuzzer_fixcrossover\n+ - entropic_fixcrossover\n+\n- experiment: 2020-08-07\nfuzzers:\n- aflplusplus_datalenrand\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/presubmit.yml", "new_path": ".github/workflows/presubmit.yml", "diff": "@@ -37,4 +37,7 @@ jobs:\n- name: Run presubmit checks\nrun: |\nFUZZBENCH_TEST_INTEGRATION=1 make presubmit\n- # TODO(zhichengcai): Add back end to end test.\n+\n+ - name: Run end to end CI test\n+ run: |\n+ make run-end-to-end-test\n" }, { "change_type": "ADD", "old_path": null, "new_path": "common/config_utils.py", "diff": "+#!/usr/bin/env python3\n+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Provides helper functions to obtain configurations.\"\"\"\n+\n+\n+def validate_and_expand(config):\n+ \"\"\"Validates |config| and returns the expanded configuration.\"\"\"\n+ # TODO: move the logic from experiment/run_experiment.py to here.\n+ return config\n" }, { "change_type": "MODIFY", "old_path": "compose/e2e-test.yaml", "new_path": "compose/e2e-test.yaml", "diff": "@@ -8,4 +8,8 @@ services:\n- queue-server\nenvironment:\nE2E_INTEGRATION_TEST: 1\n- command: python3 -m pytest -vv fuzzbench/test_e2e_run.py\n+ command: python3 -m pytest -vv fuzzbench/test_e2e/test_e2e_run.py\n+\n+ run-experiment:\n+ environment:\n+ EXPERIMENT_CONFIG: fuzzbench/test_e2e/end-to-end-test-config.yaml\n" }, { "change_type": "MODIFY", "old_path": "docker/fuzzbench/Dockerfile", "new_path": "docker/fuzzbench/Dockerfile", "diff": "@@ -31,6 +31,8 @@ COPY benchmarks benchmarks\nCOPY common common\nCOPY database database\nCOPY docker docker\n+COPY experiment/build experiment/build\n+COPY experiment/*.py experiment/\nCOPY fuzzbench fuzzbench\nCOPY fuzzers fuzzers\n" }, { "change_type": "MODIFY", "old_path": "docker/image_types.yaml", "new_path": "docker/image_types.yaml", "diff": "tag: 'dispatcher-image'\ntype: 'dispatcher'\n+# TODO: It would be better to call this benchmark builder. But that would be\n+# confusing because this doesn't involve benchmark-builder/Dockerfile. Rename\n+# that and then rename this.\n+'{benchmark}-project-builder':\n+ dockerfile: 'benchmarks/{benchmark}/Dockerfile'\n+ context: 'benchmarks/{benchmark}'\n+ tag: 'builders/benchmark/{benchmark}'\n+ type: 'builder'\n+\n'coverage-{benchmark}-builder-intermediate':\nbuild_arg:\n- 'parent_image=gcr.io/fuzzbench/builders/benchmark/{benchmark}'\ntag: 'builders/coverage/{benchmark}'\ntype: 'coverage'\n-# TODO: It would be better to call this benchmark builder. But that would be\n-# confusing because this doesn't involve benchmark-builder/Dockerfile. Rename\n-# that and then rename this.\n-'{benchmark}-project-builder':\n- dockerfile: 'benchmarks/{benchmark}/Dockerfile'\n- context: 'benchmarks/{benchmark}'\n- tag: 'builders/benchmark/{benchmark}'\n- type: 'builder'\n-\n'{fuzzer}-{benchmark}-builder-intermediate':\nbuild_arg:\n- 'parent_image=gcr.io/fuzzbench/builders/benchmark/{benchmark}'\n" }, { "change_type": "MODIFY", "old_path": "fuzzbench/jobs.py", "new_path": "fuzzbench/jobs.py", "diff": "@@ -19,14 +19,19 @@ import subprocess\nBASE_TAG = 'gcr.io/fuzzbench'\n-def build_image(name: str):\n+def build_image(image):\n\"\"\"Builds a Docker image and returns whether it succeeds.\"\"\"\n- image_tag = os.path.join(BASE_TAG, name)\n+ image_tag = os.path.join(BASE_TAG, image['tag'])\nsubprocess.run(['docker', 'pull', image_tag], check=True)\n- subprocess.run(\n- ['docker', 'build', '--tag', image_tag,\n- os.path.join('docker', name)],\n- check=True)\n+ command = ['docker', 'build', '--tag', image_tag, image['context']]\n+ cpu_options = ['--cpu-period', '100000', '--cpu-quota', '100000']\n+ command.extend(cpu_options)\n+ if 'dockerfile' in image:\n+ command.extend(['--file', image['dockerfile']])\n+ if 'build_arg' in image:\n+ for arg in image['build_arg']:\n+ command.extend(['--build-arg', arg])\n+ subprocess.run(command, check=True)\nreturn True\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzbench/local-experiment-config.yaml", "diff": "+benchmarks:\n+ - freetype2-2017\n+ - bloaty_fuzz_target\n+fuzzers:\n+ - afl\n+ - libfuzzer\n" }, { "change_type": "MODIFY", "old_path": "fuzzbench/run_experiment.py", "new_path": "fuzzbench/run_experiment.py", "diff": "@@ -18,20 +18,34 @@ import time\nimport redis\nimport rq\n+from common import config_utils, environment, yaml_utils\n+from experiment.build import docker_images\nfrom fuzzbench import jobs\n-def run_experiment():\n+def run_experiment(config):\n\"\"\"Main experiment logic.\"\"\"\nprint('Initializing the job queue.')\n# Create the queue for scheduling build jobs and run jobs.\nqueue = rq.Queue('build_n_run_queue')\n+\n+ images_to_build = docker_images.get_images_to_build(config['fuzzers'],\n+ config['benchmarks'])\njobs_list = []\n+ # TODO(#643): topological sort before enqueuing jobs.\n+ for name, image in images_to_build.items():\n+ depends = image.get('depends_on', None)\n+ if depends is not None:\n+ assert len(depends) == 1, 'image %s has %d dependencies. Multiple '\\\n+ 'dependencies are currently not supported.' % (name, len(depends))\njobs_list.append(\n- queue.enqueue(jobs.build_image,\n- 'base-image',\n- job_timeout=600,\n- job_id='base-image'))\n+ queue.enqueue(\n+ jobs.build_image,\n+ image=image,\n+ job_timeout=30 * 60,\n+ result_ttl=-1,\n+ job_id=name,\n+ depends_on=depends[0] if 'depends_on' in image else None))\nwhile True:\nprint('Current status of jobs:')\n@@ -52,8 +66,14 @@ def run_experiment():\ndef main():\n\"\"\"Set up Redis connection and start the experiment.\"\"\"\nredis_connection = redis.Redis(host=\"queue-server\")\n+\n+ config_path = environment.get('EXPERIMENT_CONFIG',\n+ 'fuzzbench/local-experiment-config.yaml')\n+ config = yaml_utils.read(config_path)\n+ config = config_utils.validate_and_expand(config)\n+\nwith rq.Connection(redis_connection):\n- return run_experiment()\n+ return run_experiment(config)\nif __name__ == '__main__':\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzbench/test_e2e/end-to-end-test-config.yaml", "diff": "+benchmarks:\n+ - bloaty_fuzz_target\n+fuzzers:\n+ - libfuzzer\n" }, { "change_type": "RENAME", "old_path": "fuzzbench/test_e2e_run.py", "new_path": "fuzzbench/test_e2e/test_e2e_run.py", "diff": "@@ -19,7 +19,17 @@ import os\nimport pytest\nimport redis\n-from rq.job import Job\n+import rq\n+\n+from common import config_utils, yaml_utils\n+from experiment.build import docker_images\n+\n+\[email protected](scope='class')\n+def experiment_config():\n+ \"\"\"Returns the default configuration for end-to-end testing.\"\"\"\n+ return config_utils.validate_and_expand(\n+ yaml_utils.read('fuzzbench/test_e2e/end-to-end-test-config.yaml'))\[email protected](scope='class')\n@@ -31,17 +41,32 @@ def redis_connection():\n# pylint: disable=no-self-use\[email protected]('E2E_INTEGRATION_TEST' not in os.environ,\nreason='Not running end-to-end test.')\[email protected]('redis_connection')\[email protected]('redis_connection', 'experiment_config')\nclass TestEndToEndRunResults:\n\"\"\"Checks the result of a test experiment run.\"\"\"\n- def test_jobs_dependency(self): # pylint: disable=redefined-outer-name\n+ def test_jobs_dependency(self, experiment_config, redis_connection): # pylint: disable=redefined-outer-name\n\"\"\"Tests that jobs dependency preserves during working.\"\"\"\n- assert True\n+ all_images = docker_images.get_images_to_build(\n+ experiment_config['fuzzers'], experiment_config['benchmarks'])\n+ jobs = {\n+ name: rq.job.Job.fetch(name, connection=redis_connection)\n+ for name in all_images\n+ }\n+ for name, image in all_images.items():\n+ if 'depends_on' in image:\n+ for dep in image['depends_on']:\n+ assert jobs[dep].ended_at <= jobs[name].started_at\n- def test_all_jobs_finished_successfully(self, redis_connection): # pylint: disable=redefined-outer-name\n+ def test_all_jobs_finished_successfully(\n+ self,\n+ experiment_config, # pylint: disable=redefined-outer-name\n+ redis_connection): # pylint: disable=redefined-outer-name\n\"\"\"Tests all jobs finished successully.\"\"\"\n- jobs = Job.fetch_many(['base-image'], connection=redis_connection)\n+ all_images = docker_images.get_images_to_build(\n+ experiment_config['fuzzers'], experiment_config['benchmarks'])\n+ jobs = rq.job.Job.fetch_many(all_images.keys(),\n+ connection=redis_connection)\nfor job in jobs:\nassert job.get_status() == 'finished'\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Gemfile.lock", "new_path": "docs/Gemfile.lock", "diff": "@@ -67,7 +67,7 @@ GEM\njekyll-theme-time-machine (= 0.1.1)\njekyll-titles-from-headings (= 0.5.3)\njemoji (= 0.11.1)\n- kramdown (= 1.17.0)\n+ kramdown (= 2.3.0)\nliquid (= 4.0.3)\nmercenary (~> 0.3)\nminima (= 2.5.1)\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/Gemfile.lock", "new_path": "docs/Gemfile.lock", "diff": "GEM\nremote: https://rubygems.org/\nspecs:\n- activesupport (6.0.3.1)\n+ activesupport (6.0.3.2)\nconcurrent-ruby (~> 1.0, >= 1.0.2)\ni18n (>= 0.7, < 2)\nminitest (~> 5.1)\n@@ -16,9 +16,9 @@ GEM\ncolorator (1.1.0)\ncommonmarker (0.17.13)\nruby-enum (~> 0.5)\n- concurrent-ruby (1.1.6)\n- dnsruby (1.61.3)\n- addressable (~> 2.5)\n+ concurrent-ruby (1.1.7)\n+ dnsruby (1.61.4)\n+ simpleidn (~> 0.1)\nem-websocket (0.5.1)\neventmachine (>= 0.12.9)\nhttp_parser.rb (~> 0.6.0)\n@@ -26,14 +26,14 @@ GEM\nffi (>= 1.3.0)\neventmachine (1.2.7)\nexecjs (2.7.0)\n- faraday (1.0.0)\n+ faraday (1.0.1)\nmultipart-post (>= 1.2, < 3)\n- ffi (1.12.1)\n+ ffi (1.13.1)\nforwardable-extended (2.6.0)\ngemoji (3.0.1)\n- github-pages (204)\n+ github-pages (207)\ngithub-pages-health-check (= 1.16.1)\n- jekyll (= 3.8.5)\n+ jekyll (= 3.9.0)\njekyll-avatar (= 0.7.0)\njekyll-coffeescript (= 1.1.1)\njekyll-commonmark-ghpages (= 0.1.6)\n@@ -68,11 +68,12 @@ GEM\njekyll-titles-from-headings (= 0.5.3)\njemoji (= 0.11.1)\nkramdown (= 2.3.0)\n+ kramdown-parser-gfm (= 1.1.0)\nliquid (= 4.0.3)\nmercenary (~> 0.3)\nminima (= 2.5.1)\nnokogiri (>= 1.10.4, < 2.0)\n- rouge (= 3.13.0)\n+ rouge (= 3.19.0)\nterminal-table (~> 1.4)\ngithub-pages-health-check (1.16.1)\naddressable (~> 2.3)\n@@ -80,20 +81,20 @@ GEM\noctokit (~> 4.0)\npublic_suffix (~> 3.0)\ntyphoeus (~> 1.3)\n- html-pipeline (2.12.3)\n+ html-pipeline (2.13.0)\nactivesupport (>= 2)\nnokogiri (>= 1.4)\nhttp_parser.rb (0.6.0)\ni18n (0.9.5)\nconcurrent-ruby (~> 1.0)\n- jekyll (3.8.5)\n+ jekyll (3.9.0)\naddressable (~> 2.4)\ncolorator (~> 1.0)\nem-websocket (~> 0.5)\ni18n (~> 0.7)\njekyll-sass-converter (~> 1.0)\njekyll-watch (~> 2.0)\n- kramdown (~> 1.14)\n+ kramdown (>= 1.17, < 3)\nliquid (~> 4.0)\nmercenary (~> 0.3.3)\npathutil (~> 0.9)\n@@ -191,7 +192,10 @@ GEM\ngemoji (~> 3.0)\nhtml-pipeline (~> 2.2)\njekyll (>= 3.0, < 5.0)\n- kramdown (1.17.0)\n+ kramdown (2.3.0)\n+ rexml\n+ kramdown-parser-gfm (1.1.0)\n+ kramdown (~> 2.0)\nliquid (4.0.3)\nlisten (3.2.1)\nrb-fsevent (~> 0.10, >= 0.10.3)\n@@ -204,21 +208,22 @@ GEM\njekyll-seo-tag (~> 2.1)\nminitest (5.14.1)\nmultipart-post (2.1.1)\n- nokogiri (1.10.9)\n+ nokogiri (1.10.10)\nmini_portile2 (~> 2.4.0)\n- octokit (4.15.0)\n+ octokit (4.18.0)\nfaraday (>= 0.9)\nsawyer (~> 0.8.0, >= 0.5.3)\npathutil (0.16.2)\nforwardable-extended (~> 2.6)\npublic_suffix (3.1.1)\n- rb-fsevent (0.10.3)\n+ rb-fsevent (0.10.4)\nrb-inotify (0.10.1)\nffi (~> 1.0)\n- rouge (3.13.0)\n- ruby-enum (0.7.2)\n+ rexml (3.2.4)\n+ rouge (3.19.0)\n+ ruby-enum (0.8.0)\ni18n\n- rubyzip (2.1.0)\n+ rubyzip (2.3.0)\nsafe_yaml (1.0.5)\nsass (3.7.4)\nsass-listen (~> 4.0.0)\n@@ -228,15 +233,20 @@ GEM\nsawyer (0.8.2)\naddressable (>= 2.3.5)\nfaraday (> 0.8, < 2.0)\n+ simpleidn (0.1.1)\n+ unf (~> 0.1.4)\nterminal-table (1.8.0)\nunicode-display_width (~> 1.1, >= 1.1.1)\nthread_safe (0.3.6)\n- typhoeus (1.3.1)\n+ typhoeus (1.4.0)\nethon (>= 0.9.0)\ntzinfo (1.2.7)\nthread_safe (~> 0.1)\n- unicode-display_width (1.6.1)\n- zeitwerk (2.3.0)\n+ unf (0.1.4)\n+ unf_ext\n+ unf_ext (0.0.7.7)\n+ unicode-display_width (1.7.0)\n+ zeitwerk (2.4.0)\nPLATFORMS\nruby\n" } ]

[ { "change_type": "MODIFY", "old_path": "analysis/benchmark_results.py", "new_path": "analysis/benchmark_results.py", "diff": "@@ -58,6 +58,12 @@ class BenchmarkResults: # pylint: disable=too-many-public-methods\ndef _benchmark_snapshot_df(self):\nreturn data_utils.get_benchmark_snapshot(self._benchmark_df)\n+ @property\n+ @functools.lru_cache()\n+ def fuzzers(self):\n+ \"\"\"Fuzzers with valid trials on this benchmark.\"\"\"\n+ return self._benchmark_df.fuzzer.unique()\n+\n@property\ndef fuzzers_with_not_enough_samples(self):\n\"\"\"Fuzzers with not enough samples.\"\"\"\n" }, { "change_type": "MODIFY", "old_path": "analysis/generate_report.py", "new_path": "analysis/generate_report.py", "diff": "@@ -27,6 +27,8 @@ from analysis import rendering\nfrom common import filesystem\nfrom common import logs\n+logger = logs.Logger('generate_report')\n+\ndef get_arg_parser():\n\"\"\"Returns argument parser.\"\"\"\n@@ -74,6 +76,11 @@ def get_arg_parser():\n'--fuzzers',\nnargs='*',\nhelp='Names of the fuzzers to include in the report.')\n+ parser.add_argument('-cov',\n+ '--coverage-report',\n+ action='store_true',\n+ default=False,\n+ help='If set, clang coverage reports are linked.')\n# It doesn't make sense to clobber and label by experiment, since nothing\n# can get clobbered like this.\n@@ -130,7 +137,8 @@ def generate_report(experiment_names,\nin_progress=False,\nend_time=None,\nmerge_with_clobber=False,\n- merge_with_clobber_nonprivate=False):\n+ merge_with_clobber_nonprivate=False,\n+ coverage_report=False):\n\"\"\"Generate report helper.\"\"\"\nif merge_with_clobber_nonprivate:\nexperiment_names = (\n@@ -174,7 +182,7 @@ def generate_report(experiment_names,\ntemplate = report_type + '.html'\ndetailed_report = rendering.render_report(experiment_ctx, template,\n- in_progress)\n+ in_progress, coverage_report)\nfilesystem.write(os.path.join(report_directory, 'index.html'),\ndetailed_report)\n@@ -198,7 +206,8 @@ def main():\nlog_scale=args.log_scale,\nfrom_cached_data=args.from_cached_data,\nend_time=args.end_time,\n- merge_with_clobber=args.merge_with_clobber)\n+ merge_with_clobber=args.merge_with_clobber,\n+ coverage_report=args.coverage_report)\nif __name__ == '__main__':\n" }, { "change_type": "MODIFY", "old_path": "analysis/rendering.py", "new_path": "analysis/rendering.py", "diff": "@@ -20,7 +20,7 @@ import jinja2\nfrom common import utils\n-def render_report(experiment_results, template, in_progress):\n+def render_report(experiment_results, template, in_progress, coverage_report):\n\"\"\"Renders report with |template| using data provided by the\n|experiment_results| context.\n@@ -37,5 +37,17 @@ def render_report(experiment_results, template, in_progress):\nloader=jinja2.FileSystemLoader(templates_dir),\n)\ntemplate = environment.get_template(template)\n+\n+ # FIXME: Use |experiment_filestore_name| from experiment db.\n+ # See #642: https://github.com/google/fuzzbench/issues/642\n+ if 'EXPERIMENT_FILESTORE' in os.environ:\n+ experiment_filestore = os.environ['EXPERIMENT_FILESTORE']\n+ prefix = \"gs://\"\n+ experiment_filestore_name = experiment_filestore[len(prefix):]\n+ else:\n+ experiment_filestore_name = 'fuzzbench-data'\n+\nreturn template.render(experiment=experiment_results,\n- in_progress=in_progress)\n+ in_progress=in_progress,\n+ coverage_report=coverage_report,\n+ experiment_filestore_name=experiment_filestore_name)\n" }, { "change_type": "MODIFY", "old_path": "analysis/report_templates/default.html", "new_path": "analysis/report_templates/default.html", "diff": "</li>\n</ul>\n+ {% if coverage_report %}\n+ <ul class=\"collapsible\">\n+ <li>\n+ <div class=\"collapsible-header\">\n+ Coverage reports for each fuzzer on this benchmark\n+ </div>\n+ <div class=\"collapsible-body\">\n+ <div class=\"row\">\n+ {% for fuzzer in benchmark.fuzzers %}\n+ <div class=\"col\">\n+ <a class=\"waves-effect waves-light btn\" href=\"https://storage.googleapis.com/{{ experiment_filestore_name }}/{{ experiment.name }}/reports/coverage/{{ benchmark.name }}/{{ fuzzer }}/index.html\">{{ fuzzer }}</a>\n+ </div>\n+ {% endfor %}\n+ </div>\n+ </div>\n+ </li>\n+ </ul>\n+ {% endif %}\n</div> <!-- id=\"{{ benchmark.name }}\" -->\n{% endfor %}\n" }, { "change_type": "MODIFY", "old_path": "common/experiment_utils.py", "new_path": "common/experiment_utils.py", "diff": "@@ -17,6 +17,7 @@ import os\nimport posixpath\nfrom common import environment\n+from common import experiment_path as exp_path\nDEFAULT_SNAPSHOT_SECONDS = 15 * 60 # Seconds.\nCONFIG_DIR = 'config'\n@@ -38,6 +39,11 @@ def get_experiment_name():\nreturn os.environ['EXPERIMENT']\n+def get_experiment_folders_dir():\n+ \"\"\"Returns experiment folders directory.\"\"\"\n+ return exp_path.path('experiment-folders')\n+\n+\ndef get_cloud_project():\n\"\"\"Returns the cloud project.\"\"\"\nreturn os.environ['CLOUD_PROJECT']\n@@ -88,11 +94,16 @@ def is_local_experiment():\ndef get_trial_dir(fuzzer, benchmark, trial_id):\n\"\"\"Returns the unique directory for |fuzzer|, |benchmark|, and\n|trial_id|.\"\"\"\n- benchmark_fuzzer_directory = '%s-%s' % (benchmark, fuzzer)\n+ benchmark_fuzzer_directory = get_benchmark_fuzzer_dir(benchmark, fuzzer)\ntrial_subdir = 'trial-%d' % trial_id\nreturn posixpath.join(benchmark_fuzzer_directory, trial_subdir)\n+def get_benchmark_fuzzer_dir(benchmark, fuzzer):\n+ \"\"\"Returns the directory for |benchmark| and |fuzzer|.\"\"\"\n+ return '%s-%s' % (benchmark, fuzzer)\n+\n+\ndef get_trial_bucket_dir(fuzzer, benchmark, trial_id):\n\"\"\"Returns the unique directory in experiment-folders int the bucket for\n|fuzzer|, |benchmark|, and |trial_id|.\"\"\"\n" }, { "change_type": "ADD", "old_path": null, "new_path": "experiment/coverage_utils.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Utility functions for coverage report generation.\"\"\"\n+\n+import os\n+import multiprocessing\n+import json\n+import queue\n+\n+from common import experiment_utils as exp_utils\n+from common import new_process\n+from common import benchmark_utils\n+from common import fuzzer_utils\n+from common import logs\n+from common import filestore_utils\n+from common import experiment_path as exp_path\n+from common import filesystem\n+from database import utils as db_utils\n+from database import models\n+from experiment.build import build_utils\n+from experiment import reporter\n+\n+logger = logs.Logger('coverage_utils') # pylint: disable=invalid-name\n+\n+COV_DIFF_QUEUE_GET_TIMEOUT = 1\n+\n+\n+def upload_coverage_reports_to_bucket():\n+ \"\"\"Copies the coverage reports to gcs bucket.\"\"\"\n+ report_dir = reporter.get_reports_dir()\n+ src_dir = get_coverage_report_dir()\n+ dst_dir = exp_path.filestore(report_dir)\n+ filestore_utils.cp(src_dir, dst_dir, recursive=True, parallel=True)\n+\n+\n+def generate_coverage_reports(experiment_config: dict):\n+ \"\"\"Generates coverage reports for each benchmark and fuzzer.\"\"\"\n+ logger.info('Start generating coverage report for benchmarks.')\n+ benchmarks = experiment_config['benchmarks'].split(',')\n+ fuzzers = experiment_config['fuzzers'].split(',')\n+ experiment = experiment_config['experiment']\n+ with multiprocessing.Pool() as pool:\n+ generate_coverage_report_args = [(experiment, benchmark, fuzzer)\n+ for benchmark in benchmarks\n+ for fuzzer in fuzzers]\n+ pool.starmap(generate_coverage_report, generate_coverage_report_args)\n+ pool.close()\n+ pool.join()\n+ logger.info('Finished generating coverage report.')\n+\n+\n+def generate_coverage_report(experiment, benchmark, fuzzer):\n+ \"\"\"Generates the coverage report for one pair of benchmark and fuzzer.\"\"\"\n+ logs.initialize()\n+ logger.info('Generating coverage report for benchmark: {benchmark} \\\n+ fuzzer: {fuzzer}.'.format(benchmark=benchmark, fuzzer=fuzzer))\n+ generator = CoverageReporter(fuzzer, benchmark, experiment)\n+ # Merges all the profdata files.\n+ generator.merge_profdata_files()\n+ # Generates the reports using llvm-cov.\n+ generator.generate_cov_report()\n+\n+ logger.info('Finished generating coverage report for '\n+ 'benchmark:{benchmark} fuzzer:{fuzzer}.'.format(\n+ benchmark=benchmark, fuzzer=fuzzer))\n+\n+\n+class CoverageReporter: # pylint: disable=too-many-instance-attributes\n+ \"\"\"Class used to generate coverage report for a pair of\n+ fuzzer and benchmark.\"\"\"\n+\n+ # pylint: disable=too-many-arguments\n+ def __init__(self, fuzzer, benchmark, experiment):\n+ self.fuzzer = fuzzer\n+ self.benchmark = benchmark\n+ self.experiment = experiment\n+ self.trial_ids = get_trial_ids(experiment, fuzzer, benchmark)\n+ cov_report_directory = get_coverage_report_dir()\n+ self.report_dir = os.path.join(cov_report_directory, benchmark, fuzzer)\n+ benchmark_fuzzer_dir = exp_utils.get_benchmark_fuzzer_dir(\n+ benchmark, fuzzer)\n+ work_dir = exp_utils.get_work_dir()\n+ self.merged_profdata_file = os.path.join(work_dir,\n+ 'measurement-folders',\n+ benchmark_fuzzer_dir,\n+ 'merged.profdata')\n+ coverage_binaries_dir = build_utils.get_coverage_binaries_dir()\n+ self.source_files = os.path.join(coverage_binaries_dir, benchmark,\n+ 'src')\n+ self.binary_file = get_coverage_binary(benchmark)\n+\n+ def merge_profdata_files(self):\n+ \"\"\"Merge profdata files from |src_files| to |dst_files|.\"\"\"\n+ logger.info('Merging profdata for fuzzer: '\n+ '{fuzzer},benchmark: {benchmark}.'.format(\n+ fuzzer=self.fuzzer, benchmark=self.benchmark))\n+ files_to_merge = [\n+ TrialCoverage(self.fuzzer, self.benchmark, trial_id,\n+ logger).profdata_file for trial_id in self.trial_ids\n+ ]\n+ result = merge_profdata_files(files_to_merge, self.merged_profdata_file)\n+ if result.retcode != 0:\n+ logger.error('Profdata files merging failed.')\n+\n+ def generate_cov_report(self):\n+ \"\"\"Generates the coverage report.\"\"\"\n+ command = [\n+ 'llvm-cov', 'show', '-format=html',\n+ '-path-equivalence=/,{prefix}'.format(prefix=self.source_files),\n+ '-output-dir={dst_dir}'.format(dst_dir=self.report_dir),\n+ '-Xdemangler', 'c++filt', '-Xdemangler', '-n', self.binary_file,\n+ '-instr-profile={profdata}'.format(\n+ profdata=self.merged_profdata_file)\n+ ]\n+ result = new_process.execute(command)\n+ if result.retcode != 0:\n+ logger.error('Coverage report generation failed for '\n+ 'fuzzer: {fuzzer},benchmark: {benchmark}.'.format(\n+ fuzzer=self.fuzzer, benchmark=self.benchmark))\n+\n+\n+def get_coverage_archive_name(benchmark):\n+ \"\"\"Gets the archive name for |benchmark|.\"\"\"\n+ return 'coverage-build-%s.tar.gz' % benchmark\n+\n+\n+def get_fuzzer_benchmark_key(fuzzer: str, benchmark: str):\n+ \"\"\"Returns the key in coverage dict for a pair of fuzzer-benchmark.\"\"\"\n+ return fuzzer + ' ' + benchmark\n+\n+\n+def get_profdata_file_name(trial_id):\n+ \"\"\"Returns the profdata file name for |trial_id|.\"\"\"\n+ return 'data-{id}.profdata'.format(id=trial_id)\n+\n+\n+def get_coverage_report_dir():\n+ \"\"\"Returns the directory to store all the coverage reports.\"\"\"\n+ report_dir = reporter.get_reports_dir()\n+ return os.path.join(report_dir, 'coverage')\n+\n+\n+def get_coverage_binary(benchmark: str) -> str:\n+ \"\"\"Gets the coverage binary for benchmark.\"\"\"\n+ coverage_binaries_dir = build_utils.get_coverage_binaries_dir()\n+ fuzz_target = benchmark_utils.get_fuzz_target(benchmark)\n+ return fuzzer_utils.get_fuzz_target_binary(coverage_binaries_dir /\n+ benchmark,\n+ fuzz_target_name=fuzz_target)\n+\n+\n+def get_trial_ids(experiment: str, fuzzer: str, benchmark: str):\n+ \"\"\"Gets ids of all finished trials for a pair of fuzzer and benchmark.\"\"\"\n+ trial_ids = [\n+ trial_id_tuple[0]\n+ for trial_id_tuple in db_utils.query(models.Trial.id).filter(\n+ models.Trial.experiment == experiment, models.Trial.fuzzer ==\n+ fuzzer, models.Trial.benchmark == benchmark,\n+ ~models.Trial.preempted)\n+ ]\n+ return trial_ids\n+\n+\n+def merge_profdata_files(src_files, dst_file):\n+ \"\"\"Uses llvm-profdata to merge |src_files| to |dst_files|.\"\"\"\n+ command = ['llvm-profdata', 'merge', '-sparse']\n+ command.extend(src_files)\n+ command.extend(['-o', dst_file])\n+ result = new_process.execute(command, expect_zero=False)\n+ return result\n+\n+\n+def get_coverage_infomation(coverage_summary_file):\n+ \"\"\"Reads the coverage information from |coverage_summary_file|\n+ and skip possible warnings in the file.\"\"\"\n+ with open(coverage_summary_file) as summary:\n+ return json.loads(summary.readlines()[-1])\n+\n+\n+def store_coverage_data(experiment_config: dict):\n+ \"\"\"Generates the specific coverage data and store in cloud bucket.\"\"\"\n+ logger.info('Start storing coverage data')\n+ with multiprocessing.Pool() as pool, multiprocessing.Manager() as manager:\n+ q = manager.Queue() # pytype: disable=attribute-error\n+ covered_regions = get_all_covered_regions(experiment_config, pool, q)\n+ json_src_dir = reporter.get_reports_dir()\n+ filesystem.recreate_directory(json_src_dir)\n+ json_src = os.path.join(json_src_dir, 'covered_regions.json')\n+ with open(json_src, 'w') as src_file:\n+ json.dump(covered_regions, src_file)\n+ json_dst = exp_path.filestore(json_src)\n+ filestore_utils.cp(json_src, json_dst)\n+\n+ logger.info('Finished storing coverage data')\n+\n+\n+def get_all_covered_regions(experiment_config: dict, pool, q) -> dict:\n+ \"\"\"Gets regions covered for each pair for fuzzer and benchmark.\"\"\"\n+ logger.info('Measuring all fuzzer-benchmark pairs for final coverage data.')\n+\n+ benchmarks = experiment_config['benchmarks'].split(',')\n+ fuzzers = experiment_config['fuzzers'].split(',')\n+ experiment = experiment_config['experiment']\n+\n+ get_covered_region_args = [(experiment, fuzzer, benchmark, q)\n+ for fuzzer in fuzzers\n+ for benchmark in benchmarks]\n+\n+ result = pool.starmap_async(get_covered_region, get_covered_region_args)\n+\n+ # Poll the queue for covered region data and save them in a dict until the\n+ # pool is done processing each combination of fuzzers and benchmarks.\n+ all_covered_regions = {}\n+\n+ while True:\n+ try:\n+ covered_regions = q.get(timeout=COV_DIFF_QUEUE_GET_TIMEOUT)\n+ all_covered_regions.update(covered_regions)\n+ except queue.Empty:\n+ if result.ready():\n+ # If \"ready\" that means pool has finished. Since it is\n+ # finished and the queue is empty, we can stop checking\n+ # the queue for more covered regions.\n+ logger.debug(\n+ 'Finished call to map with get_all_covered_regions.')\n+ break\n+\n+ for key in all_covered_regions:\n+ all_covered_regions[key] = list(all_covered_regions[key])\n+ logger.info('Done measuring all coverage data.')\n+ return all_covered_regions\n+\n+\n+def get_covered_region(experiment: str, fuzzer: str, benchmark: str,\n+ q: multiprocessing.Queue):\n+ \"\"\"Gets the final covered region for a specific pair of fuzzer-benchmark.\"\"\"\n+ logs.initialize()\n+ logger.debug('Measuring covered region: fuzzer: %s, benchmark: %s.', fuzzer,\n+ benchmark)\n+ key = get_fuzzer_benchmark_key(fuzzer, benchmark)\n+ covered_regions = {key: set()}\n+ trial_ids = get_trial_ids(experiment, fuzzer, benchmark)\n+ for trial_id in trial_ids:\n+ logger.info('Measuring covered region: trial_id = %d.', trial_id)\n+ snapshot_logger = logs.Logger('measurer',\n+ default_extras={\n+ 'fuzzer': fuzzer,\n+ 'benchmark': benchmark,\n+ 'trial_id': str(trial_id),\n+ })\n+ trial_coverage = TrialCoverage(fuzzer, benchmark, trial_id,\n+ snapshot_logger)\n+ trial_coverage.generate_summary(0, summary_only=False)\n+ new_covered_regions = trial_coverage.get_current_covered_regions()\n+ covered_regions[key] = covered_regions[key].union(new_covered_regions)\n+ q.put(covered_regions)\n+ logger.debug('Done measuring covered region: fuzzer: %s, benchmark: %s.',\n+ fuzzer, benchmark)\n+\n+\n+class TrialCoverage: # pylint: disable=too-many-instance-attributes\n+ \"\"\"Base class for storing and getting coverage data for a trial.\"\"\"\n+\n+ def __init__(self, fuzzer: str, benchmark: str, trial_num: int,\n+ trial_logger: logs.Logger):\n+ self.fuzzer = fuzzer\n+ self.benchmark = benchmark\n+ self.trial_num = trial_num\n+ self.logger = trial_logger\n+ self.benchmark_fuzzer_trial_dir = exp_utils.get_trial_dir(\n+ fuzzer, benchmark, trial_num)\n+ self.work_dir = exp_utils.get_work_dir()\n+ self.measurement_dir = os.path.join(self.work_dir,\n+ 'measurement-folders',\n+ self.benchmark_fuzzer_trial_dir)\n+ self.report_dir = os.path.join(self.measurement_dir, 'reports')\n+\n+ # Store the profdata file for the current trial.\n+ self.profdata_file = os.path.join(self.report_dir, 'data.profdata')\n+\n+ # Store the coverage information in json form.\n+ self.cov_summary_file = os.path.join(self.report_dir,\n+ 'cov_summary.json')\n+\n+ def get_current_covered_regions(self):\n+ \"\"\"Get the covered regions for the current trial.\"\"\"\n+ covered_regions = set()\n+ try:\n+ coverage_info = get_coverage_infomation(self.cov_summary_file)\n+ functions_data = coverage_info['data'][0]['functions']\n+ # The fourth number in the region-list indicates if the region\n+ # is hit.\n+ hit_index = 4\n+ # The last number in the region-list indicates what type of the\n+ # region it is; 'code_region' is used to obtain various code\n+ # coverage statistic and is represented by number 0.\n+ type_index = -1\n+ for function_data in functions_data:\n+ for region in function_data['regions']:\n+ if region[hit_index] != 0 and region[type_index] == 0:\n+ covered_regions.add(tuple(region[:hit_index]))\n+ except Exception: # pylint: disable=broad-except\n+ self.logger.error(\n+ 'Coverage summary json file defective or missing.')\n+ return covered_regions\n+\n+ def generate_summary(self, cycle: int, summary_only=True):\n+ \"\"\"Transforms the .profdata file into json form.\"\"\"\n+ coverage_binary = get_coverage_binary(self.benchmark)\n+ command = [\n+ 'llvm-cov', 'export', '-format=text', coverage_binary,\n+ '-instr-profile=%s' % self.profdata_file\n+ ]\n+\n+ if summary_only:\n+ command.append('-summary-only')\n+\n+ with open(self.cov_summary_file, 'w') as output_file:\n+ result = new_process.execute(command,\n+ output_file=output_file,\n+ expect_zero=False)\n+ if result.retcode != 0:\n+ self.logger.error(\n+ 'Coverage summary json file generation failed for \\\n+ cycle: %d.', cycle)\n+ if cycle != 0:\n+ self.logger.error(\n+ 'Coverage summary json file generation failed for \\\n+ cycle: %d.', cycle)\n+ else:\n+ self.logger.error(\n+ 'Coverage summary json file generation failed in the end.')\n" }, { "change_type": "MODIFY", "old_path": "experiment/dispatcher.py", "new_path": "experiment/dispatcher.py", "diff": "@@ -155,7 +155,9 @@ def dispatcher_main():\nis_complete = not measurer_main_process.is_alive()\n# Generate periodic output reports.\n- reporter.output_report(experiment.config, in_progress=not is_complete)\n+ reporter.output_report(experiment.config,\n+ in_progress=not is_complete,\n+ coverage_report=is_complete)\nif is_complete:\n# Experiment is complete, bail out.\n" }, { "change_type": "MODIFY", "old_path": "experiment/measurer.py", "new_path": "experiment/measurer.py", "diff": "@@ -24,25 +24,22 @@ import tarfile\nimport time\nfrom typing import List, Set\nimport queue\n-import json\nfrom sqlalchemy import func\nfrom sqlalchemy import orm\n-from common import benchmark_utils\nfrom common import experiment_utils\nfrom common import experiment_path as exp_path\nfrom common import filesystem\n-from common import fuzzer_utils\nfrom common import filestore_utils\nfrom common import logs\nfrom common import utils\n-from common import new_process\nfrom database import utils as db_utils\nfrom database import models\nfrom experiment.build import build_utils\nfrom experiment import run_coverage\nfrom experiment import scheduler\n+from experiment import coverage_utils\nlogger = logs.Logger('measurer') # pylint: disable=invalid-name\n@@ -53,124 +50,15 @@ NUM_RETRIES = 3\nRETRY_DELAY = 3\nFAIL_WAIT_SECONDS = 30\nSNAPSHOT_QUEUE_GET_TIMEOUT = 1\n-COV_DIFF_QUEUE_GET_TIMEOUT = 1\nSNAPSHOTS_BATCH_SAVE_SIZE = 100\n-def get_experiment_folders_dir():\n- \"\"\"Return experiment folders directory.\"\"\"\n- return exp_path.path('experiment-folders')\n-\n-\ndef exists_in_experiment_filestore(path: pathlib.Path) -> bool:\n\"\"\"Returns True if |path| exists in the experiment_filestore.\"\"\"\nreturn filestore_utils.ls(exp_path.filestore(path),\nmust_exist=False).retcode == 0\n-def get_fuzzer_benchmark_key(fuzzer: str, benchmark: str):\n- \"\"\"Return the key in coverage dict for a pair of fuzzer-benchmark.\"\"\"\n- return fuzzer + ' ' + benchmark\n-\n-\n-def get_trial_ids(experiment: str, fuzzer: str, benchmark: str):\n- \"\"\"Get ids of all finished trials for a pair of fuzzer and benchmark.\"\"\"\n- trial_ids = [\n- trial_id_tuple[0]\n- for trial_id_tuple in db_utils.query(models.Trial.id).filter(\n- models.Trial.experiment == experiment, models.Trial.fuzzer ==\n- fuzzer, models.Trial.benchmark == benchmark,\n- ~models.Trial.preempted)\n- ]\n- return trial_ids\n-\n-\n-def get_coverage_infomation(coverage_summary_file):\n- \"\"\"Read the coverage information from |coverage_summary_file|\n- and skip possible warnings in the file.\"\"\"\n- with open(coverage_summary_file) as summary:\n- return json.loads(summary.readlines()[-1])\n-\n-\n-def store_coverage_data(experiment_config: dict):\n- \"\"\"Generate the specific coverage data and store in cloud bucket.\"\"\"\n- logger.info('Start storing coverage data')\n- with multiprocessing.Pool() as pool, multiprocessing.Manager() as manager:\n- q = manager.Queue() # pytype: disable=attribute-error\n- covered_regions = get_all_covered_regions(experiment_config, pool, q)\n- json_src_dir = get_experiment_folders_dir()\n- json_src = os.path.join(json_src_dir, 'covered_regions.json')\n- with open(json_src, 'w') as src_file:\n- json.dump(covered_regions, src_file)\n- json_dst = exp_path.filestore(json_src)\n- filestore_utils.cp(json_src, json_dst)\n- logger.info('Finished storing coverage data')\n-\n-\n-def get_all_covered_regions(experiment_config: dict, pool, q) -> dict:\n- \"\"\"Get regions covered for each pair for fuzzer and benchmark.\"\"\"\n- logger.info('Measuring all fuzzer-benchmark pairs for final coverage data.')\n-\n- benchmarks = experiment_config['benchmarks'].split(',')\n- fuzzers = experiment_config['fuzzers'].split(',')\n- experiment = experiment_config['experiment']\n-\n- get_covered_region_args = [(experiment, fuzzer, benchmark, q)\n- for fuzzer in fuzzers\n- for benchmark in benchmarks]\n-\n- result = pool.starmap_async(get_covered_region, get_covered_region_args)\n-\n- # Poll the queue for covered region data and save them in a dict until the\n- # pool is done processing each combination of fuzzers and benchmarks.\n- all_covered_regions = {}\n-\n- while True:\n- try:\n- covered_regions = q.get(timeout=COV_DIFF_QUEUE_GET_TIMEOUT)\n- all_covered_regions.update(covered_regions)\n- except queue.Empty:\n- if result.ready():\n- # If \"ready\" that means pool has finished. Since it is\n- # finished and the queue is empty, we can stop checking\n- # the queue for more covered regions.\n- logger.debug(\n- 'Finished call to map with get_all_covered_regions.')\n- break\n-\n- for key in all_covered_regions:\n- all_covered_regions[key] = list(all_covered_regions[key])\n- logger.info('Done measuring all coverage data.')\n- return all_covered_regions\n-\n-\n-def get_covered_region(experiment: str, fuzzer: str, benchmark: str,\n- q: multiprocessing.Queue):\n- \"\"\"Get the final covered region for a specific pair of fuzzer-benchmark.\"\"\"\n- initialize_logs()\n- logger.debug('Measuring covered region: fuzzer: %s, benchmark: %s.', fuzzer,\n- benchmark)\n- key = get_fuzzer_benchmark_key(fuzzer, benchmark)\n- covered_regions = {key: set()}\n- trial_ids = get_trial_ids(experiment, fuzzer, benchmark)\n- for trial_id in trial_ids:\n- logger.info('Measuring covered region: trial_id = %d.', trial_id)\n- snapshot_logger = logs.Logger('measurer',\n- default_extras={\n- 'fuzzer': fuzzer,\n- 'benchmark': benchmark,\n- 'trial_id': str(trial_id),\n- })\n- snapshot_measurer = SnapshotMeasurer(fuzzer, benchmark, trial_id,\n- snapshot_logger)\n- snapshot_measurer.generate_summary(0, summary_only=False)\n- new_covered_regions = snapshot_measurer.get_current_covered_regions()\n- covered_regions[key] = covered_regions[key].union(new_covered_regions)\n- q.put(covered_regions)\n- logger.debug('Done measuring covered region: fuzzer: %s, benchmark: %s.',\n- fuzzer, benchmark)\n-\n-\ndef measure_main(experiment_config):\n\"\"\"Do the continuously measuring and the final measuring.\"\"\"\ninitialize_logs()\n@@ -182,7 +70,9 @@ def measure_main(experiment_config):\nmeasure_loop(experiment, max_total_time)\n# Do the final measuring and store the coverage data.\n- store_coverage_data(experiment_config)\n+ coverage_utils.store_coverage_data(experiment_config)\n+ coverage_utils.generate_coverage_reports(experiment_config)\n+ coverage_utils.upload_coverage_reports_to_bucket()\nlogger.info('Finished measuring.')\n@@ -222,7 +112,7 @@ def measure_all_trials(experiment: str, max_total_time: int, pool, q) -> bool:\nwas called first. Otherwise it will use fork which breaks logging.\"\"\"\nlogger.info('Measuring all trials.')\n- experiment_folders_dir = get_experiment_folders_dir()\n+ experiment_folders_dir = experiment_utils.get_experiment_folders_dir()\nif not exists_in_experiment_filestore(experiment_folders_dir):\nreturn True\n@@ -420,7 +310,7 @@ def extract_corpus(corpus_archive: str, sha_blacklist: Set[str],\nfilesystem.write(file_path, member_contents, 'wb')\n-class SnapshotMeasurer: # pylint: disable=too-many-instance-attributes\n+class SnapshotMeasurer(coverage_utils.TrialCoverage): # pylint: disable=too-many-instance-attributes\n\"\"\"Class used for storing details needed to measure coverage of a particular\ntrial.\"\"\"\n@@ -428,22 +318,13 @@ class SnapshotMeasurer: # pylint: disable=too-many-instance-attributes\ndef __init__(self, fuzzer: str, benchmark: str, trial_num: int,\ntrial_logger: logs.Logger):\n- self.fuzzer = fuzzer\n- self.benchmark = benchmark\n- self.trial_num = trial_num\n- self.logger = trial_logger\n- benchmark_fuzzer_trial_dir = experiment_utils.get_trial_dir(\n- fuzzer, benchmark, trial_num)\n- work_dir = experiment_utils.get_work_dir()\n- measurement_dir = os.path.join(work_dir, 'measurement-folders',\n- benchmark_fuzzer_trial_dir)\n- self.corpus_dir = os.path.join(measurement_dir, 'corpus')\n-\n- self.crashes_dir = os.path.join(measurement_dir, 'crashes')\n- self.coverage_dir = os.path.join(measurement_dir, 'coverage')\n- self.report_dir = os.path.join(measurement_dir, 'reports')\n- self.trial_dir = os.path.join(work_dir, 'experiment-folders',\n- benchmark_fuzzer_trial_dir)\n+ super().__init__(fuzzer, benchmark, trial_num, trial_logger)\n+ self.corpus_dir = os.path.join(self.measurement_dir, 'corpus')\n+\n+ self.crashes_dir = os.path.join(self.measurement_dir, 'crashes')\n+ self.coverage_dir = os.path.join(self.measurement_dir, 'coverage')\n+ self.trial_dir = os.path.join(self.work_dir, 'experiment-folders',\n+ self.benchmark_fuzzer_trial_dir)\n# Stores the files that have already been measured for a trial.\nself.measured_files_path = os.path.join(self.report_dir,\n@@ -473,7 +354,7 @@ class SnapshotMeasurer: # pylint: disable=too-many-instance-attributes\ndef run_cov_new_units(self):\n\"\"\"Run the coverage binary on new units.\"\"\"\n- coverage_binary = get_coverage_binary(self.benchmark)\n+ coverage_binary = coverage_utils.get_coverage_binary(self.benchmark)\ncrashing_units = run_coverage.do_coverage_run(coverage_binary,\nself.corpus_dir,\nself.profraw_file,\n@@ -482,35 +363,14 @@ class SnapshotMeasurer: # pylint: disable=too-many-instance-attributes\nself.UNIT_BLACKLIST[self.benchmark] = (\nself.UNIT_BLACKLIST[self.benchmark].union(set(crashing_units)))\n- def get_current_covered_regions(self):\n- \"\"\"Get the covered regions for the current trial.\"\"\"\n- covered_regions = set()\n- try:\n- coverage_info = get_coverage_infomation(self.cov_summary_file)\n- functions_data = coverage_info['data'][0]['functions']\n- # The fourth number in the region-list indicates if the region\n- # is hit.\n- hit_index = 4\n- # The last number in the region-list indicates what type of the\n- # region it is; 'code_region' is used to obtain various code\n- # coverage statistic and is represented by number 0.\n- type_index = -1\n- for function_data in functions_data:\n- for region in function_data['regions']:\n- if region[hit_index] != 0 and region[type_index] == 0:\n- covered_regions.add(tuple(region[:hit_index]))\n- except Exception: # pylint: disable=broad-except\n- self.logger.error(\n- 'Coverage summary json file defective or missing.')\n- return covered_regions\n-\ndef get_current_coverage(self) -> int:\n\"\"\"Get the current number of lines covered.\"\"\"\nif not os.path.exists(self.cov_summary_file):\nself.logger.warning('No coverage summary json file found.')\nreturn 0\ntry:\n- coverage_info = get_coverage_infomation(self.cov_summary_file)\n+ coverage_info = coverage_utils.get_coverage_infomation(\n+ self.cov_summary_file)\ncoverage_data = coverage_info[\"data\"][0]\nsummary_data = coverage_data[\"totals\"]\nregions_coverage_data = summary_data[\"regions\"]\n@@ -529,38 +389,13 @@ class SnapshotMeasurer: # pylint: disable=too-many-instance-attributes\nfiles_to_merge = [self.profraw_file, self.profdata_file]\nelse:\nfiles_to_merge = [self.profraw_file]\n- command = ['llvm-profdata', 'merge', '-sparse'\n- ] + files_to_merge + ['-o', self.profdata_file]\n- result = new_process.execute(command, expect_zero=False)\n+ result = coverage_utils.merge_profdata_files(files_to_merge,\n+ self.profdata_file)\nif result.retcode != 0:\nself.logger.error(\n'Coverage profdata generation failed for cycle: %d.', cycle)\n- def generate_summary(self, cycle: int, summary_only=True):\n- \"\"\"Transform the .profdata file into json form.\"\"\"\n- coverage_binary = get_coverage_binary(self.benchmark)\n- command = [\n- 'llvm-cov', 'export', '-format=text', coverage_binary,\n- '-instr-profile=%s' % self.profdata_file\n- ]\n-\n- if summary_only:\n- command.append('-summary-only')\n-\n- with open(self.cov_summary_file, 'w') as output_file:\n- result = new_process.execute(command,\n- output_file=output_file,\n- expect_zero=False)\n- if result.retcode != 0:\n- if cycle != 0:\n- self.logger.error(\n- 'Coverage summary json file generation failed for \\\n- cycle: %d.', cycle)\n- else:\n- self.logger.error(\n- 'Coverage summary json file generation failed in the end.')\n-\ndef generate_coverage_information(self, cycle: int):\n\"\"\"Generate the .profdata file and then transform it into\njson summary.\"\"\"\n@@ -790,15 +625,6 @@ def set_up_coverage_binary(benchmark):\nos.remove(archive_path)\n-def get_coverage_binary(benchmark: str) -> str:\n- \"\"\"Get the coverage binary for benchmark.\"\"\"\n- coverage_binaries_dir = build_utils.get_coverage_binaries_dir()\n- fuzz_target = benchmark_utils.get_fuzz_target(benchmark)\n- return fuzzer_utils.get_fuzz_target_binary(coverage_binaries_dir /\n- benchmark,\n- fuzz_target_name=fuzz_target)\n-\n-\ndef initialize_logs():\n\"\"\"Initialize logs. This must be called on process start.\"\"\"\nlogs.initialize(default_extras={\n" }, { "change_type": "MODIFY", "old_path": "experiment/reporter.py", "new_path": "experiment/reporter.py", "diff": "@@ -32,7 +32,9 @@ def get_reports_dir():\nreturn exp_path.path('reports')\n-def output_report(experiment_config: dict, in_progress=False):\n+def output_report(experiment_config: dict,\n+ in_progress=False,\n+ coverage_report=False):\n\"\"\"Generate the HTML report and write it to |web_bucket|.\"\"\"\nexperiment_name = experiment_utils.get_experiment_name()\nweb_filestore_path = posixpath.join(experiment_config['report_filestore'],\n@@ -54,7 +56,8 @@ def output_report(experiment_config: dict, in_progress=False):\nstr(reports_dir),\nreport_name=experiment_name,\nin_progress=in_progress,\n- merge_with_clobber_nonprivate=merge_with_nonprivate)\n+ merge_with_clobber_nonprivate=merge_with_nonprivate,\n+ coverage_report=coverage_report)\nfilestore_utils.rsync(str(reports_dir),\nweb_filestore_path,\ngsutil_options=[\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_measurer.py", "new_path": "experiment/test_measurer.py", "diff": "@@ -144,7 +144,7 @@ def test_generate_profdata_merge(mocked_execute, experiment, fs):\[email protected]('common.new_process.execute')\[email protected]('experiment.measurer.get_coverage_binary')\[email protected]('experiment.coverage_utils.get_coverage_binary')\ndef test_generate_summary(mocked_get_coverage_binary, mocked_execute,\nexperiment, fs):\n\"\"\"Tests that generate_summary can run the correct command.\"\"\"\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_reporter.py", "new_path": "experiment/test_reporter.py", "diff": "@@ -50,5 +50,6 @@ def test_output_report_filestore(fs, experiment):\n[experiment_name],\nreports_dir,\nreport_name=experiment_name,\n+ coverage_report=False,\nin_progress=False,\nmerge_with_clobber_nonprivate=False)\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/coverage/builder.Dockerfile", "new_path": "fuzzers/coverage/builder.Dockerfile", "diff": "@@ -26,3 +26,7 @@ RUN git clone https://github.com/llvm/llvm-project.git /llvm-project && \\\ncd /llvm-project/compiler-rt/lib/fuzzer && \\\nbash build.sh && \\\ncp libFuzzer.a /usr/lib\n+\n+# Copy source code files of benchmark to $OUT for the report generation.\n+RUN mkdir $OUT/src && \\\n+ cp -rL --parent $SRC $OUT/src\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/build_and_test_run_fuzzer_benchmarks.py", "new_path": ".github/workflows/build_and_test_run_fuzzer_benchmarks.py", "diff": "@@ -52,11 +52,9 @@ STANDARD_BENCHMARKS = {\ndef get_make_targets(benchmarks, fuzzer):\n- \"\"\"Return pull and test targets for |fuzzer| and each benchmark\n+ \"\"\"Returns and test targets for |fuzzer| and each benchmark\nin |benchmarks| to pass to make.\"\"\"\n- return [('pull-%s-%s' % (fuzzer, benchmark),\n- 'test-run-%s-%s' % (fuzzer, benchmark))\n- for benchmark in benchmarks]\n+ return ['test-run-%s-%s' % (fuzzer, benchmark) for benchmark in benchmarks]\ndef delete_docker_images():\n@@ -81,18 +79,14 @@ def delete_docker_images():\ndef make_builds(benchmarks, fuzzer):\n- \"\"\"Use make to build each target in |build_targets|.\"\"\"\n+ \"\"\"Use make to test the fuzzer on each benchmark in |benchmarks|.\"\"\"\nprint('Building benchmarks: {} for fuzzer: {}'.format(\n', '.join(benchmarks), fuzzer))\nmake_targets = get_make_targets(benchmarks, fuzzer)\n- for pull_target, build_target in make_targets:\n- # Pull target first.\n- subprocess.run(['make', '-j', pull_target], check=False)\n-\n- # Then build.\n- build_command = ['make', 'RUNNING_ON_CI=yes', '-j', build_target]\n- print('Running command:', ' '.join(build_command))\n- result = subprocess.run(build_command, check=False)\n+ for make_target in make_targets:\n+ make_command = ['make', 'RUNNING_ON_CI=yes', '-j', make_target]\n+ print('Running command:', ' '.join(make_command))\n+ result = subprocess.run(make_command, check=False)\nif not result.returncode == 0:\nreturn False\n# Delete docker images so disk doesn't fill up.\n" }, { "change_type": "MODIFY", "old_path": "docker/benchmark-runner/Dockerfile", "new_path": "docker/benchmark-runner/Dockerfile", "diff": "@@ -54,13 +54,16 @@ ENV WORKDIR /out\nRUN mkdir -p $WORKDIR\nWORKDIR $WORKDIR\n+# The argument needs to be re-declared otherwise it becomes an empty string.\n+ARG fuzzer\n+\n# Copy over all the build artifacts (without * to preserve directory structure).\n# This also copies seed and dictionary files if they are available.\nCOPY --from=builder /out/ ./\n# Copy the fuzzer.py file.\n-COPY --from=builder /src/fuzzer.py .\n+COPY fuzzers/$fuzzer/fuzzer.py .\n# Copy the fuzzers directory.\n-COPY --from=builder /src/fuzzers fuzzers\n+COPY fuzzers ./fuzzers\n# Create empty __init__.py to allow python deps to work.\nRUN touch __init__.py\n" } ]

[ { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-08-14\n+ fuzzers:\n+ - libfuzzer_fixcrossover\n+ - entropic_fixcrossover\n+\n- experiment: 2020-08-13\nfuzzers:\n- aflplusplus\n" } ]

[ { "change_type": "MODIFY", "old_path": "fuzzers/entropic_keepseed/builder.Dockerfile", "new_path": "fuzzers/entropic_keepseed/builder.Dockerfile", "diff": "ARG parent_image\nFROM $parent_image\n+COPY patch-crossover.diff /\nCOPY patch.diff /\nRUN git clone https://github.com/llvm/llvm-project.git /llvm-project && \\\ncd /llvm-project && \\\ngit checkout b52b2e1c188072e3cbc91500cfd503fb26d50ffc && \\\n+ patch -p1 < /patch-crossover.diff && \\\npatch -p1 < /patch.diff && \\\ncd /llvm-project/compiler-rt/lib/fuzzer && \\\n(for f in *.cpp; do \\\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_keepseed/patch-crossover.diff", "diff": "+commit df6a841a7c8b1c540f0e3c31c3697a11b2a00f51\n+Author: Dokyung Song <[email protected]>\n+Date: Wed Aug 5 23:12:19 2020 +0000\n+\n+ [libFuzzer] Fix arguments of InsertPartOf/CopyPartOf calls in CrossOver mutator.\n+\n+ The CrossOver mutator is meant to cross over two given buffers\n+ (referred to as the first/second buffer below). Previously\n+ InsertPartOf/CopyPartOf calls used in the CrossOver mutator\n+ incorrectly inserted/copied part of the second buffer into a \"scratch\n+ buffer\" (MutateInPlaceHere of the size CurrentMaxMutationLen), rather\n+ than the first buffer. This is not intended behavior, because the\n+ scratch buffer does not always (i) contain the content of the first\n+ buffer, and (ii) have the same size as the first buffer;\n+ CurrentMaxMutationLen is typically a lot larger than the size of the\n+ first buffer. This patch fixes the issue by using the first buffer\n+ instead of the scratch buffer in InsertPartOf/CopyPartOf calls.\n+\n+ This patch also adds two new tests, namely \"cross_over_insert\" and\n+ \"cross_over_copy\", which specifically target InsertPartOf and\n+ CopyPartOf, respectively.\n+\n+ - cross_over_insert.test checks if the fuzzer can use InsertPartOf to\n+ trigger the crash.\n+\n+ - cross_over_copy.test checks if the fuzzer can use CopyPartOf to\n+ trigger the crash.\n+\n+ These newly added tests were designed to pass with the current patch,\n+ but not without the it (with b216c80cc2496b87bf827260ce7e24dc62247d71\n+ these tests do no pass). To achieve this, -max_len was intentionally\n+ given a high value. Without this patch, InsertPartOf/CopyPartOf will\n+ generate larger inputs, possibly with unpredictable data in it,\n+ thereby failing to trigger the crash.\n+\n+ The test pass condition for these new tests is narrowed down by (i)\n+ limiting mutation depth to 1 (i.e., a single CrossOver mutation should\n+ be able to trigger the crash) and (ii) checking whether the mutation\n+ sequence of \"CrossOver-\" leads to the crash.\n+\n+ Also note that these newly added tests and an existing test\n+ (cross_over.test) all use \"-reduce_inputs=0\" flags to prevent reducing\n+ inputs; it's easier to force the fuzzer to keep original input string\n+ this way than tweaking cov-instrumented basic blocks in the source\n+ code of the fuzzer executable.\n+\n+ Differential Revision: https://reviews.llvm.org/D85554\n+\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerMutate.cpp b/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n+index 29541eac5dc..df9ada45bb0 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n++++ b/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n+@@ -425,26 +425,26 @@ size_t MutationDispatcher::Mutate_CrossOver(uint8_t *Data, size_t Size,\n+ if (!CrossOverWith) return 0;\n+ const Unit &O = *CrossOverWith;\n+ if (O.empty()) return 0;\n+- MutateInPlaceHere.resize(MaxSize);\n+- auto &U = MutateInPlaceHere;\n+ size_t NewSize = 0;\n+ switch(Rand(3)) {\n+ case 0:\n+- NewSize = CrossOver(Data, Size, O.data(), O.size(), U.data(), U.size());\n++ MutateInPlaceHere.resize(MaxSize);\n++ NewSize = CrossOver(Data, Size, O.data(), O.size(),\n++ MutateInPlaceHere.data(), MaxSize);\n++ memcpy(Data, MutateInPlaceHere.data(), NewSize);\n+ break;\n+ case 1:\n+- NewSize = InsertPartOf(O.data(), O.size(), U.data(), U.size(), MaxSize);\n++ NewSize = InsertPartOf(O.data(), O.size(), Data, Size, MaxSize);\n+ if (!NewSize)\n+- NewSize = CopyPartOf(O.data(), O.size(), U.data(), U.size());\n++ NewSize = CopyPartOf(O.data(), O.size(), Data, Size);\n+ break;\n+ case 2:\n+- NewSize = CopyPartOf(O.data(), O.size(), U.data(), U.size());\n++ NewSize = CopyPartOf(O.data(), O.size(), Data, Size);\n+ break;\n+ default: assert(0);\n+ }\n+ assert(NewSize > 0 && \"CrossOver returned empty unit\");\n+ assert(NewSize <= MaxSize && \"CrossOver returned overisized unit\");\n+- memcpy(Data, U.data(), NewSize);\n+ return NewSize;\n+ }\n+\n+diff --git a/compiler-rt/test/fuzzer/CrossOverTest.cpp b/compiler-rt/test/fuzzer/CrossOverTest.cpp\n+index a7643570a92..3ca53a8a851 100644\n+--- a/compiler-rt/test/fuzzer/CrossOverTest.cpp\n++++ b/compiler-rt/test/fuzzer/CrossOverTest.cpp\n+@@ -4,10 +4,10 @@\n+\n+ // Test for a fuzzer. The fuzzer must find the string\n+ // ABCDEFGHIJ\n+-// We use it as a test for CrossOver functionality\n+-// by passing two inputs to it:\n+-// ABCDE00000\n+-// ZZZZZFGHIJ\n++// We use it as a test for each of CrossOver functionalities\n++// by passing the following sets of two inputs to it:\n++// {ABCDEHIJ, ZFG} to test InsertPartOf\n++// {ABCDE00HIJ, ZFG} to test CopyPartOf\n+ //\n+ #include <assert.h>\n+ #include <cstddef>\n+@@ -16,6 +16,17 @@\n+ #include <iostream>\n+ #include <ostream>\n+\n++#ifndef INPUT_A\n++#define INPUT_A \"ABCDE00000\"\n++#endif\n++\n++#ifndef INPUT_B\n++#define INPUT_B \"ZZZZZFGHIJ\"\n++#endif\n++\n++const char *InputA = INPUT_A;\n++const char *InputB = INPUT_B;\n++\n+ static volatile int Sink;\n+ static volatile int *NullPtr;\n+\n+@@ -42,13 +53,11 @@ static const uint32_t ExpectedHash = 0xe1677acb;\n+\n+ extern \"C\" int LLVMFuzzerTestOneInput(const uint8_t *Data, size_t Size) {\n+ // fprintf(stderr, \"ExpectedHash: %x\\n\", ExpectedHash);\n+- if (Size != 10) return 0;\n+- if (*Data == 'A')\n++ if (Size == 10 && ExpectedHash == simple_hash(Data, Size))\n++ *NullPtr = 0;\n++ if (*Data == InputA[0])\n+ Sink++;\n+- if (*Data == 'Z')\n++ if (*Data == InputB[0])\n+ Sink--;\n+- if (ExpectedHash == simple_hash(Data, Size))\n+- *NullPtr = 0;\n+ return 0;\n+ }\n+-\n+diff --git a/compiler-rt/test/fuzzer/cross_over.test b/compiler-rt/test/fuzzer/cross_over.test\n+index 058b5eb2c85..64e06e8cd36 100644\n+--- a/compiler-rt/test/fuzzer/cross_over.test\n++++ b/compiler-rt/test/fuzzer/cross_over.test\n+@@ -12,7 +12,7 @@ RUN: echo -n ABCDE00000 > %t-corpus/A\n+ RUN: echo -n ZZZZZFGHIJ > %t-corpus/B\n+\n+\n+-RUN: not %run %t-CrossOverTest -max_len=10 -seed=1 -runs=10000000 %t-corpus\n++RUN: not %run %t-CrossOverTest -max_len=10 -reduce_inputs=0 -seed=1 -runs=10000000 %t-corpus\n+\n+ # Test the same thing but using -seed_inputs instead of passing the corpus dir.\n+-RUN: not %run %t-CrossOverTest -max_len=10 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B\n++RUN: not %run %t-CrossOverTest -max_len=10 -reduce_inputs=0 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B\n+diff --git a/compiler-rt/test/fuzzer/cross_over_copy.test b/compiler-rt/test/fuzzer/cross_over_copy.test\n+new file mode 100644\n+index 00000000000..f8f45c974e2\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/cross_over_copy.test\n+@@ -0,0 +1,20 @@\n++# Tests CrossOver CopyPartOf.\n++# We want to make sure that the test can find the input\n++# ABCDEFGHIJ when given two other inputs in the seed corpus:\n++# ABCDE00HIJ and\n++# (Z) FG\n++#\n++RUN: %cpp_compiler -DINPUT_A='\"ABCDE00HIJ\"' -DINPUT_B='\"ZFG\"' %S/CrossOverTest.cpp -o %t-CrossOverTest\n++\n++RUN: rm -rf %t-corpus\n++RUN: mkdir %t-corpus\n++RUN: echo -n ABCDE00HIJ > %t-corpus/A\n++RUN: echo -n ZFG > %t-corpus/B\n++\n++\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 %t-corpus 2>&1 | FileCheck %s\n++\n++# Test the same thing but using -seed_inputs instead of passing the corpus dir.\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B 2>&1 | FileCheck %s\n++\n++CHECK: MS: 1 CrossOver-\n+diff --git a/compiler-rt/test/fuzzer/cross_over_insert.test b/compiler-rt/test/fuzzer/cross_over_insert.test\n+new file mode 100644\n+index 00000000000..5ad2ff0a633\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/cross_over_insert.test\n+@@ -0,0 +1,20 @@\n++# Tests CrossOver InsertPartOf.\n++# We want to make sure that the test can find the input\n++# ABCDEFGHIJ when given two other inputs in the seed corpus:\n++# ABCDE HIJ and\n++# (Z) FG\n++#\n++RUN: %cpp_compiler -DINPUT_A='\"ABCDEHIJ\"' -DINPUT_B='\"ZFG\"' %S/CrossOverTest.cpp -o %t-CrossOverTest\n++\n++RUN: rm -rf %t-corpus\n++RUN: mkdir %t-corpus\n++RUN: echo -n ABCDEHIJ > %t-corpus/A\n++RUN: echo -n ZFG > %t-corpus/B\n++\n++\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 %t-corpus 2>&1 | FileCheck %s\n++\n++# Test the same thing but using -seed_inputs instead of passing the corpus dir.\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B 2>&1 | FileCheck %s\n++\n++CHECK: MS: 1 CrossOver-\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/libfuzzer_keepseed/builder.Dockerfile", "new_path": "fuzzers/libfuzzer_keepseed/builder.Dockerfile", "diff": "ARG parent_image\nFROM $parent_image\n+COPY patch-crossover.diff /\nCOPY patch.diff /\nRUN git clone https://github.com/llvm/llvm-project.git /llvm-project && \\\ncd /llvm-project/ && \\\ngit checkout b52b2e1c188072e3cbc91500cfd503fb26d50ffc && \\\n+ patch -p1 < /patch-crossover.diff && \\\npatch -p1 < /patch.diff && \\\ncd compiler-rt/lib/fuzzer && \\\n(for f in *.cpp; do \\\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/libfuzzer_keepseed/patch-crossover.diff", "diff": "+commit df6a841a7c8b1c540f0e3c31c3697a11b2a00f51\n+Author: Dokyung Song <[email protected]>\n+Date: Wed Aug 5 23:12:19 2020 +0000\n+\n+ [libFuzzer] Fix arguments of InsertPartOf/CopyPartOf calls in CrossOver mutator.\n+\n+ The CrossOver mutator is meant to cross over two given buffers\n+ (referred to as the first/second buffer below). Previously\n+ InsertPartOf/CopyPartOf calls used in the CrossOver mutator\n+ incorrectly inserted/copied part of the second buffer into a \"scratch\n+ buffer\" (MutateInPlaceHere of the size CurrentMaxMutationLen), rather\n+ than the first buffer. This is not intended behavior, because the\n+ scratch buffer does not always (i) contain the content of the first\n+ buffer, and (ii) have the same size as the first buffer;\n+ CurrentMaxMutationLen is typically a lot larger than the size of the\n+ first buffer. This patch fixes the issue by using the first buffer\n+ instead of the scratch buffer in InsertPartOf/CopyPartOf calls.\n+\n+ This patch also adds two new tests, namely \"cross_over_insert\" and\n+ \"cross_over_copy\", which specifically target InsertPartOf and\n+ CopyPartOf, respectively.\n+\n+ - cross_over_insert.test checks if the fuzzer can use InsertPartOf to\n+ trigger the crash.\n+\n+ - cross_over_copy.test checks if the fuzzer can use CopyPartOf to\n+ trigger the crash.\n+\n+ These newly added tests were designed to pass with the current patch,\n+ but not without the it (with b216c80cc2496b87bf827260ce7e24dc62247d71\n+ these tests do no pass). To achieve this, -max_len was intentionally\n+ given a high value. Without this patch, InsertPartOf/CopyPartOf will\n+ generate larger inputs, possibly with unpredictable data in it,\n+ thereby failing to trigger the crash.\n+\n+ The test pass condition for these new tests is narrowed down by (i)\n+ limiting mutation depth to 1 (i.e., a single CrossOver mutation should\n+ be able to trigger the crash) and (ii) checking whether the mutation\n+ sequence of \"CrossOver-\" leads to the crash.\n+\n+ Also note that these newly added tests and an existing test\n+ (cross_over.test) all use \"-reduce_inputs=0\" flags to prevent reducing\n+ inputs; it's easier to force the fuzzer to keep original input string\n+ this way than tweaking cov-instrumented basic blocks in the source\n+ code of the fuzzer executable.\n+\n+ Differential Revision: https://reviews.llvm.org/D85554\n+\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerMutate.cpp b/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n+index 29541eac5dc..df9ada45bb0 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n++++ b/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n+@@ -425,26 +425,26 @@ size_t MutationDispatcher::Mutate_CrossOver(uint8_t *Data, size_t Size,\n+ if (!CrossOverWith) return 0;\n+ const Unit &O = *CrossOverWith;\n+ if (O.empty()) return 0;\n+- MutateInPlaceHere.resize(MaxSize);\n+- auto &U = MutateInPlaceHere;\n+ size_t NewSize = 0;\n+ switch(Rand(3)) {\n+ case 0:\n+- NewSize = CrossOver(Data, Size, O.data(), O.size(), U.data(), U.size());\n++ MutateInPlaceHere.resize(MaxSize);\n++ NewSize = CrossOver(Data, Size, O.data(), O.size(),\n++ MutateInPlaceHere.data(), MaxSize);\n++ memcpy(Data, MutateInPlaceHere.data(), NewSize);\n+ break;\n+ case 1:\n+- NewSize = InsertPartOf(O.data(), O.size(), U.data(), U.size(), MaxSize);\n++ NewSize = InsertPartOf(O.data(), O.size(), Data, Size, MaxSize);\n+ if (!NewSize)\n+- NewSize = CopyPartOf(O.data(), O.size(), U.data(), U.size());\n++ NewSize = CopyPartOf(O.data(), O.size(), Data, Size);\n+ break;\n+ case 2:\n+- NewSize = CopyPartOf(O.data(), O.size(), U.data(), U.size());\n++ NewSize = CopyPartOf(O.data(), O.size(), Data, Size);\n+ break;\n+ default: assert(0);\n+ }\n+ assert(NewSize > 0 && \"CrossOver returned empty unit\");\n+ assert(NewSize <= MaxSize && \"CrossOver returned overisized unit\");\n+- memcpy(Data, U.data(), NewSize);\n+ return NewSize;\n+ }\n+\n+diff --git a/compiler-rt/test/fuzzer/CrossOverTest.cpp b/compiler-rt/test/fuzzer/CrossOverTest.cpp\n+index a7643570a92..3ca53a8a851 100644\n+--- a/compiler-rt/test/fuzzer/CrossOverTest.cpp\n++++ b/compiler-rt/test/fuzzer/CrossOverTest.cpp\n+@@ -4,10 +4,10 @@\n+\n+ // Test for a fuzzer. The fuzzer must find the string\n+ // ABCDEFGHIJ\n+-// We use it as a test for CrossOver functionality\n+-// by passing two inputs to it:\n+-// ABCDE00000\n+-// ZZZZZFGHIJ\n++// We use it as a test for each of CrossOver functionalities\n++// by passing the following sets of two inputs to it:\n++// {ABCDEHIJ, ZFG} to test InsertPartOf\n++// {ABCDE00HIJ, ZFG} to test CopyPartOf\n+ //\n+ #include <assert.h>\n+ #include <cstddef>\n+@@ -16,6 +16,17 @@\n+ #include <iostream>\n+ #include <ostream>\n+\n++#ifndef INPUT_A\n++#define INPUT_A \"ABCDE00000\"\n++#endif\n++\n++#ifndef INPUT_B\n++#define INPUT_B \"ZZZZZFGHIJ\"\n++#endif\n++\n++const char *InputA = INPUT_A;\n++const char *InputB = INPUT_B;\n++\n+ static volatile int Sink;\n+ static volatile int *NullPtr;\n+\n+@@ -42,13 +53,11 @@ static const uint32_t ExpectedHash = 0xe1677acb;\n+\n+ extern \"C\" int LLVMFuzzerTestOneInput(const uint8_t *Data, size_t Size) {\n+ // fprintf(stderr, \"ExpectedHash: %x\\n\", ExpectedHash);\n+- if (Size != 10) return 0;\n+- if (*Data == 'A')\n++ if (Size == 10 && ExpectedHash == simple_hash(Data, Size))\n++ *NullPtr = 0;\n++ if (*Data == InputA[0])\n+ Sink++;\n+- if (*Data == 'Z')\n++ if (*Data == InputB[0])\n+ Sink--;\n+- if (ExpectedHash == simple_hash(Data, Size))\n+- *NullPtr = 0;\n+ return 0;\n+ }\n+-\n+diff --git a/compiler-rt/test/fuzzer/cross_over.test b/compiler-rt/test/fuzzer/cross_over.test\n+index 058b5eb2c85..64e06e8cd36 100644\n+--- a/compiler-rt/test/fuzzer/cross_over.test\n++++ b/compiler-rt/test/fuzzer/cross_over.test\n+@@ -12,7 +12,7 @@ RUN: echo -n ABCDE00000 > %t-corpus/A\n+ RUN: echo -n ZZZZZFGHIJ > %t-corpus/B\n+\n+\n+-RUN: not %run %t-CrossOverTest -max_len=10 -seed=1 -runs=10000000 %t-corpus\n++RUN: not %run %t-CrossOverTest -max_len=10 -reduce_inputs=0 -seed=1 -runs=10000000 %t-corpus\n+\n+ # Test the same thing but using -seed_inputs instead of passing the corpus dir.\n+-RUN: not %run %t-CrossOverTest -max_len=10 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B\n++RUN: not %run %t-CrossOverTest -max_len=10 -reduce_inputs=0 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B\n+diff --git a/compiler-rt/test/fuzzer/cross_over_copy.test b/compiler-rt/test/fuzzer/cross_over_copy.test\n+new file mode 100644\n+index 00000000000..f8f45c974e2\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/cross_over_copy.test\n+@@ -0,0 +1,20 @@\n++# Tests CrossOver CopyPartOf.\n++# We want to make sure that the test can find the input\n++# ABCDEFGHIJ when given two other inputs in the seed corpus:\n++# ABCDE00HIJ and\n++# (Z) FG\n++#\n++RUN: %cpp_compiler -DINPUT_A='\"ABCDE00HIJ\"' -DINPUT_B='\"ZFG\"' %S/CrossOverTest.cpp -o %t-CrossOverTest\n++\n++RUN: rm -rf %t-corpus\n++RUN: mkdir %t-corpus\n++RUN: echo -n ABCDE00HIJ > %t-corpus/A\n++RUN: echo -n ZFG > %t-corpus/B\n++\n++\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 %t-corpus 2>&1 | FileCheck %s\n++\n++# Test the same thing but using -seed_inputs instead of passing the corpus dir.\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B 2>&1 | FileCheck %s\n++\n++CHECK: MS: 1 CrossOver-\n+diff --git a/compiler-rt/test/fuzzer/cross_over_insert.test b/compiler-rt/test/fuzzer/cross_over_insert.test\n+new file mode 100644\n+index 00000000000..5ad2ff0a633\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/cross_over_insert.test\n+@@ -0,0 +1,20 @@\n++# Tests CrossOver InsertPartOf.\n++# We want to make sure that the test can find the input\n++# ABCDEFGHIJ when given two other inputs in the seed corpus:\n++# ABCDE HIJ and\n++# (Z) FG\n++#\n++RUN: %cpp_compiler -DINPUT_A='\"ABCDEHIJ\"' -DINPUT_B='\"ZFG\"' %S/CrossOverTest.cpp -o %t-CrossOverTest\n++\n++RUN: rm -rf %t-corpus\n++RUN: mkdir %t-corpus\n++RUN: echo -n ABCDEHIJ > %t-corpus/A\n++RUN: echo -n ZFG > %t-corpus/B\n++\n++\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 %t-corpus 2>&1 | FileCheck %s\n++\n++# Test the same thing but using -seed_inputs instead of passing the corpus dir.\n++RUN: not %run %t-CrossOverTest -mutate_depth=1 -max_len=1024 -reduce_inputs=0 -seed=1 -runs=10000000 -seed_inputs=%t-corpus/A,%t-corpus/B 2>&1 | FileCheck %s\n++\n++CHECK: MS: 1 CrossOver-\n" }, { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "fuzzers:\n- libfuzzer_fixcrossover\n- entropic_fixcrossover\n+ - libfuzzer_keepseed\n+ - entropic_keepseed\n- experiment: 2020-08-13\nfuzzers:\n" } ]

[ { "change_type": "MODIFY", "old_path": "fuzzers/entropic_keepseed/fuzzer.py", "new_path": "fuzzers/entropic_keepseed/fuzzer.py", "diff": "@@ -27,4 +27,7 @@ def fuzz(input_corpus, output_corpus, target_binary):\nlibfuzzer_fuzzer.run_fuzzer(input_corpus,\noutput_corpus,\ntarget_binary,\n- extra_flags=['-entropic=1', '-keep_seed=1'])\n+ extra_flags=[\n+ '-entropic=1', '-keep_seed=1',\n+ '-cross_over_uniformdist=1'\n+ ])\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/entropic_keepseed/patch.diff", "new_path": "fuzzers/entropic_keepseed/patch.diff", "diff": "-commit aac9771fa16e3fc00725d4bbd662d71186a09532\n+commit 72e16fc7160185305eee536c257a478ae84f7082\nAuthor: Dokyung Song <[email protected]>\nDate: Fri Jul 31 00:07:20 2020 +0000\n[libFuzzer] Optionally keep initial seed inputs regardless of whether they discover new features or not.\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerCorpus.h b/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n-index 54d1e09ec6d..80398a9d7ce 100644\n+index 54d1e09ec6d..5c687013c59 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n+++ b/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n@@ -33,6 +33,7 @@ struct InputInfo {\n@@ -16,7 +16,7 @@ index 54d1e09ec6d..80398a9d7ce 100644\nbool MayDeleteFile = false;\nbool Reduced = false;\nbool HasFocusFunction = false;\n-@@ -131,9 +132,11 @@ class InputCorpus {\n+@@ -131,9 +132,12 @@ class InputCorpus {\nEntropicOptions Entropic;\n@@ -25,12 +25,13 @@ index 54d1e09ec6d..80398a9d7ce 100644\npublic:\n- InputCorpus(const std::string &OutputCorpus, EntropicOptions Entropic)\n- : Entropic(Entropic), OutputCorpus(OutputCorpus) {\n-+ InputCorpus(const std::string &OutputCorpus, EntropicOptions Entropic, bool KeepSeed)\n++ InputCorpus(const std::string &OutputCorpus, EntropicOptions Entropic,\n++ bool KeepSeed)\n+ : Entropic(Entropic), OutputCorpus(OutputCorpus), KeepSeed(KeepSeed) {\nmemset(InputSizesPerFeature, 0, sizeof(InputSizesPerFeature));\nmemset(SmallestElementPerFeature, 0, sizeof(SmallestElementPerFeature));\n}\n-@@ -177,7 +180,7 @@ public:\n+@@ -177,7 +181,7 @@ public:\nbool empty() const { return Inputs.empty(); }\nconst Unit &operator[] (size_t Idx) const { return Inputs[Idx]->U; }\nInputInfo *AddToCorpus(const Unit &U, size_t NumFeatures, bool MayDeleteFile,\n@@ -39,7 +40,7 @@ index 54d1e09ec6d..80398a9d7ce 100644\nconst Vector<uint32_t> &FeatureSet,\nconst DataFlowTrace &DFT, const InputInfo *BaseII) {\nassert(!U.empty());\n-@@ -187,6 +190,7 @@ public:\n+@@ -187,6 +191,7 @@ public:\nInputInfo &II = *Inputs.back();\nII.U = U;\nII.NumFeatures = NumFeatures;\n@@ -47,26 +48,20 @@ index 54d1e09ec6d..80398a9d7ce 100644\nII.MayDeleteFile = MayDeleteFile;\nII.UniqFeatureSet = FeatureSet;\nII.HasFocusFunction = HasFocusFunction;\n-@@ -471,7 +475,7 @@ private:\n-\n- for (size_t i = 0; i < N; i++) {\n-\n-- if (Inputs[i]->NumFeatures == 0) {\n-+ if (Inputs[i]->NumFeatures == 0 && !(Inputs[i]->SeedInput && KeepSeed)) {\n- // If the seed doesn't represent any features, assign zero energy.\n- Weights[i] = 0.;\n- } else if (Inputs[i]->NumExecutedMutations / kMaxMutationFactor >\n-@@ -491,7 +495,7 @@ private:\n-\n- if (VanillaSchedule) {\n- for (size_t i = 0; i < N; i++)\n-- Weights[i] = Inputs[i]->NumFeatures\n-+ Weights[i] = (Inputs[i]->NumFeatures || (KeepSeed && Inputs[i]->SeedInput))\n- ? (i + 1) * (Inputs[i]->HasFocusFunction ? 1000 : 1)\n- : 0.;\n+@@ -276,6 +281,11 @@ public:\n+ return Idx;\n}\n+\n++ InputInfo &ChooseUnitToCrossOverWith(Random &Rand) {\n++ InputInfo &II = *Inputs[Rand(Inputs.size())];\n++ return II;\n++ }\n++\n+ void PrintStats() {\n+ for (size_t i = 0; i < Inputs.size(); i++) {\n+ const auto &II = *Inputs[i];\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerDriver.cpp b/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n-index 00a33a413d2..ef7991c1e27 100644\n+index 8339697396c..0933af56804 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n+++ b/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n@@ -649,6 +649,7 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n@@ -77,33 +72,49 @@ index 00a33a413d2..ef7991c1e27 100644\nOptions.UnitTimeoutSec = Flags.timeout;\nOptions.ErrorExitCode = Flags.error_exitcode;\nOptions.TimeoutExitCode = Flags.timeout_exitcode;\n-@@ -753,7 +754,7 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n+@@ -657,6 +658,8 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n+ Options.IgnoreCrashes = Flags.ignore_crashes;\n+ Options.MaxTotalTimeSec = Flags.max_total_time;\n+ Options.DoCrossOver = Flags.cross_over;\n++ Options.CrossOverUniformDist = Flags.cross_over_uniformdist;\n++ Options.TraceSeedInput = Flags.trace_seed_input;\n+ Options.MutateDepth = Flags.mutate_depth;\n+ Options.ReduceDepth = Flags.reduce_depth;\n+ Options.UseCounters = Flags.use_counters;\n+@@ -753,7 +756,8 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\nRandom Rand(Seed);\nauto *MD = new MutationDispatcher(Rand, Options);\n- auto *Corpus = new InputCorpus(Options.OutputCorpus, Entropic);\n-+ auto *Corpus = new InputCorpus(Options.OutputCorpus, Entropic, Options.KeepSeed);\n++ auto *Corpus =\n++ new InputCorpus(Options.OutputCorpus, Entropic, Options.KeepSeed);\nauto *F = new Fuzzer(Callback, *Corpus, *MD, Options);\nfor (auto &U: Dictionary)\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerFlags.def b/compiler-rt/lib/fuzzer/FuzzerFlags.def\n-index 832224a705d..0dac7e705a3 100644\n+index 832224a705d..10b1f5f539a 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerFlags.def\n+++ b/compiler-rt/lib/fuzzer/FuzzerFlags.def\n-@@ -23,6 +23,8 @@ FUZZER_FLAG_INT(len_control, 100, \"Try generating small inputs first, \"\n+@@ -23,7 +23,14 @@ FUZZER_FLAG_INT(len_control, 100, \"Try generating small inputs first, \"\nFUZZER_FLAG_STRING(seed_inputs, \"A comma-separated list of input files \"\n\"to use as an additional seed corpus. Alternatively, an \\\"@\\\" followed by \"\n\"the name of a file containing the comma-separated list.\")\n-+FUZZER_FLAG_INT(keep_seed, 0, \"If 1, keep seed inputs for mutation even if \"\n-+ \"they do not produce new coverage.\")\n++FUZZER_FLAG_INT(keep_seed, 0, \"If 1, keep all seed inputs in the corpus even if \"\n++ \"they do not produce new coverage. This also invalidates -reduce_inputs for \"\n++ \"seed input mutations.\")\nFUZZER_FLAG_INT(cross_over, 1, \"If 1, cross over inputs.\")\n++FUZZER_FLAG_INT(cross_over_uniformdist, 0, \"Experimental. If 1, use a uniform \"\n++ \"probability distribution when choosing inputs to cross over with.\")\n++FUZZER_FLAG_INT(trace_seed_input, 0, \"Internal. Print all seed inputs picked \"\n++ \"up by the fuzzer.\")\nFUZZER_FLAG_INT(mutate_depth, 5,\n\"Apply this number of consecutive mutations to each input.\")\n+ FUZZER_FLAG_INT(reduce_depth, 0, \"Experimental/internal. \"\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerFork.cpp b/compiler-rt/lib/fuzzer/FuzzerFork.cpp\n-index d9e6b79443e..38fb82fc12d 100644\n+index d9e6b79443e..97e91cbe869 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerFork.cpp\n+++ b/compiler-rt/lib/fuzzer/FuzzerFork.cpp\n-@@ -309,11 +309,17 @@ void FuzzWithFork(Random &Rand, const FuzzingOptions &Options,\n+@@ -309,11 +309,20 @@ void FuzzWithFork(Random &Rand, const FuzzingOptions &Options,\nelse\nEnv.MainCorpusDir = CorpusDirs[0];\n@@ -115,13 +126,16 @@ index d9e6b79443e..38fb82fc12d 100644\n+ if (Options.KeepSeed) {\n+ for (auto &File : SeedFiles)\n+ Env.Files.push_back(File.File);\n-+ }\n-+ else {\n++ } else {\n+ auto CFPath = DirPlusFile(Env.TempDir, \"merge.txt\");\n+ CrashResistantMerge(Env.Args, {}, SeedFiles, &Env.Files, {}, &Env.Features,\n-+ {}, &Env.Cov,\n-+ CFPath, false);\n++ {}, &Env.Cov, CFPath, false);\n+ RemoveFile(CFPath);\n++ }\n++ if (Options.TraceSeedInput) {\n++ for (auto &File : Env.Files) {\n++ Printf(\"INFO: seed - %s\\n\", File.c_str());\n++ }\n+ }\nPrintf(\"INFO: -fork=%d: %zd seed inputs, starting to fuzz in %s\\n\", NumJobs,\nEnv.Files.size(), Env.TempDir.c_str());\n@@ -140,23 +154,53 @@ index 31096ce804b..e75807209f5 100644\nsize_t NumberOfLeakDetectionAttempts = 0;\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerLoop.cpp b/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n-index 02db6d27b0a..a9af25a3070 100644\n+index 02db6d27b0a..28d5f32c0d6 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n+++ b/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n-@@ -487,10 +487,11 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\n+@@ -478,7 +478,7 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\n+ UniqFeatureSetTmp.push_back(Feature);\n+ if (Options.Entropic)\n+ Corpus.UpdateFeatureFrequency(II, Feature);\n+- if (Options.ReduceInputs && II)\n++ if (Options.ReduceInputs && II && !(Options.KeepSeed && II->SeedInput))\n+ if (std::binary_search(II->UniqFeatureSet.begin(),\n+ II->UniqFeatureSet.end(), Feature))\n+ FoundUniqFeaturesOfII++;\n+@@ -487,11 +487,12 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\n*FoundUniqFeatures = FoundUniqFeaturesOfII;\nPrintPulseAndReportSlowInput(Data, Size);\nsize_t NumNewFeatures = Corpus.NumFeatureUpdates() - NumUpdatesBefore;\n- if (NumNewFeatures) {\n+ if (NumNewFeatures || (Options.KeepSeed && IsExecutingSeedCorpora)) {\nTPC.UpdateObservedPCs();\n- auto NewII = Corpus.AddToCorpus({Data, Data + Size}, NumNewFeatures,\n- MayDeleteFile, TPC.ObservedFocusFunction(),\n-+ IsExecutingSeedCorpora,\n- UniqFeatureSetTmp, DFT, II);\n+- auto NewII = Corpus.AddToCorpus({Data, Data + Size}, NumNewFeatures,\n+- MayDeleteFile, TPC.ObservedFocusFunction(),\n+- UniqFeatureSetTmp, DFT, II);\n++ auto NewII =\n++ Corpus.AddToCorpus({Data, Data + Size}, NumNewFeatures, MayDeleteFile,\n++ TPC.ObservedFocusFunction(),\n++ IsExecutingSeedCorpora, UniqFeatureSetTmp, DFT, II);\nWriteFeatureSetToFile(Options.FeaturesDir, Sha1ToString(NewII->Sha1),\nNewII->UniqFeatureSet);\n-@@ -764,6 +765,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+ return true;\n+@@ -664,8 +665,14 @@ void Fuzzer::MutateAndTestOne() {\n+ MD.StartMutationSequence();\n+\n+ auto &II = Corpus.ChooseUnitToMutate(MD.GetRand());\n+- if (Options.DoCrossOver)\n+- MD.SetCrossOverWith(&Corpus.ChooseUnitToMutate(MD.GetRand()).U);\n++ if (Options.DoCrossOver) {\n++ if (Options.CrossOverUniformDist) {\n++ MD.SetCrossOverWith(&Corpus.ChooseUnitToCrossOverWith(MD.GetRand()).U);\n++ }\n++ else {\n++ MD.SetCrossOverWith(&Corpus.ChooseUnitToMutate(MD.GetRand()).U);\n++ }\n++ }\n+ const auto &U = II.U;\n+ memcpy(BaseSha1, II.Sha1, sizeof(BaseSha1));\n+ assert(CurrentUnitData);\n+@@ -764,6 +771,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\nassert(CorporaFiles.front().Size <= CorporaFiles.back().Size);\n}\n@@ -165,7 +209,7 @@ index 02db6d27b0a..a9af25a3070 100644\n// Load and execute inputs one by one.\nfor (auto &SF : CorporaFiles) {\nauto U = FileToVector(SF.File, MaxInputLen, /*ExitOnError=*/false);\n-@@ -773,6 +776,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+@@ -773,6 +782,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\nTryDetectingAMemoryLeak(U.data(), U.size(),\n/*DuringInitialCorpusExecution*/ true);\n}\n@@ -174,7 +218,7 @@ index 02db6d27b0a..a9af25a3070 100644\n}\nPrintStats(\"INITED\");\n-@@ -785,6 +790,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+@@ -785,6 +796,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\nCorpus.NumInputsThatTouchFocusFunction());\n}\n@@ -184,7 +228,7 @@ index 02db6d27b0a..a9af25a3070 100644\nPrintf(\"ERROR: no interesting inputs were found. \"\n\"Is the code instrumented for coverage? Exiting.\\n\");\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerOptions.h b/compiler-rt/lib/fuzzer/FuzzerOptions.h\n-index 9d975bd61fe..ccd0b3dcb56 100644\n+index 9d975bd61fe..bb2c14be47b 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerOptions.h\n+++ b/compiler-rt/lib/fuzzer/FuzzerOptions.h\n@@ -18,6 +18,7 @@ struct FuzzingOptions {\n@@ -195,6 +239,15 @@ index 9d975bd61fe..ccd0b3dcb56 100644\nint UnitTimeoutSec = 300;\nint TimeoutExitCode = 70;\nint OOMExitCode = 71;\n+@@ -30,6 +31,8 @@ struct FuzzingOptions {\n+ int RssLimitMb = 0;\n+ int MallocLimitMb = 0;\n+ bool DoCrossOver = true;\n++ bool CrossOverUniformDist = false;\n++ bool TraceSeedInput = false;\n+ int MutateDepth = 5;\n+ bool ReduceDepth = false;\n+ bool UseCounters = false;\ndiff --git a/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp b/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp\nindex 0e9435ab8fc..dfc642ab6d0 100644\n--- a/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/libfuzzer_keepseed/fuzzer.py", "new_path": "fuzzers/libfuzzer_keepseed/fuzzer.py", "diff": "@@ -23,7 +23,8 @@ def build():\ndef fuzz(input_corpus, output_corpus, target_binary):\n\"\"\"Run fuzzer.\"\"\"\n- libfuzzer_fuzzer.run_fuzzer(input_corpus,\n+ libfuzzer_fuzzer.run_fuzzer(\n+ input_corpus,\noutput_corpus,\ntarget_binary,\n- extra_flags=['-keep_seed=1'])\n+ extra_flags=['-keep_seed=1', '-cross_over_uniformdist=1'])\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/libfuzzer_keepseed/patch.diff", "new_path": "fuzzers/libfuzzer_keepseed/patch.diff", "diff": "-commit aac9771fa16e3fc00725d4bbd662d71186a09532\n+commit 72e16fc7160185305eee536c257a478ae84f7082\nAuthor: Dokyung Song <[email protected]>\nDate: Fri Jul 31 00:07:20 2020 +0000\n[libFuzzer] Optionally keep initial seed inputs regardless of whether they discover new features or not.\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerCorpus.h b/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n-index 54d1e09ec6d..80398a9d7ce 100644\n+index 54d1e09ec6d..5c687013c59 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n+++ b/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n@@ -33,6 +33,7 @@ struct InputInfo {\n@@ -16,7 +16,7 @@ index 54d1e09ec6d..80398a9d7ce 100644\nbool MayDeleteFile = false;\nbool Reduced = false;\nbool HasFocusFunction = false;\n-@@ -131,9 +132,11 @@ class InputCorpus {\n+@@ -131,9 +132,12 @@ class InputCorpus {\nEntropicOptions Entropic;\n@@ -25,12 +25,13 @@ index 54d1e09ec6d..80398a9d7ce 100644\npublic:\n- InputCorpus(const std::string &OutputCorpus, EntropicOptions Entropic)\n- : Entropic(Entropic), OutputCorpus(OutputCorpus) {\n-+ InputCorpus(const std::string &OutputCorpus, EntropicOptions Entropic, bool KeepSeed)\n++ InputCorpus(const std::string &OutputCorpus, EntropicOptions Entropic,\n++ bool KeepSeed)\n+ : Entropic(Entropic), OutputCorpus(OutputCorpus), KeepSeed(KeepSeed) {\nmemset(InputSizesPerFeature, 0, sizeof(InputSizesPerFeature));\nmemset(SmallestElementPerFeature, 0, sizeof(SmallestElementPerFeature));\n}\n-@@ -177,7 +180,7 @@ public:\n+@@ -177,7 +181,7 @@ public:\nbool empty() const { return Inputs.empty(); }\nconst Unit &operator[] (size_t Idx) const { return Inputs[Idx]->U; }\nInputInfo *AddToCorpus(const Unit &U, size_t NumFeatures, bool MayDeleteFile,\n@@ -39,7 +40,7 @@ index 54d1e09ec6d..80398a9d7ce 100644\nconst Vector<uint32_t> &FeatureSet,\nconst DataFlowTrace &DFT, const InputInfo *BaseII) {\nassert(!U.empty());\n-@@ -187,6 +190,7 @@ public:\n+@@ -187,6 +191,7 @@ public:\nInputInfo &II = *Inputs.back();\nII.U = U;\nII.NumFeatures = NumFeatures;\n@@ -47,26 +48,20 @@ index 54d1e09ec6d..80398a9d7ce 100644\nII.MayDeleteFile = MayDeleteFile;\nII.UniqFeatureSet = FeatureSet;\nII.HasFocusFunction = HasFocusFunction;\n-@@ -471,7 +475,7 @@ private:\n-\n- for (size_t i = 0; i < N; i++) {\n-\n-- if (Inputs[i]->NumFeatures == 0) {\n-+ if (Inputs[i]->NumFeatures == 0 && !(Inputs[i]->SeedInput && KeepSeed)) {\n- // If the seed doesn't represent any features, assign zero energy.\n- Weights[i] = 0.;\n- } else if (Inputs[i]->NumExecutedMutations / kMaxMutationFactor >\n-@@ -491,7 +495,7 @@ private:\n-\n- if (VanillaSchedule) {\n- for (size_t i = 0; i < N; i++)\n-- Weights[i] = Inputs[i]->NumFeatures\n-+ Weights[i] = (Inputs[i]->NumFeatures || (KeepSeed && Inputs[i]->SeedInput))\n- ? (i + 1) * (Inputs[i]->HasFocusFunction ? 1000 : 1)\n- : 0.;\n+@@ -276,6 +281,11 @@ public:\n+ return Idx;\n}\n+\n++ InputInfo &ChooseUnitToCrossOverWith(Random &Rand) {\n++ InputInfo &II = *Inputs[Rand(Inputs.size())];\n++ return II;\n++ }\n++\n+ void PrintStats() {\n+ for (size_t i = 0; i < Inputs.size(); i++) {\n+ const auto &II = *Inputs[i];\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerDriver.cpp b/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n-index 00a33a413d2..ef7991c1e27 100644\n+index 8339697396c..0933af56804 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n+++ b/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n@@ -649,6 +649,7 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n@@ -77,33 +72,49 @@ index 00a33a413d2..ef7991c1e27 100644\nOptions.UnitTimeoutSec = Flags.timeout;\nOptions.ErrorExitCode = Flags.error_exitcode;\nOptions.TimeoutExitCode = Flags.timeout_exitcode;\n-@@ -753,7 +754,7 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n+@@ -657,6 +658,8 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n+ Options.IgnoreCrashes = Flags.ignore_crashes;\n+ Options.MaxTotalTimeSec = Flags.max_total_time;\n+ Options.DoCrossOver = Flags.cross_over;\n++ Options.CrossOverUniformDist = Flags.cross_over_uniformdist;\n++ Options.TraceSeedInput = Flags.trace_seed_input;\n+ Options.MutateDepth = Flags.mutate_depth;\n+ Options.ReduceDepth = Flags.reduce_depth;\n+ Options.UseCounters = Flags.use_counters;\n+@@ -753,7 +756,8 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\nRandom Rand(Seed);\nauto *MD = new MutationDispatcher(Rand, Options);\n- auto *Corpus = new InputCorpus(Options.OutputCorpus, Entropic);\n-+ auto *Corpus = new InputCorpus(Options.OutputCorpus, Entropic, Options.KeepSeed);\n++ auto *Corpus =\n++ new InputCorpus(Options.OutputCorpus, Entropic, Options.KeepSeed);\nauto *F = new Fuzzer(Callback, *Corpus, *MD, Options);\nfor (auto &U: Dictionary)\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerFlags.def b/compiler-rt/lib/fuzzer/FuzzerFlags.def\n-index 832224a705d..0dac7e705a3 100644\n+index 832224a705d..10b1f5f539a 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerFlags.def\n+++ b/compiler-rt/lib/fuzzer/FuzzerFlags.def\n-@@ -23,6 +23,8 @@ FUZZER_FLAG_INT(len_control, 100, \"Try generating small inputs first, \"\n+@@ -23,7 +23,14 @@ FUZZER_FLAG_INT(len_control, 100, \"Try generating small inputs first, \"\nFUZZER_FLAG_STRING(seed_inputs, \"A comma-separated list of input files \"\n\"to use as an additional seed corpus. Alternatively, an \\\"@\\\" followed by \"\n\"the name of a file containing the comma-separated list.\")\n-+FUZZER_FLAG_INT(keep_seed, 0, \"If 1, keep seed inputs for mutation even if \"\n-+ \"they do not produce new coverage.\")\n++FUZZER_FLAG_INT(keep_seed, 0, \"If 1, keep all seed inputs in the corpus even if \"\n++ \"they do not produce new coverage. This also invalidates -reduce_inputs for \"\n++ \"seed input mutations.\")\nFUZZER_FLAG_INT(cross_over, 1, \"If 1, cross over inputs.\")\n++FUZZER_FLAG_INT(cross_over_uniformdist, 0, \"Experimental. If 1, use a uniform \"\n++ \"probability distribution when choosing inputs to cross over with.\")\n++FUZZER_FLAG_INT(trace_seed_input, 0, \"Internal. Print all seed inputs picked \"\n++ \"up by the fuzzer.\")\nFUZZER_FLAG_INT(mutate_depth, 5,\n\"Apply this number of consecutive mutations to each input.\")\n+ FUZZER_FLAG_INT(reduce_depth, 0, \"Experimental/internal. \"\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerFork.cpp b/compiler-rt/lib/fuzzer/FuzzerFork.cpp\n-index d9e6b79443e..38fb82fc12d 100644\n+index d9e6b79443e..97e91cbe869 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerFork.cpp\n+++ b/compiler-rt/lib/fuzzer/FuzzerFork.cpp\n-@@ -309,11 +309,17 @@ void FuzzWithFork(Random &Rand, const FuzzingOptions &Options,\n+@@ -309,11 +309,20 @@ void FuzzWithFork(Random &Rand, const FuzzingOptions &Options,\nelse\nEnv.MainCorpusDir = CorpusDirs[0];\n@@ -115,13 +126,16 @@ index d9e6b79443e..38fb82fc12d 100644\n+ if (Options.KeepSeed) {\n+ for (auto &File : SeedFiles)\n+ Env.Files.push_back(File.File);\n-+ }\n-+ else {\n++ } else {\n+ auto CFPath = DirPlusFile(Env.TempDir, \"merge.txt\");\n+ CrashResistantMerge(Env.Args, {}, SeedFiles, &Env.Files, {}, &Env.Features,\n-+ {}, &Env.Cov,\n-+ CFPath, false);\n++ {}, &Env.Cov, CFPath, false);\n+ RemoveFile(CFPath);\n++ }\n++ if (Options.TraceSeedInput) {\n++ for (auto &File : Env.Files) {\n++ Printf(\"INFO: seed - %s\\n\", File.c_str());\n++ }\n+ }\nPrintf(\"INFO: -fork=%d: %zd seed inputs, starting to fuzz in %s\\n\", NumJobs,\nEnv.Files.size(), Env.TempDir.c_str());\n@@ -140,23 +154,53 @@ index 31096ce804b..e75807209f5 100644\nsize_t NumberOfLeakDetectionAttempts = 0;\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerLoop.cpp b/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n-index 02db6d27b0a..a9af25a3070 100644\n+index 02db6d27b0a..28d5f32c0d6 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n+++ b/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n-@@ -487,10 +487,11 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\n+@@ -478,7 +478,7 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\n+ UniqFeatureSetTmp.push_back(Feature);\n+ if (Options.Entropic)\n+ Corpus.UpdateFeatureFrequency(II, Feature);\n+- if (Options.ReduceInputs && II)\n++ if (Options.ReduceInputs && II && !(Options.KeepSeed && II->SeedInput))\n+ if (std::binary_search(II->UniqFeatureSet.begin(),\n+ II->UniqFeatureSet.end(), Feature))\n+ FoundUniqFeaturesOfII++;\n+@@ -487,11 +487,12 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\n*FoundUniqFeatures = FoundUniqFeaturesOfII;\nPrintPulseAndReportSlowInput(Data, Size);\nsize_t NumNewFeatures = Corpus.NumFeatureUpdates() - NumUpdatesBefore;\n- if (NumNewFeatures) {\n+ if (NumNewFeatures || (Options.KeepSeed && IsExecutingSeedCorpora)) {\nTPC.UpdateObservedPCs();\n- auto NewII = Corpus.AddToCorpus({Data, Data + Size}, NumNewFeatures,\n- MayDeleteFile, TPC.ObservedFocusFunction(),\n-+ IsExecutingSeedCorpora,\n- UniqFeatureSetTmp, DFT, II);\n+- auto NewII = Corpus.AddToCorpus({Data, Data + Size}, NumNewFeatures,\n+- MayDeleteFile, TPC.ObservedFocusFunction(),\n+- UniqFeatureSetTmp, DFT, II);\n++ auto NewII =\n++ Corpus.AddToCorpus({Data, Data + Size}, NumNewFeatures, MayDeleteFile,\n++ TPC.ObservedFocusFunction(),\n++ IsExecutingSeedCorpora, UniqFeatureSetTmp, DFT, II);\nWriteFeatureSetToFile(Options.FeaturesDir, Sha1ToString(NewII->Sha1),\nNewII->UniqFeatureSet);\n-@@ -764,6 +765,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+ return true;\n+@@ -664,8 +665,14 @@ void Fuzzer::MutateAndTestOne() {\n+ MD.StartMutationSequence();\n+\n+ auto &II = Corpus.ChooseUnitToMutate(MD.GetRand());\n+- if (Options.DoCrossOver)\n+- MD.SetCrossOverWith(&Corpus.ChooseUnitToMutate(MD.GetRand()).U);\n++ if (Options.DoCrossOver) {\n++ if (Options.CrossOverUniformDist) {\n++ MD.SetCrossOverWith(&Corpus.ChooseUnitToCrossOverWith(MD.GetRand()).U);\n++ }\n++ else {\n++ MD.SetCrossOverWith(&Corpus.ChooseUnitToMutate(MD.GetRand()).U);\n++ }\n++ }\n+ const auto &U = II.U;\n+ memcpy(BaseSha1, II.Sha1, sizeof(BaseSha1));\n+ assert(CurrentUnitData);\n+@@ -764,6 +771,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\nassert(CorporaFiles.front().Size <= CorporaFiles.back().Size);\n}\n@@ -165,7 +209,7 @@ index 02db6d27b0a..a9af25a3070 100644\n// Load and execute inputs one by one.\nfor (auto &SF : CorporaFiles) {\nauto U = FileToVector(SF.File, MaxInputLen, /*ExitOnError=*/false);\n-@@ -773,6 +776,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+@@ -773,6 +782,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\nTryDetectingAMemoryLeak(U.data(), U.size(),\n/*DuringInitialCorpusExecution*/ true);\n}\n@@ -174,7 +218,7 @@ index 02db6d27b0a..a9af25a3070 100644\n}\nPrintStats(\"INITED\");\n-@@ -785,6 +790,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+@@ -785,6 +796,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\nCorpus.NumInputsThatTouchFocusFunction());\n}\n@@ -184,7 +228,7 @@ index 02db6d27b0a..a9af25a3070 100644\nPrintf(\"ERROR: no interesting inputs were found. \"\n\"Is the code instrumented for coverage? Exiting.\\n\");\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerOptions.h b/compiler-rt/lib/fuzzer/FuzzerOptions.h\n-index 9d975bd61fe..ccd0b3dcb56 100644\n+index 9d975bd61fe..bb2c14be47b 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerOptions.h\n+++ b/compiler-rt/lib/fuzzer/FuzzerOptions.h\n@@ -18,6 +18,7 @@ struct FuzzingOptions {\n@@ -195,6 +239,15 @@ index 9d975bd61fe..ccd0b3dcb56 100644\nint UnitTimeoutSec = 300;\nint TimeoutExitCode = 70;\nint OOMExitCode = 71;\n+@@ -30,6 +31,8 @@ struct FuzzingOptions {\n+ int RssLimitMb = 0;\n+ int MallocLimitMb = 0;\n+ bool DoCrossOver = true;\n++ bool CrossOverUniformDist = false;\n++ bool TraceSeedInput = false;\n+ int MutateDepth = 5;\n+ bool ReduceDepth = false;\n+ bool UseCounters = false;\ndiff --git a/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp b/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp\nindex 0e9435ab8fc..dfc642ab6d0 100644\n--- a/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp\n" }, { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-08-15\n+ fuzzers:\n+ - libfuzzer_fixcrossover\n+ - entropic_fixcrossover\n+ - libfuzzer_keepseed\n+ - entropic_keepseed\n+\n- experiment: 2020-08-14\nfuzzers:\n- aflplusplus\n" } ]

[ { "change_type": "MODIFY", "old_path": "analysis/benchmark_results.py", "new_path": "analysis/benchmark_results.py", "diff": "@@ -17,10 +17,12 @@ import os\nimport functools\nfrom analysis import data_utils\n+from analysis import coverage_data_utils\nfrom analysis import stat_tests\n-class BenchmarkResults: # pylint: disable=too-many-public-methods\n+# pylint: disable=too-many-public-methods, too-many-arguments\n+class BenchmarkResults:\n\"\"\"Represents results of various analysis done on benchmark data.\nNOTE: Do not create this class manually! Instead, use the |benchmarks|\n@@ -31,11 +33,12 @@ class BenchmarkResults: # pylint: disable=too-many-public-methods\ntemplate, properties are computed on demand and only once.\n\"\"\"\n- def __init__(self, benchmark_name, experiment_df, output_directory,\n- plotter):\n+ def __init__(self, benchmark_name, experiment_df, coverage_dict,\n+ output_directory, plotter):\nself.name = benchmark_name\nself._experiment_df = experiment_df\n+ self._coverage_dict = coverage_dict\nself._output_directory = output_directory\nself._plotter = plotter\n@@ -45,6 +48,19 @@ class BenchmarkResults: # pylint: disable=too-many-public-methods\ndef _get_full_path(self, filename):\nreturn os.path.join(self._output_directory, filename)\n+ def _get_experiment_filestore_path(self, fuzzer_name):\n+ return coverage_data_utils.get_fuzzer_filestore_path(\n+ self._benchmark_df, fuzzer_name)\n+\n+ def get_filestore_name(self, fuzzer_name):\n+ \"\"\"Returns the filestore name of the |fuzzer_name|.\"\"\"\n+ filestore_path = self._get_experiment_filestore_path(fuzzer_name)\n+ gcs_prefix = 'gs://'\n+ gcs_http_prefix = 'https://storage.googleapis.com/'\n+ if filestore_path.startswith(gcs_prefix):\n+ filestore_path = filestore_path.replace(gcs_prefix, gcs_http_prefix)\n+ return filestore_path\n+\n@property\[email protected]_cache()\n# TODO(lszekeres): With python3.8+, replace above two decorators with:\n@@ -53,6 +69,12 @@ class BenchmarkResults: # pylint: disable=too-many-public-methods\nexp_df = self._experiment_df\nreturn exp_df[exp_df.benchmark == self.name]\n+ @property\n+ @functools.lru_cache()\n+ def fuzzer_names(self):\n+ \"\"\"Names of all fuzzers.\"\"\"\n+ return self._benchmark_df.fuzzer.unique()\n+\n@property\[email protected]_cache()\ndef _benchmark_snapshot_df(self):\n@@ -60,9 +82,31 @@ class BenchmarkResults: # pylint: disable=too-many-public-methods\n@property\[email protected]_cache()\n- def fuzzers(self):\n- \"\"\"Fuzzers with valid trials on this benchmark.\"\"\"\n- return self._benchmark_df.fuzzer.unique()\n+ def _benchmark_coverage_dict(self):\n+ \"\"\"Covered regions of each fuzzer on this benchmark.\"\"\"\n+ return coverage_data_utils.get_benchmark_cov_dict(\n+ self._coverage_dict, self.name)\n+\n+ @property\n+ @functools.lru_cache()\n+ def _benchmark_aggregated_coverage_df(self):\n+ \"\"\"Aggregated covered regions of each fuzzer on this benchmark.\"\"\"\n+ return coverage_data_utils.get_benchmark_aggregated_cov_df(\n+ self._benchmark_coverage_dict)\n+\n+ @property\n+ @functools.lru_cache()\n+ def _unique_region_dict(self):\n+ \"\"\"Unique regions with the fuzzers that cover it.\"\"\"\n+ return coverage_data_utils.get_unique_region_dict(\n+ self._benchmark_coverage_dict)\n+\n+ @property\n+ @functools.lru_cache()\n+ def _unique_region_cov_df(self):\n+ \"\"\"Fuzzers with the number of covered unique regions.\"\"\"\n+ return coverage_data_utils.get_unique_region_cov_df(\n+ self._unique_region_dict, self.fuzzer_names)\n@property\ndef fuzzers_with_not_enough_samples(self):\n@@ -236,3 +280,30 @@ class BenchmarkResults: # pylint: disable=too-many-public-methods\nself._plotter.write_better_than_plot(better_than_table,\nself._get_full_path(plot_filename))\nreturn plot_filename\n+\n+ @property\n+ def unique_coverage_ranking_plot(self):\n+ \"\"\"Ranking plot for unique coverage.\"\"\"\n+ plot_filename = self._prefix_with_benchmark('ranking_unique_region.svg')\n+ unique_region_cov_df_combined = self._unique_region_cov_df.merge(\n+ self._benchmark_aggregated_coverage_df, on='fuzzer')\n+ self._plotter.write_unique_coverage_ranking_plot(\n+ unique_region_cov_df_combined, self._get_full_path(plot_filename))\n+ return plot_filename\n+\n+ @property\n+ @functools.lru_cache()\n+ def pairwise_unique_coverage_table(self):\n+ \"\"\"Pairwise unique coverage table for each pair of fuzzers.\"\"\"\n+ return coverage_data_utils.get_pairwise_unique_coverage_table(\n+ self._benchmark_coverage_dict)\n+\n+ @property\n+ def pairwise_unique_coverage_plot(self):\n+ \"\"\"Pairwise unique coverage plot for each pair of fuzzers.\"\"\"\n+ plot_filename = self._prefix_with_benchmark(\n+ 'pairwise_unique_coverage_plot.svg')\n+ self._plotter.write_pairwise_unique_coverage_heatmap_plot(\n+ self.pairwise_unique_coverage_table,\n+ self._get_full_path(plot_filename))\n+ return plot_filename\n" }, { "change_type": "ADD", "old_path": null, "new_path": "analysis/coverage_data_utils.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Utility functions for coverage data calculation.\"\"\"\n+\n+import collections\n+import json\n+import os\n+import posixpath\n+import tempfile\n+import pandas as pd\n+\n+from common import filestore_utils\n+\n+\n+def get_fuzzer_benchmark_key(fuzzer: str, benchmark: str):\n+ \"\"\"Returns the key in coverage dict for a pair of fuzzer-benchmark.\"\"\"\n+ return fuzzer + ' ' + benchmark\n+\n+\n+def get_fuzzer_filestore_path(benchmark_df, fuzzer):\n+ \"\"\"Gets the filestore_path for |fuzzer| in |benchmark_df|.\"\"\"\n+ fuzzer_df = benchmark_df[benchmark_df.fuzzer == fuzzer]\n+ filestore_path = fuzzer_df.experiment_filestore.unique()[0]\n+ exp_name = fuzzer_df.experiment.unique()[0]\n+ return posixpath.join(filestore_path, exp_name)\n+\n+\n+def get_covered_regions_dict(experiment_df):\n+ \"\"\"Combines json files for different fuzzer-benchmark pair\n+ in |experiment_df| and returns a dictionary of the covered regions.\"\"\"\n+ covered_regions_dict = {}\n+ benchmarks = experiment_df.benchmark.unique()\n+ for benchmark in benchmarks:\n+ benchmark_df = experiment_df[experiment_df.benchmark == benchmark]\n+ fuzzers = benchmark_df.fuzzer.unique()\n+ for fuzzer in fuzzers:\n+ fuzzer_covered_regions = get_fuzzer_covered_regions(\n+ benchmark_df, benchmark, fuzzer)\n+ key = get_fuzzer_benchmark_key(fuzzer, benchmark)\n+ covered_regions_dict[key] = fuzzer_covered_regions\n+ return covered_regions_dict\n+\n+\n+def get_fuzzer_covered_regions(benchmark_df, benchmark, fuzzer):\n+ \"\"\"Gets the covered regions for |fuzzer| in |benchmark_df| from the json\n+ file in the bucket.\"\"\"\n+ with tempfile.TemporaryDirectory() as temp_dir:\n+ dst_file = os.path.join(temp_dir, 'tmp.json')\n+ src_filestore_path = get_fuzzer_filestore_path(benchmark_df, fuzzer)\n+ src_file = posixpath.join(src_filestore_path, 'coverage', 'data',\n+ benchmark, fuzzer, 'covered_regions.json')\n+ filestore_utils.cp(src_file, dst_file)\n+ with open(dst_file) as json_file:\n+ return json.load(json_file)\n+\n+\n+def get_unique_region_dict(benchmark_coverage_dict):\n+ \"\"\"Returns a dictionary containing the covering fuzzers for each\n+ unique region, where the |threshold| defines which regions are unique.\"\"\"\n+ region_dict = collections.defaultdict(list)\n+ unique_region_dict = {}\n+ threshold_count = 1\n+ for fuzzer in benchmark_coverage_dict:\n+ for region in benchmark_coverage_dict[fuzzer]:\n+ region_dict[region].append(fuzzer)\n+ for region, fuzzers in region_dict.items():\n+ if len(fuzzers) <= threshold_count:\n+ unique_region_dict[region] = fuzzers\n+ return unique_region_dict\n+\n+\n+def get_unique_region_cov_df(unique_region_dict, fuzzer_names):\n+ \"\"\"Returns a DataFrame where the two columns are fuzzers and the number\n+ of unique regions covered.\"\"\"\n+ fuzzers = collections.defaultdict(int)\n+ for region in unique_region_dict:\n+ for fuzzer in unique_region_dict[region]:\n+ fuzzers[fuzzer] += 1\n+ dict_to_transform = {'fuzzer': [], 'unique_regions_covered': []}\n+ for fuzzer in fuzzer_names:\n+ covered_num = fuzzers[fuzzer]\n+ dict_to_transform['fuzzer'].append(fuzzer)\n+ dict_to_transform['unique_regions_covered'].append(covered_num)\n+ return pd.DataFrame(dict_to_transform)\n+\n+\n+def get_benchmark_cov_dict(coverage_dict, benchmark):\n+ \"\"\"Returns a dictionary to store the covered regions of each fuzzer.\n+ Uses a set of tuples to store the covered regions.\"\"\"\n+ benchmark_cov_dict = {}\n+ for key_pair, covered_regions in coverage_dict.items():\n+ current_fuzzer, current_benchmark = key_pair.split()\n+ if current_benchmark == benchmark:\n+ covered_regions_in_set = set()\n+ for region in covered_regions:\n+ covered_regions_in_set.add(tuple(region))\n+ benchmark_cov_dict[current_fuzzer] = covered_regions_in_set\n+ return benchmark_cov_dict\n+\n+\n+def get_benchmark_aggregated_cov_df(benchmark_coverage_dict):\n+ \"\"\"Returns a dataframe where each row represents a fuzzer and its\n+ aggregated coverage number.\"\"\"\n+ dict_to_transform = {'fuzzer': [], 'aggregated_edges_covered': []}\n+ for fuzzer in benchmark_coverage_dict:\n+ aggregated_edges_covered = len(benchmark_coverage_dict[fuzzer])\n+ dict_to_transform['fuzzer'].append(fuzzer)\n+ dict_to_transform['aggregated_edges_covered'].append(\n+ aggregated_edges_covered)\n+ return pd.DataFrame(dict_to_transform)\n+\n+\n+def get_pairwise_unique_coverage_table(benchmark_coverage_dict):\n+ \"\"\"Returns a table that shows the unique coverage between\n+ each pair of fuzzers.\n+\n+ The pairwise unique coverage table is a square matrix where each\n+ row and column represents a fuzzer, and each cell contains a number\n+ showing the regions covered by the fuzzer of the column but not by\n+ the fuzzer of the row.\"\"\"\n+\n+ fuzzers = benchmark_coverage_dict.keys()\n+\n+ pairwise_unique_coverage_values = []\n+ for fuzzer_in_row in fuzzers:\n+ row = []\n+ for fuzzer_in_col in fuzzers:\n+ pairwise_unique_coverage_value = get_unique_covered_percentage(\n+ benchmark_coverage_dict[fuzzer_in_row],\n+ benchmark_coverage_dict[fuzzer_in_col])\n+ row.append(pairwise_unique_coverage_value)\n+ pairwise_unique_coverage_values.append(row)\n+\n+ return pd.DataFrame(pairwise_unique_coverage_values,\n+ index=fuzzers,\n+ columns=fuzzers)\n+\n+\n+def get_unique_covered_percentage(fuzzer_row_covered_regions,\n+ fuzzer_col_covered_regions):\n+ \"\"\"Returns the number of regions covered by the fuzzer of the column\n+ but not by the fuzzer of the row.\"\"\"\n+\n+ unique_region_count = 0\n+ for region in fuzzer_col_covered_regions:\n+ if region not in fuzzer_row_covered_regions:\n+ unique_region_count += 1\n+ return unique_region_count\n" }, { "change_type": "MODIFY", "old_path": "analysis/data_utils.py", "new_path": "analysis/data_utils.py", "diff": "@@ -38,7 +38,8 @@ def validate_data(experiment_df):\ndef drop_uninteresting_columns(experiment_df):\n\"\"\"Returns table with only interesting columns.\"\"\"\nreturn experiment_df[[\n- 'benchmark', 'fuzzer', 'trial_id', 'time', 'edges_covered'\n+ 'benchmark', 'fuzzer', 'trial_id', 'time', 'edges_covered',\n+ 'experiment', 'experiment_filestore'\n]]\n" }, { "change_type": "MODIFY", "old_path": "analysis/experiment_results.py", "new_path": "analysis/experiment_results.py", "diff": "@@ -21,7 +21,7 @@ from analysis import data_utils\nfrom analysis import stat_tests\n-class ExperimentResults:\n+class ExperimentResults: # pylint: disable=too-many-instance-attributes\n\"\"\"Provides the main interface for getting various analysis results and\nplots about an experiment, represented by |experiment_df|.\n@@ -31,8 +31,10 @@ class ExperimentResults:\ntemplate, only the properties needed for the given report will be computed.\n\"\"\"\n- def __init__(self,\n+ def __init__( # pylint: disable=too-many-arguments\n+ self,\nexperiment_df,\n+ coverage_dict,\noutput_directory,\nplotter,\nexperiment_name=None):\n@@ -63,6 +65,9 @@ class ExperimentResults:\nself._plotter = plotter\n+ # Dictionary to store the full coverage data.\n+ self._coverage_dict = coverage_dict\n+\ndef _get_full_path(self, filename):\nreturn os.path.join(self._output_directory, filename)\n@@ -87,6 +92,7 @@ class ExperimentResults:\nbenchmark_names = self._experiment_df.benchmark.unique()\nreturn [\nbenchmark_results.BenchmarkResults(name, self._experiment_df,\n+ self._coverage_dict,\nself._output_directory,\nself._plotter)\nfor name in sorted(benchmark_names)\n" }, { "change_type": "MODIFY", "old_path": "analysis/generate_report.py", "new_path": "analysis/generate_report.py", "diff": "@@ -20,6 +20,7 @@ import sys\nimport pandas as pd\nfrom analysis import data_utils\n+from analysis import coverage_data_utils\nfrom analysis import experiment_results\nfrom analysis import plotting\nfrom analysis import queries\n@@ -76,11 +77,12 @@ def get_arg_parser():\n'--fuzzers',\nnargs='*',\nhelp='Names of the fuzzers to include in the report.')\n- parser.add_argument('-cov',\n+ parser.add_argument(\n+ '-cov',\n'--coverage-report',\naction='store_true',\n- default=False,\n- help='If set, clang coverage reports are linked.')\n+ default=True,\n+ help='If set, clang coverage reports and differential plots are shown.')\n# It doesn't make sense to clobber and label by experiment, since nothing\n# can get clobbered like this.\n@@ -159,6 +161,12 @@ def generate_report(experiment_names,\ndata_utils.validate_data(experiment_df)\n+ # Load the json summary file.\n+ coverage_dict = {}\n+ if coverage_report:\n+ coverage_dict = coverage_data_utils.get_covered_regions_dict(\n+ experiment_df)\n+\nif benchmarks is not None:\nexperiment_df = data_utils.filter_benchmarks(experiment_df, benchmarks)\n@@ -178,7 +186,11 @@ def generate_report(experiment_names,\nfuzzer_names = experiment_df.fuzzer.unique()\nplotter = plotting.Plotter(fuzzer_names, quick, log_scale)\nexperiment_ctx = experiment_results.ExperimentResults(\n- experiment_df, report_directory, plotter, experiment_name=report_name)\n+ experiment_df,\n+ coverage_dict,\n+ report_directory,\n+ plotter,\n+ experiment_name=report_name)\ntemplate = report_type + '.html'\ndetailed_report = rendering.render_report(experiment_ctx, template,\n@@ -195,8 +207,7 @@ def main():\nparser = get_arg_parser()\nargs = parser.parse_args()\n- generate_report(\n- experiment_names=args.experiments,\n+ generate_report(experiment_names=args.experiments,\nreport_directory=args.report_dir,\nreport_name=args.report_name,\nlabel_by_experiment=args.label_by_experiment,\n@@ -208,7 +219,6 @@ def main():\nfrom_cached_data=args.from_cached_data,\nend_time=args.end_time,\nmerge_with_clobber=args.merge_with_clobber,\n- merge_with_clobber_nonprivate=args.merge_with_clobber_nonprivate,\ncoverage_report=args.coverage_report)\n" }, { "change_type": "MODIFY", "old_path": "analysis/plotting.py", "new_path": "analysis/plotting.py", "diff": "@@ -336,3 +336,73 @@ class Plotter:\nfig.savefig(image_path, bbox_inches=\"tight\")\nfinally:\nplt.close(fig)\n+\n+ def unique_coverage_ranking_plot(self,\n+ unique_region_cov_df_combined,\n+ axes=None):\n+ \"\"\"Draws unique_coverage_ranking plot. The fuzzer labels will be in\n+ the order of their coverage.\"\"\"\n+\n+ fuzzer_order = unique_region_cov_df_combined.sort_values(\n+ by='unique_regions_covered', ascending=False).fuzzer\n+\n+ axes = sns.barplot(y='unique_regions_covered',\n+ x='fuzzer',\n+ data=unique_region_cov_df_combined,\n+ order=fuzzer_order,\n+ palette=self._fuzzer_colors,\n+ ax=axes)\n+\n+ for patch in axes.patches:\n+ axes.annotate(\n+ format(patch.get_height(), '.2f'),\n+ (patch.get_x() + patch.get_width() / 2., patch.get_height()),\n+ ha='center',\n+ va='center',\n+ xytext=(0, 10),\n+ textcoords='offset points')\n+\n+ sns.barplot(y='aggregated_edges_covered',\n+ x='fuzzer',\n+ data=unique_region_cov_df_combined,\n+ order=fuzzer_order,\n+ facecolor=(1, 1, 1, 0),\n+ edgecolor='0.2',\n+ ax=axes)\n+\n+ axes.set(ylabel='Reached unique edge coverage')\n+ axes.set(xlabel='Fuzzer (highest coverage on the left)')\n+ axes.set_xticklabels(axes.get_xticklabels(),\n+ rotation=_DEFAULT_LABEL_ROTATION,\n+ horizontalalignment='right')\n+\n+ sns.despine(ax=axes, trim=True)\n+\n+ def write_unique_coverage_ranking_plot(self, unique_region_cov_df_combined,\n+ image_path):\n+ \"\"\"Writes ranking plot for unique coverage.\"\"\"\n+ self._write_plot_to_image(self.unique_coverage_ranking_plot,\n+ unique_region_cov_df_combined, image_path)\n+\n+ def pairwise_unique_coverage_heatmap_plot(self,\n+ pairwise_unique_coverage_table,\n+ axes=None):\n+ \"\"\"Draws the heatmap to visualize the unique coverage between\n+ each pair of fuzzers.\"\"\"\n+ heatmap_args = {\n+ 'annot': True,\n+ 'fmt': 'd',\n+ 'cmap': 'Blues',\n+ 'linewidths': 0.5\n+ }\n+ axes = sns.heatmap(pairwise_unique_coverage_table,\n+ ax=axes,\n+ **heatmap_args)\n+ axes.set(ylabel='Not covered by')\n+ axes.set(xlabel='Covered by')\n+\n+ def write_pairwise_unique_coverage_heatmap_plot(\n+ self, pairwise_unique_coverage_table, image_path):\n+ \"\"\"Writes pairwise unique coverage heatmap plot.\"\"\"\n+ self._write_plot_to_image(self.pairwise_unique_coverage_heatmap_plot,\n+ pairwise_unique_coverage_table, image_path)\n" }, { "change_type": "MODIFY", "old_path": "analysis/queries.py", "new_path": "analysis/queries.py", "diff": "@@ -23,7 +23,7 @@ def get_experiment_data(experiment_names):\n\"\"\"Get measurements (such as coverage) on experiments from the database.\"\"\"\nsnapshots_query = db_utils.query(\n- Experiment.git_hash,\\\n+ Experiment.git_hash, Experiment.experiment_filestore,\\\nTrial.experiment, Trial.fuzzer, Trial.benchmark,\\\nTrial.time_started, Trial.time_ended,\\\nSnapshot.trial_id, Snapshot.time, Snapshot.edges_covered)\\\n" }, { "change_type": "MODIFY", "old_path": "analysis/rendering.py", "new_path": "analysis/rendering.py", "diff": "@@ -38,16 +38,6 @@ def render_report(experiment_results, template, in_progress, coverage_report):\n)\ntemplate = environment.get_template(template)\n- # FIXME: Use |experiment_filestore_name| from experiment db.\n- # See #642: https://github.com/google/fuzzbench/issues/642\n- if 'EXPERIMENT_FILESTORE' in os.environ:\n- experiment_filestore = os.environ['EXPERIMENT_FILESTORE']\n- prefix = \"gs://\"\n- experiment_filestore_name = experiment_filestore[len(prefix):]\n- else:\n- experiment_filestore_name = 'fuzzbench-data'\n-\nreturn template.render(experiment=experiment_results,\nin_progress=in_progress,\n- coverage_report=coverage_report,\n- experiment_filestore_name=experiment_filestore_name)\n+ coverage_report=coverage_report)\n" }, { "change_type": "MODIFY", "old_path": "analysis/report_templates/default.html", "new_path": "analysis/report_templates/default.html", "diff": "</ul>\n{% if coverage_report %}\n+ <ul class=\"collapsible\">\n+ <li>\n+ <div class=\"collapsible-header\">\n+ Unique coverage plots\n+ </div>\n+ <div class=\"collapsible-body\">\n+\n+ <div class=\"row\">\n+ <div class=\"col s6 offset-s3\">\n+ <h5 class=\"center-align\">Ranking by unique edge covered</h4>\n+ <img class=\"responsive-img materialboxed\"\n+ src=\"{{ benchmark.unique_coverage_ranking_plot }}\">\n+ Each bar shows the total number of edges found by a given fuzzer.\n+ The colored area shows the number of unique edges\n+ (i.e., edges that were not covered by any other fuzzers).\n+ </div>\n+ </div> <!-- row -->\n+\n+ <div class=\"row\">\n+ <div class=\"col s6 offset-s3\">\n+ <h5 class=\"center-align\">Pairwise unique coverage</h4>\n+ <img class=\"responsive-img materialboxed\"\n+ src=\"{{ benchmark.pairwise_unique_coverage_plot }}\">\n+ Each cell represents the number of edges covered by the fuzzer\n+ of the column but not by the fuzzer of the row\n+ </div>\n+ </div> <!-- row -->\n+\n+ </div>\n+ </li>\n+ </ul>\n+\n<ul class=\"collapsible\">\n<li>\n<div class=\"collapsible-header\">\n</div>\n<div class=\"collapsible-body\">\n<div class=\"row\">\n- {% for fuzzer in benchmark.fuzzers %}\n+ {% for fuzzer in benchmark.fuzzer_names %}\n<div class=\"col\">\n- <a class=\"waves-effect waves-light btn\" href=\"https://storage.googleapis.com/{{ experiment_filestore_name }}/{{ experiment.name }}/reports/coverage/{{ benchmark.name }}/{{ fuzzer }}/index.html\">{{ fuzzer }}</a>\n+ <a class=\"waves-effect waves-light btn\" href=\"{{ benchmark.get_filestore_name(fuzzer) }}/coverage/reports/{{ benchmark.name }}/{{ fuzzer }}/index.html\">{{ fuzzer }}</a>\n</div>\n{% endfor %}\n</div>\n" }, { "change_type": "ADD", "old_path": null, "new_path": "analysis/test_coverage_data_utils.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions andsss\n+# limitations under the License.\n+\"\"\"Tests for coverage_data_utils.py\"\"\"\n+\n+import pandas as pd\n+import pandas.testing as pd_test\n+\n+from analysis import coverage_data_utils\n+\n+\n+def create_coverage_data():\n+ \"\"\"Utility function to create test data.\"\"\"\n+ return {\n+ \"afl libpng-1.2.56\": [[0, 0, 1, 1], [0, 0, 2, 2], [0, 0, 3, 3]],\n+ \"libfuzzer libpng-1.2.56\": [[0, 0, 1, 1], [0, 0, 2, 3], [0, 0, 3, 3],\n+ [0, 0, 4, 4]]\n+ }\n+\n+\n+def test_get_unique_region_dict():\n+ \"\"\"Tests get_unique_region_dict() function.\"\"\"\n+ coverage_dict = create_coverage_data()\n+ benchmark_coverage_dict = coverage_data_utils.get_benchmark_cov_dict(\n+ coverage_dict, 'libpng-1.2.56')\n+ unique_region_dict = coverage_data_utils.get_unique_region_dict(\n+ benchmark_coverage_dict)\n+ expected_dict = {\n+ (0, 0, 2, 2): ['afl'],\n+ (0, 0, 2, 3): ['libfuzzer'],\n+ (0, 0, 4, 4): ['libfuzzer']\n+ }\n+ assert expected_dict == unique_region_dict\n+\n+\n+def test_get_unique_region_cov_df():\n+ \"\"\"Tests get_unique_region_cov_df() function.\"\"\"\n+ coverage_dict = create_coverage_data()\n+ benchmark_coverage_dict = coverage_data_utils.get_benchmark_cov_dict(\n+ coverage_dict, 'libpng-1.2.56')\n+ unique_region_dict = coverage_data_utils.get_unique_region_dict(\n+ benchmark_coverage_dict)\n+ fuzzer_names = ['afl', 'libfuzzer']\n+ unique_region_df = coverage_data_utils.get_unique_region_cov_df(\n+ unique_region_dict, fuzzer_names)\n+ unique_region_df = unique_region_df.sort_values(by=['fuzzer']).reset_index(\n+ drop=True)\n+ expected_df = pd.DataFrame([{\n+ 'fuzzer': 'afl',\n+ 'unique_regions_covered': 1\n+ }, {\n+ 'fuzzer': 'libfuzzer',\n+ 'unique_regions_covered': 2\n+ }])\n+ assert unique_region_df.equals(expected_df)\n+\n+\n+def test_get_benchmark_cov_dict():\n+ \"\"\"Tests that get_benchmark_cov_dict() returns correct dictionary.\"\"\"\n+ coverage_dict = create_coverage_data()\n+ benchmark = 'libpng-1.2.56'\n+ benchmark_cov_dict = coverage_data_utils.get_benchmark_cov_dict(\n+ coverage_dict, benchmark)\n+ expected_cov_dict = {\n+ \"afl\": {(0, 0, 3, 3), (0, 0, 2, 2), (0, 0, 1, 1)},\n+ \"libfuzzer\": {(0, 0, 4, 4), (0, 0, 3, 3), (0, 0, 2, 3), (0, 0, 1, 1)}\n+ }\n+ assert expected_cov_dict == benchmark_cov_dict\n+\n+\n+def test_get_pairwise_unique_coverage_table():\n+ \"\"\"Tests that get_pairwise_unique_coverage_table() gives the\n+ correct dataframe.\"\"\"\n+ coverage_dict = create_coverage_data()\n+ benchmark_coverage_dict = coverage_data_utils.get_benchmark_cov_dict(\n+ coverage_dict, 'libpng-1.2.56')\n+ table = coverage_data_utils.get_pairwise_unique_coverage_table(\n+ benchmark_coverage_dict)\n+ fuzzers = ['afl', 'libfuzzer']\n+ expected_table = pd.DataFrame([[0, 2], [1, 0]],\n+ index=fuzzers,\n+ columns=fuzzers)\n+ pd_test.assert_frame_equal(table, expected_table)\n" }, { "change_type": "MODIFY", "old_path": "analysis/test_data_utils.py", "new_path": "analysis/test_data_utils.py", "diff": "@@ -23,7 +23,8 @@ from analysis import data_utils\ndef create_trial_data( # pylint: disable=too-many-arguments\n- trial_id, benchmark, fuzzer, cycles, reached_coverage, experiment):\n+ trial_id, benchmark, fuzzer, cycles, reached_coverage, experiment,\n+ experiment_filestore):\n\"\"\"Utility function to create test trial data.\"\"\"\nreturn pd.DataFrame([{\n'experiment': experiment,\n@@ -34,22 +35,31 @@ def create_trial_data( # pylint: disable=too-many-arguments\n'time_ended': None,\n'time': t,\n'edges_covered': reached_coverage,\n+ 'experiment_filestore': experiment_filestore\n} for t in range(cycles)])\n-def create_experiment_data(experiment='test_experiment', incomplete=False):\n+def create_experiment_data(experiment='test_experiment',\n+ incomplete=False,\n+ experiment_filestore='gs://fuzzbench-data'):\n\"\"\"Utility function to create test experiment data.\"\"\"\nreturn pd.concat([\n- create_trial_data(0, 'libpng', 'afl', 10, 100, experiment),\n- create_trial_data(1, 'libpng', 'afl', 10, 200, experiment),\n- create_trial_data(2, 'libpng', 'libfuzzer', 10, 200, experiment),\n- create_trial_data(3, 'libpng', 'libfuzzer', 10, 300, experiment),\n+ create_trial_data(0, 'libpng', 'afl', 10, 100, experiment,\n+ experiment_filestore),\n+ create_trial_data(1, 'libpng', 'afl', 10, 200, experiment,\n+ experiment_filestore),\n+ create_trial_data(2, 'libpng', 'libfuzzer', 10, 200, experiment,\n+ experiment_filestore),\n+ create_trial_data(3, 'libpng', 'libfuzzer', 10, 300, experiment,\n+ experiment_filestore),\ncreate_trial_data(4, 'libxml', 'afl', 6 if incomplete else 10, 1000,\n- experiment),\n- create_trial_data(5, 'libxml', 'afl', 10, 1200, experiment),\n+ experiment, experiment_filestore),\n+ create_trial_data(5, 'libxml', 'afl', 10, 1200, experiment,\n+ experiment_filestore),\ncreate_trial_data(6, 'libxml', 'libfuzzer', 8 if incomplete else 10,\n- 600, experiment),\n- create_trial_data(7, 'libxml', 'libfuzzer', 10, 800, experiment),\n+ 600, experiment, experiment_filestore),\n+ create_trial_data(7, 'libxml', 'libfuzzer', 10, 800, experiment,\n+ experiment_filestore),\n])\n" }, { "change_type": "MODIFY", "old_path": "experiment/coverage_utils.py", "new_path": "experiment/coverage_utils.py", "diff": "import os\nimport multiprocessing\nimport json\n-import queue\nfrom common import experiment_utils as exp_utils\nfrom common import new_process\n@@ -24,23 +23,27 @@ from common import benchmark_utils\nfrom common import fuzzer_utils\nfrom common import logs\nfrom common import filestore_utils\n-from common import experiment_path as exp_path\nfrom common import filesystem\nfrom database import utils as db_utils\nfrom database import models\nfrom experiment.build import build_utils\n-from experiment import reporter\nlogger = logs.Logger('coverage_utils') # pylint: disable=invalid-name\nCOV_DIFF_QUEUE_GET_TIMEOUT = 1\n-def upload_coverage_reports_to_bucket():\n- \"\"\"Copies the coverage reports to gcs bucket.\"\"\"\n- report_dir = reporter.get_reports_dir()\n- src_dir = get_coverage_report_dir()\n- dst_dir = exp_path.filestore(report_dir)\n+def get_coverage_info_dir():\n+ \"\"\"Returns the directory to store coverage information including\n+ coverage report and json summary file.\"\"\"\n+ work_dir = exp_utils.get_work_dir()\n+ return os.path.join(work_dir, 'coverage')\n+\n+\n+def upload_coverage_info_to_bucket():\n+ \"\"\"Copies the coverage reports and json summary files to gcs bucket.\"\"\"\n+ src_dir = get_coverage_info_dir()\n+ dst_dir = exp_utils.get_experiment_filestore_path()\nfilestore_utils.cp(src_dir, dst_dir, recursive=True, parallel=True)\n@@ -55,8 +58,6 @@ def generate_coverage_reports(experiment_config: dict):\nfor benchmark in benchmarks\nfor fuzzer in fuzzers]\npool.starmap(generate_coverage_report, generate_coverage_report_args)\n- pool.close()\n- pool.join()\nlogger.info('Finished generating coverage report.')\n@@ -88,6 +89,9 @@ class CoverageReporter: # pylint: disable=too-many-instance-attributes\nself.trial_ids = get_trial_ids(experiment, fuzzer, benchmark)\ncov_report_directory = get_coverage_report_dir()\nself.report_dir = os.path.join(cov_report_directory, benchmark, fuzzer)\n+ coverage_info_dir = get_coverage_info_dir()\n+ self.json_file_dir = os.path.join(coverage_info_dir, 'data', benchmark,\n+ fuzzer)\nbenchmark_fuzzer_dir = exp_utils.get_benchmark_fuzzer_dir(\nbenchmark, fuzzer)\nwork_dir = exp_utils.get_work_dir()\n@@ -129,17 +133,20 @@ class CoverageReporter: # pylint: disable=too-many-instance-attributes\n'fuzzer: {fuzzer},benchmark: {benchmark}.'.format(\nfuzzer=self.fuzzer, benchmark=self.benchmark))\n+ def generate_json_summary_file(self, covered_regions):\n+ \"\"\"Stores the coverage data in a json file.\"\"\"\n+ json_src_dir = self.json_file_dir\n+ filesystem.create_directory(json_src_dir)\n+ json_src = os.path.join(json_src_dir, 'covered_regions.json')\n+ with open(json_src, 'w') as src_file:\n+ json.dump(covered_regions, src_file)\n+\ndef get_coverage_archive_name(benchmark):\n\"\"\"Gets the archive name for |benchmark|.\"\"\"\nreturn 'coverage-build-%s.tar.gz' % benchmark\n-def get_fuzzer_benchmark_key(fuzzer: str, benchmark: str):\n- \"\"\"Returns the key in coverage dict for a pair of fuzzer-benchmark.\"\"\"\n- return fuzzer + ' ' + benchmark\n-\n-\ndef get_profdata_file_name(trial_id):\n\"\"\"Returns the profdata file name for |trial_id|.\"\"\"\nreturn 'data-{id}.profdata'.format(id=trial_id)\n@@ -147,8 +154,8 @@ def get_profdata_file_name(trial_id):\ndef get_coverage_report_dir():\n\"\"\"Returns the directory to store all the coverage reports.\"\"\"\n- report_dir = reporter.get_reports_dir()\n- return os.path.join(report_dir, 'coverage')\n+ coverage_info_dir = get_coverage_info_dir()\n+ return os.path.join(coverage_info_dir, 'reports')\ndef get_coverage_binary(benchmark: str) -> str:\n@@ -191,65 +198,46 @@ def get_coverage_infomation(coverage_summary_file):\ndef store_coverage_data(experiment_config: dict):\n\"\"\"Generates the specific coverage data and store in cloud bucket.\"\"\"\nlogger.info('Start storing coverage data')\n- with multiprocessing.Pool() as pool, multiprocessing.Manager() as manager:\n- q = manager.Queue() # pytype: disable=attribute-error\n- covered_regions = get_all_covered_regions(experiment_config, pool, q)\n- json_src_dir = reporter.get_reports_dir()\n- filesystem.recreate_directory(json_src_dir)\n- json_src = os.path.join(json_src_dir, 'covered_regions.json')\n- with open(json_src, 'w') as src_file:\n- json.dump(covered_regions, src_file)\n- json_dst = exp_path.filestore(json_src)\n- filestore_utils.cp(json_src, json_dst)\n+ with multiprocessing.Pool() as pool:\n+ store_all_covered_regions(experiment_config, pool)\nlogger.info('Finished storing coverage data')\n-def get_all_covered_regions(experiment_config: dict, pool, q) -> dict:\n- \"\"\"Gets regions covered for each pair for fuzzer and benchmark.\"\"\"\n- logger.info('Measuring all fuzzer-benchmark pairs for final coverage data.')\n+def store_all_covered_regions(experiment_config: dict, pool):\n+ \"\"\"Stores regions covered for each pair for fuzzer and benchmark.\"\"\"\n+ logger.info('Storing all fuzzer-benchmark pairs for final coverage data.')\nbenchmarks = experiment_config['benchmarks'].split(',')\nfuzzers = experiment_config['fuzzers'].split(',')\nexperiment = experiment_config['experiment']\n- get_covered_region_args = [(experiment, fuzzer, benchmark, q)\n+ store_covered_region_args = [(experiment, fuzzer, benchmark)\nfor fuzzer in fuzzers\nfor benchmark in benchmarks]\n- result = pool.starmap_async(get_covered_region, get_covered_region_args)\n+ pool.starmap(store_covered_region, store_covered_region_args)\n+ logger.info('Done storing all coverage data.')\n- # Poll the queue for covered region data and save them in a dict until the\n- # pool is done processing each combination of fuzzers and benchmarks.\n- all_covered_regions = {}\n- while True:\n- try:\n- covered_regions = q.get(timeout=COV_DIFF_QUEUE_GET_TIMEOUT)\n- all_covered_regions.update(covered_regions)\n- except queue.Empty:\n- if result.ready():\n- # If \"ready\" that means pool has finished. Since it is\n- # finished and the queue is empty, we can stop checking\n- # the queue for more covered regions.\n- logger.debug(\n- 'Finished call to map with get_all_covered_regions.')\n- break\n-\n- for key in all_covered_regions:\n- all_covered_regions[key] = list(all_covered_regions[key])\n- logger.info('Done measuring all coverage data.')\n- return all_covered_regions\n-\n-\n-def get_covered_region(experiment: str, fuzzer: str, benchmark: str,\n- q: multiprocessing.Queue):\n+def store_covered_region(experiment: str, fuzzer: str, benchmark: str):\n\"\"\"Gets the final covered region for a specific pair of fuzzer-benchmark.\"\"\"\nlogs.initialize()\n+ logger.debug('Storing covered region: fuzzer: %s, benchmark: %s.', fuzzer,\n+ benchmark)\n+ covered_regions = get_covered_region(experiment, fuzzer, benchmark)\n+ generator = CoverageReporter(fuzzer, benchmark, experiment)\n+ generator.generate_json_summary_file(covered_regions)\n+\n+ logger.debug('Finished storing covered region: fuzzer: %s, benchmark: %s.',\n+ fuzzer, benchmark)\n+\n+\n+def get_covered_region(experiment: str, fuzzer: str, benchmark: str):\n+ \"\"\"Gets covered regions for |fuzzer| on |benchmark|.\"\"\"\nlogger.debug('Measuring covered region: fuzzer: %s, benchmark: %s.', fuzzer,\nbenchmark)\n- key = get_fuzzer_benchmark_key(fuzzer, benchmark)\n- covered_regions = {key: set()}\n+ covered_regions = set()\ntrial_ids = get_trial_ids(experiment, fuzzer, benchmark)\nfor trial_id in trial_ids:\nlogger.info('Measuring covered region: trial_id = %d.', trial_id)\n@@ -263,10 +251,10 @@ def get_covered_region(experiment: str, fuzzer: str, benchmark: str,\nsnapshot_logger)\ntrial_coverage.generate_summary(0, summary_only=False)\nnew_covered_regions = trial_coverage.get_current_covered_regions()\n- covered_regions[key] = covered_regions[key].union(new_covered_regions)\n- q.put(covered_regions)\n+ covered_regions = covered_regions.union(new_covered_regions)\nlogger.debug('Done measuring covered region: fuzzer: %s, benchmark: %s.',\nfuzzer, benchmark)\n+ return list(covered_regions)\nclass TrialCoverage: # pylint: disable=too-many-instance-attributes\n@@ -306,10 +294,13 @@ class TrialCoverage: # pylint: disable=too-many-instance-attributes\n# region it is; 'code_region' is used to obtain various code\n# coverage statistic and is represented by number 0.\ntype_index = -1\n+ # The number of index 5 represents the file number.\n+ file_index = 5\nfor function_data in functions_data:\nfor region in function_data['regions']:\nif region[hit_index] != 0 and region[type_index] == 0:\n- covered_regions.add(tuple(region[:hit_index]))\n+ covered_regions.add(\n+ tuple(region[:hit_index] + region[file_index:]))\nexcept Exception: # pylint: disable=broad-except\nself.logger.error(\n'Coverage summary json file defective or missing.')\n" }, { "change_type": "MODIFY", "old_path": "experiment/measurer.py", "new_path": "experiment/measurer.py", "diff": "@@ -72,7 +72,8 @@ def measure_main(experiment_config):\n# Do the final measuring and store the coverage data.\ncoverage_utils.store_coverage_data(experiment_config)\ncoverage_utils.generate_coverage_reports(experiment_config)\n- coverage_utils.upload_coverage_reports_to_bucket()\n+ coverage_utils.upload_coverage_info_to_bucket()\n+\nlogger.info('Finished measuring.')\n" } ]

[ { "change_type": "MODIFY", "old_path": "common/gce.py", "new_path": "common/gce.py", "diff": "\"\"\"Module for using the Google Compute Engine (GCE) API.\"\"\"\nimport threading\n-import dateutil.parser\n-\nfrom googleapiclient import discovery\nfrom oauth2client.client import GoogleCredentials\n@@ -31,60 +29,27 @@ def initialize():\ncredentials=credentials)\n-def get_operations(project, zone):\n- \"\"\"Generator that yields GCE operations for compute engine |project| and\n- |zone| in descendending order by time.\"\"\"\n- zone_operations = thread_local.service.zoneOperations() # pylint: disable=no-member\n- request = zone_operations.list(project=project,\n- zone=zone,\n- orderBy='creationTimestamp desc')\n+def _get_instance_items(project, zone):\n+ \"\"\"Return an iterator of all instance response items for a project.\"\"\"\n+ instances = thread_local.service.instances()\n+ request = instances.list(project=project, zone=zone)\nwhile request is not None:\nresponse = request.execute()\n- for operation in response['items']:\n- yield operation\n-\n- request = zone_operations.list_next(previous_request=request,\n+ for instance in response['items']:\n+ yield instance\n+ request = instances.list_next(previous_request=request,\nprevious_response=response)\n-def get_preempted_operations(operations):\n- \"\"\"Generator that yields GCE preempted operations in |operations|.\"\"\"\n- # This endpoint doesn't support filtering by time (despite implications\n- # otherwise). Instead it supports ordering by time. It also supports\n- # filtering by operation but doesn't support it when combined with ordering.\n- # So we must filter manually.\n- # See the link below for an example of this query in action.\n- # https://cloud.google.com/compute/docs/reference/rest/v1/zoneOperations/list?apix_params=%7B%22project%22%3A%22fuzzbench%22%2C%22zone%22%3A%22us-central1-a%22%2C%22filter%22%3A%22(operationType%20%3D%20%5C%22compute.instances.preempted%5C%22)%22%2C%22orderBy%22%3A%22creationTimestamp%20desc%22%7D\n- for operation in operations:\n- if operation['operationType'] == 'compute.instances.preempted':\n- yield operation\n-\n-\n-def filter_by_end_time(min_end_time, operations):\n- \"\"\"Generator that yields GCE operations in |operations| that finished before\n- |min_end_time|. |operations| must be an iterable that is ordered by time.\"\"\"\n- for operation in operations:\n- end_time = operation.get('endTime')\n- if not end_time:\n- # Try to handle cases where the operation hasn't finished.\n- yield operation\n- continue\n-\n- operation_end_time = dateutil.parser.isoparse(end_time)\n- if operation_end_time < min_end_time:\n- break\n- yield operation\n-\n-\n-def get_base_target_link(experiment_config):\n- \"\"\"Returns the base of the target link for this experiment so that\n- get_instance_from_preempted_operation can return the instance.\"\"\"\n- return ('https://www.googleapis.com/compute/v1/projects/{project}/zones/'\n- '{zone}/instances/').format(\n- project=experiment_config['cloud_project'],\n- zone=experiment_config['cloud_compute_zone'])\n+def get_instances(project, zone):\n+ \"\"\"Return a list of all instance names in |project| and |zone|.\"\"\"\n+ for instance in _get_instance_items(project, zone):\n+ yield instance['name']\n-def get_instance_from_preempted_operation(operation, base_target_link) -> str:\n- \"\"\"Returns the instance name from a preempted |operation|.\"\"\"\n- return operation['targetLink'][len(base_target_link):]\n+def get_preempted_instances(project, zone):\n+ \"\"\"Return a list of preempted instance names in |project| and |zone|.\"\"\"\n+ for instance in _get_instance_items(project, zone):\n+ if (instance['scheduling']['preemptible'] and\n+ instance['status'] == 'TERMINATED'):\n+ yield instance['name']\n" }, { "change_type": "MODIFY", "old_path": "common/gcloud.py", "new_path": "common/gcloud.py", "diff": "@@ -100,12 +100,6 @@ def delete_instances(instance_names: List[str], zone: str, **kwargs) -> bool:\nreturn not error_occurred\n-def list_instances() -> List[str]:\n- \"\"\"Return list of current running instances.\"\"\"\n- result = new_process.execute(['gcloud', 'compute', 'instances', 'list'])\n- return [instance.split(' ')[0] for instance in result.output.splitlines()]\n-\n-\ndef set_default_project(cloud_project: str):\n\"\"\"Set default project for future gcloud and gsutil commands.\"\"\"\nreturn new_process.execute(\n" }, { "change_type": "DELETE", "old_path": "common/test_gce.py", "new_path": null, "diff": "-# Copyright 2020 Google LLC\n-#\n-# Licensed under the Apache License, Version 2.0 (the \"License\");\n-# you may not use this file except in compliance with the License.\n-# You may obtain a copy of the License at\n-#\n-# http://www.apache.org/licenses/LICENSE-2.0\n-#\n-# Unless required by applicable law or agreed to in writing, software\n-# distributed under the License is distributed on an \"AS IS\" BASIS,\n-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n-# See the License for the specific language governing permissions and\n-# limitations under the License.\n-\"\"\"Tests for gce.py.\"\"\"\n-import posixpath\n-\n-from common import gce\n-\n-\n-def test_get_instance_from_preempted_operation():\n- \"\"\"Tests that _get_instance_from_preemption_operation returns the correct\n- value.\"\"\"\n- expected_instance = 'r-my-experiment-100'\n- base_target_link = 'www.myresourceurl/'\n- target_link = posixpath.join(base_target_link, expected_instance)\n- operation = {'targetLink': target_link}\n- instance = gce.get_instance_from_preempted_operation(\n- operation, base_target_link)\n-\n- assert instance == expected_instance\n" }, { "change_type": "MODIFY", "old_path": "experiment/scheduler.py", "new_path": "experiment/scheduler.py", "diff": "@@ -106,8 +106,12 @@ def all_trials_ended(experiment: str) -> bool:\ndef delete_instances(instances, experiment_config):\n\"\"\"Deletes |instances|.\"\"\"\n- running_instances = gcloud.list_instances()\n- instances_to_delete = [i for i in instances if i in running_instances]\n+ cloud_project = experiment_config['cloud_project']\n+ cloud_compute_zone = experiment_config['cloud_compute_zone']\n+ instances_to_delete = [\n+ i for i in gce.get_instances(cloud_project, cloud_compute_zone)\n+ if i in instances\n+ ]\nreturn gcloud.delete_instances(instances_to_delete,\nexperiment_config['cloud_compute_zone'])\n@@ -231,8 +235,6 @@ class TrialInstanceManager: # pylint: disable=too-many-instance-attributes\nself.preempted_trials = {}\nself.preemptible_starts_futile = False\n- self.base_target_link = gce.get_base_target_link(experiment_config)\n-\n# Filter operations happening before the experiment started.\nself.last_preemptible_query = (db_utils.query(models.Experiment).filter(\nmodels.Experiment.name == experiment_config['experiment']).one(\n@@ -410,7 +412,7 @@ class TrialInstanceManager: # pylint: disable=too-many-instance-attributes\nstarted_instances = self._get_started_unfinished_instances()\nquery_time = datetime_now()\n- preempted_instances = list(self._query_preempted_instances())\n+ preempted_instances = self._get_preempted_instances_with_retries()\ntrials = []\nfor instance in preempted_instances:\ntrial = started_instances.get(instance)\n@@ -428,30 +430,20 @@ class TrialInstanceManager: # pylint: disable=too-many-instance-attributes\n# Update this now when we know that we have succeded processing the\n# query. It's far worse if we update the query too early than if we\n# don't update the query at this point (which will only result in\n- # redundant work.\n+ # redundant work).\nself.last_preemptible_query = query_time\n# Return all preempted instances, those we knew from beforehand and\n# those we discovered in the query.\nreturn trials\n- @retry.wrap(\n- NUM_RETRIES, RETRY_WAIT_SECONDS,\n- 'experiment.scheduler.TrialInstanceManager._query_preempted_instances')\n- def _query_preempted_instances(self):\n+ @retry.wrap(NUM_RETRIES, RETRY_WAIT_SECONDS,\n+ 'experiment.scheduler.TrialInstanceManager.'\n+ '_get_preempted_instances_with_retries')\n+ def _get_preempted_instances_with_retries(self):\nproject = self.experiment_config['cloud_project']\nzone = self.experiment_config['cloud_compute_zone']\n- operations = gce.filter_by_end_time(self.last_preemptible_query,\n- gce.get_operations(project, zone))\n- instances = []\n- for operation in gce.get_preempted_operations(operations):\n- if operation is None:\n- logs.error('Operation is None.')\n- continue\n- instance = gce.get_instance_from_preempted_operation(\n- operation, self.base_target_link)\n- instances.append(instance)\n- return instances\n+ return list(gce.get_preempted_instances(project, zone))\ndef handle_preempted_trials(self):\n\"\"\"Handle preempted trials by marking them as preempted and creating\n" }, { "change_type": "MODIFY", "old_path": "experiment/stop_experiment.py", "new_path": "experiment/stop_experiment.py", "diff": "@@ -18,6 +18,7 @@ import sys\nfrom common import experiment_utils\nfrom common import logs\n+from common import gce\nfrom common import gcloud\nfrom common import yaml_utils\n@@ -27,14 +28,16 @@ logger = logs.Logger('stop_experiment') # pylint: disable=invalid-name\ndef stop_experiment(experiment_name, experiment_config_filename):\n\"\"\"Stop the experiment specified by |experiment_config_filename|.\"\"\"\nexperiment_config = yaml_utils.read(experiment_config_filename)\n-\nif experiment_config.get('local_experiment', False):\nraise NotImplementedError(\n'Local experiment stop logic is not implemented.')\n- instances = gcloud.list_instances()\n-\n+ cloud_project = experiment_config['cloud_project']\ncloud_compute_zone = experiment_config['cloud_compute_zone']\n+\n+ gce.initialize()\n+ instances = list(gce.get_instances(cloud_project, cloud_compute_zone))\n+\ntrial_prefix = 'r-' + experiment_name\nexperiment_instances = [\ninstance for instance in instances if instance.startswith(trial_prefix)\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_scheduler.py", "new_path": "experiment/test_scheduler.py", "diff": "import datetime\nfrom multiprocessing.pool import ThreadPool\nimport os\n-import posixpath\nimport time\nfrom unittest import mock\n@@ -215,7 +214,8 @@ def test_start_trials_not_started(mocked_create_instance, pending_trials,\[email protected]('experiment.scheduler.datetime_now')\[email protected]('common.benchmark_utils.get_fuzz_target',\nreturn_value='fuzz-target')\n-def test_schedule(_, mocked_datetime_now, mocked_execute, pending_trials,\[email protected]('common.gce._get_instance_items', return_value=[])\n+def test_schedule(_, __, mocked_datetime_now, mocked_execute, pending_trials,\nexperiment_config):\n\"\"\"Tests that schedule() ends expired trials and starts new ones as\nneeded.\"\"\"\n@@ -376,7 +376,7 @@ def test_get_preempted_trials_nonpreemptible(experiment_config, db):\nassert trial_instance_manager.get_preempted_trials() == []\[email protected]('common.gce.get_operations', return_value=[])\[email protected]('common.gce._get_instance_items', return_value=[])\ndef test_get_preempted_trials_stale_preempted(_, preempt_exp_conf):\n\"\"\"Tests that TrialInstanceManager.get_preempted_trials doesn't return\ntrials that we already know were preempted.\"\"\"\n@@ -395,37 +395,21 @@ def test_get_preempted_trials_stale_preempted(_, preempt_exp_conf):\nassert trial_instance_manager.get_preempted_trials() == []\n-def _get_preemption_operation(trial_id, exp_conf):\n- zone_url = (\n- 'https://www.googleapis.com/compute/v1/projects/{project}/zones/'\n- '{zone}').format(zone=exp_conf['cloud_compute_zone'],\n- project=exp_conf['cloud_project'])\n+def _get_preempted_instance_item(trial_id, exp_conf):\ninstance_name = experiment_utils.get_trial_instance_name(\nexp_conf['experiment'], trial_id)\n- target_link = posixpath.join('instances', zone_url, instance_name)\n- name = 'systemevent-blah'\n- self_link = posixpath.join(zone_url, name)\nreturn {\n'id': '1',\n- 'name': name,\n- 'zone': zone_url,\n- 'operationType': 'compute.instances.preempted',\n- 'targetLink': target_link,\n- 'targetId': '1',\n- 'status': 'DONE',\n- 'statusMessage': 'Instance was preempted.',\n- 'user': 'system',\n- 'progress': 100,\n- 'insertTime': '2020-01-24T29:16:46.842-02:00',\n- 'startTime': '2020-01-24T29:16:46.842-02:00',\n- 'endTime': '2020-01-24T29:16:46.842-02:00',\n- 'selfLink': self_link,\n- 'kind': 'compute#operation'\n+ 'name': instance_name,\n+ 'status': 'TERMINATED',\n+ 'scheduling': {\n+ 'preemptible': True,\n+ }\n}\[email protected]('common.gce.get_preempted_operations')\n-def test_get_preempted_trials_new_preempted(mocked_get_preempted_operations,\[email protected]('common.gce._get_instance_items')\n+def test_get_preempted_trials_new_preempted(mocked_get_instance_items,\npreempt_exp_conf):\n\"\"\"Tests that TrialInstanceManager.get_preempted_trials returns trials that\nnew preempted trials we don't know about until we query for them and not\n@@ -445,8 +429,8 @@ def test_get_preempted_trials_new_preempted(mocked_get_preempted_operations,\ntime_started=time_started)\ntrials = [known_preempted, unknown_preempted]\ndb_utils.add_all(trials)\n- mocked_get_preempted_operations.return_value = [\n- _get_preemption_operation(trial.id, preempt_exp_conf)\n+ mocked_get_instance_items.return_value = [\n+ _get_preempted_instance_item(trial.id, preempt_exp_conf)\nfor trial in trials\n]\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/fuzzers.yml", "new_path": ".github/workflows/fuzzers.yml", "diff": "@@ -37,6 +37,7 @@ jobs:\n# Temporary variants.\n- libfuzzer_keepseed\n- entropic_keepseed\n+ - entropic_exectime\n- aflplusplus_classic\n- aflplusplus_lto_pcguard\n- aflplusplus_lto\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_exectime/builder.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+ARG parent_image\n+FROM $parent_image\n+\n+COPY patch.diff /\n+\n+RUN git clone https://github.com/llvm/llvm-project.git /llvm-project && \\\n+ cd /llvm-project && \\\n+ git checkout bb54bcf84970c04c9748004f3a4cf59b0c1832a7 && \\\n+ patch -p1 < /patch.diff && \\\n+ cd /llvm-project/compiler-rt/lib/fuzzer && \\\n+ (for f in *.cpp; do \\\n+ clang++ -stdlib=libc++ -fPIC -O2 -std=c++11 $f -c & \\\n+ done && wait) && \\\n+ ar r /libEntropic.a *.o\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_exectime/fuzzer.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Integration code for libFuzzer fuzzer.\"\"\"\n+\n+from fuzzers.entropic import fuzzer as entropic_fuzzer\n+from fuzzers.libfuzzer import fuzzer as libfuzzer_fuzzer\n+\n+\n+def build():\n+ \"\"\"Build benchmark.\"\"\"\n+ entropic_fuzzer.build()\n+\n+\n+def fuzz(input_corpus, output_corpus, target_binary):\n+ \"\"\"Run fuzzer.\"\"\"\n+ libfuzzer_fuzzer.run_fuzzer(\n+ input_corpus,\n+ output_corpus,\n+ target_binary,\n+ extra_flags=['-entropic=1', '-entropic_scale_per_exec_time=1'])\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_exectime/patch.diff", "diff": "+commit cf575f9bb0d180fa2069c30b6b78f8ada3502462\n+Author: Dokyung Song <[email protected]>\n+Date: Mon Aug 17 16:59:59 2020 +0000\n+\n+ [libFuzzer] Scale energy assigned to each input based on input execution time.\n+\n+ This is an experimental patch uploaded to get early feedback.\n+\n+ This patch scales the energy computed by the Entropic schedule based on the\n+ execution time of each input. The input execution time is compared with the\n+ average execution time of inputs in the corpus, and, based on the amount by\n+ which they differ, the energy is scaled from 0.1x (for inputs executing slow) to\n+ 30x (for inputs executing fast). Note that the exact formula is borrowed from\n+ AFL.\n+\n+ In my short, local experiments, this gives a sizeable throughput increase (which\n+ in turn leads to more coverage) on the SQLITE3 benchmark, which I will test\n+ further with longer and more experiments with more benchmarks. I am not sure for\n+ now that how this execution-time-based energy scaling works with other factors\n+ (e.g., the number of globally rare features) that are already considered by the\n+ Entropic schedule to compute the energy; I am going to do some more profiling\n+ and testing to examine this.\n+\n+ Differential Revision: https://reviews.llvm.org/D86092\n+\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerCorpus.h b/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n+index 54d1e09ec6d..66f137e9c7c 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n++++ b/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n+@@ -18,6 +18,7 @@\n+ #include \"FuzzerSHA1.h\"\n+ #include \"FuzzerTracePC.h\"\n+ #include <algorithm>\n++#include <chrono>\n+ #include <numeric>\n+ #include <random>\n+ #include <unordered_set>\n+@@ -26,6 +27,7 @@ namespace fuzzer {\n+\n+ struct InputInfo {\n+ Unit U; // The actual input data.\n++ std::chrono::microseconds TimeOfUnit;\n+ uint8_t Sha1[kSHA1NumBytes]; // Checksum.\n+ // Number of features that this input has and no smaller input has.\n+ size_t NumFeatures = 0;\n+@@ -65,7 +67,8 @@ struct InputInfo {\n+ // of the seed. Since we do not know the entropy of a seed that has\n+ // never been executed we assign fresh seeds maximum entropy and\n+ // let II->Energy approach the true entropy from above.\n+- void UpdateEnergy(size_t GlobalNumberOfFeatures) {\n++ void UpdateEnergy(size_t GlobalNumberOfFeatures, bool ScalePerExecTime,\n++ uint64_t AverageTimeOfUnit) {\n+ Energy = 0.0;\n+ SumIncidence = 0;\n+\n+@@ -88,6 +91,27 @@ struct InputInfo {\n+ // Normalize.\n+ if (SumIncidence != 0)\n+ Energy = (Energy / SumIncidence) + logl(SumIncidence);\n++\n++ if (ScalePerExecTime) {\n++ // Scaling to favor inputs with lower execution time.\n++ uint32_t PerfScore = 100;\n++ if (TimeOfUnit.count() * 0.1 > AverageTimeOfUnit)\n++ PerfScore = 10;\n++ else if (TimeOfUnit.count() * 0.25 > AverageTimeOfUnit)\n++ PerfScore = 25;\n++ else if (TimeOfUnit.count() * 0.5 > AverageTimeOfUnit)\n++ PerfScore = 50;\n++ else if (TimeOfUnit.count() * 0.75 > AverageTimeOfUnit)\n++ PerfScore = 75;\n++ else if (TimeOfUnit.count() * 4 < AverageTimeOfUnit)\n++ PerfScore = 300;\n++ else if (TimeOfUnit.count() * 3 < AverageTimeOfUnit)\n++ PerfScore = 200;\n++ else if (TimeOfUnit.count() * 2 < AverageTimeOfUnit)\n++ PerfScore = 150;\n++\n++ Energy *= PerfScore;\n++ }\n+ }\n+\n+ // Increment the frequency of the feature Idx.\n+@@ -120,6 +144,7 @@ struct EntropicOptions {\n+ bool Enabled;\n+ size_t NumberOfRarestFeatures;\n+ size_t FeatureFrequencyThreshold;\n++ bool ScalePerExecTime;\n+ };\n+\n+ class InputCorpus {\n+@@ -178,6 +203,7 @@ public:\n+ const Unit &operator[] (size_t Idx) const { return Inputs[Idx]->U; }\n+ InputInfo *AddToCorpus(const Unit &U, size_t NumFeatures, bool MayDeleteFile,\n+ bool HasFocusFunction,\n++ std::chrono::microseconds TimeOfUnit,\n+ const Vector<uint32_t> &FeatureSet,\n+ const DataFlowTrace &DFT, const InputInfo *BaseII) {\n+ assert(!U.empty());\n+@@ -187,6 +213,7 @@ public:\n+ InputInfo &II = *Inputs.back();\n+ II.U = U;\n+ II.NumFeatures = NumFeatures;\n++ II.TimeOfUnit = TimeOfUnit;\n+ II.MayDeleteFile = MayDeleteFile;\n+ II.UniqFeatureSet = FeatureSet;\n+ II.HasFocusFunction = HasFocusFunction;\n+@@ -460,12 +487,18 @@ private:\n+ Weights.resize(N);\n+ std::iota(Intervals.begin(), Intervals.end(), 0);\n+\n++ std::chrono::microseconds TotalTimeOfUnit(0);\n++ for (auto II : Inputs) {\n++ TotalTimeOfUnit += II->TimeOfUnit;\n++ }\n++\n+ bool VanillaSchedule = true;\n+ if (Entropic.Enabled) {\n+ for (auto II : Inputs) {\n+ if (II->NeedsEnergyUpdate && II->Energy != 0.0) {\n+ II->NeedsEnergyUpdate = false;\n+- II->UpdateEnergy(RareFeatures.size());\n++ II->UpdateEnergy(RareFeatures.size(), Entropic.ScalePerExecTime,\n++ TotalTimeOfUnit.count() / N);\n+ }\n+ }\n+\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerDriver.cpp b/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n+index bed9e84de67..d31783debff 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n++++ b/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n+@@ -718,6 +718,7 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n+ (size_t)Flags.entropic_feature_frequency_threshold;\n+ Options.EntropicNumberOfRarestFeatures =\n+ (size_t)Flags.entropic_number_of_rarest_features;\n++ Options.EntropicScalePerExecTime = Flags.entropic_scale_per_exec_time;\n+ if (Options.Entropic) {\n+ if (!Options.FocusFunction.empty()) {\n+ Printf(\"ERROR: The parameters `--entropic` and `--focus_function` cannot \"\n+@@ -733,6 +734,7 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n+ Entropic.FeatureFrequencyThreshold =\n+ Options.EntropicFeatureFrequencyThreshold;\n+ Entropic.NumberOfRarestFeatures = Options.EntropicNumberOfRarestFeatures;\n++ Entropic.ScalePerExecTime = Options.EntropicScalePerExecTime;\n+\n+ unsigned Seed = Flags.seed;\n+ // Initialize Seed.\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerFlags.def b/compiler-rt/lib/fuzzer/FuzzerFlags.def\n+index 832224a705d..dda66e012a8 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerFlags.def\n++++ b/compiler-rt/lib/fuzzer/FuzzerFlags.def\n+@@ -161,6 +161,7 @@ FUZZER_FLAG_INT(entropic_number_of_rarest_features, 100, \"Experimental. If \"\n+ \"entropic is enabled, we keep track of the frequencies only for the \"\n+ \"Top-X least abundant features (union features that are considered as \"\n+ \"rare).\")\n++FUZZER_FLAG_INT(entropic_scale_per_exec_time, 0, \"Experimental.\")\n+\n+ FUZZER_FLAG_INT(analyze_dict, 0, \"Experimental\")\n+ FUZZER_DEPRECATED_FLAG(use_clang_coverage)\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerLoop.cpp b/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n+index 02db6d27b0a..9bb788c4efe 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n++++ b/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n+@@ -469,6 +469,7 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\n+ return false;\n+\n+ ExecuteCallback(Data, Size);\n++ auto TimeOfUnit = duration_cast<microseconds>(UnitStopTime - UnitStartTime);\n+\n+ UniqFeatureSetTmp.clear();\n+ size_t FoundUniqFeaturesOfII = 0;\n+@@ -491,7 +492,7 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\n+ TPC.UpdateObservedPCs();\n+ auto NewII = Corpus.AddToCorpus({Data, Data + Size}, NumNewFeatures,\n+ MayDeleteFile, TPC.ObservedFocusFunction(),\n+- UniqFeatureSetTmp, DFT, II);\n++ TimeOfUnit, UniqFeatureSetTmp, DFT, II);\n+ WriteFeatureSetToFile(Options.FeaturesDir, Sha1ToString(NewII->Sha1),\n+ NewII->UniqFeatureSet);\n+ return true;\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerOptions.h b/compiler-rt/lib/fuzzer/FuzzerOptions.h\n+index b75e7c7af70..ed309eabf5c 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerOptions.h\n++++ b/compiler-rt/lib/fuzzer/FuzzerOptions.h\n+@@ -47,6 +47,7 @@ struct FuzzingOptions {\n+ bool Entropic = false;\n+ size_t EntropicFeatureFrequencyThreshold = 0xFF;\n+ size_t EntropicNumberOfRarestFeatures = 100;\n++ bool EntropicScalePerExecTime = false;\n+ std::string OutputCorpus;\n+ std::string ArtifactPrefix = \"./\";\n+ std::string ExactArtifactPath;\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_exectime/runner.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+FROM gcr.io/fuzzbench/base-runner\n" }, { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-08-21\n+ fuzzers:\n+ - entropic_exectime\n+\n- experiment: 2020-08-20\nfuzzers:\n- aflplusplus\n" } ]

[ { "change_type": "MODIFY", "old_path": "fuzzers/entropic_keepseed/fuzzer.py", "new_path": "fuzzers/entropic_keepseed/fuzzer.py", "diff": "@@ -29,5 +29,6 @@ def fuzz(input_corpus, output_corpus, target_binary):\ntarget_binary,\nextra_flags=[\n'-entropic=1', '-keep_seed=1',\n- '-cross_over_uniformdist=1'\n+ '-cross_over_uniformdist=1',\n+ '-entropic_scale_per_exec_time=1'\n])\n" }, { "change_type": "MODIFY", "old_path": "fuzzers/entropic_keepseed/patch.diff", "new_path": "fuzzers/entropic_keepseed/patch.diff", "diff": "-commit 72e16fc7160185305eee536c257a478ae84f7082\n-Author: Dokyung Song <[email protected]>\n-Date: Fri Jul 31 00:07:20 2020 +0000\n-\n- [libFuzzer] Optionally keep initial seed inputs regardless of whether they discover new features or not.\n-\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerCorpus.h b/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n-index 54d1e09ec6d..5c687013c59 100644\n+index 54d1e09ec6d..ed7d0837659 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n+++ b/compiler-rt/lib/fuzzer/FuzzerCorpus.h\n-@@ -33,6 +33,7 @@ struct InputInfo {\n+@@ -18,6 +18,7 @@\n+ #include \"FuzzerSHA1.h\"\n+ #include \"FuzzerTracePC.h\"\n+ #include <algorithm>\n++#include <chrono>\n+ #include <numeric>\n+ #include <random>\n+ #include <unordered_set>\n+@@ -26,6 +27,7 @@ namespace fuzzer {\n+\n+ struct InputInfo {\n+ Unit U; // The actual input data.\n++ std::chrono::microseconds TimeOfUnit;\n+ uint8_t Sha1[kSHA1NumBytes]; // Checksum.\n+ // Number of features that this input has and no smaller input has.\n+ size_t NumFeatures = 0;\n+@@ -33,6 +35,7 @@ struct InputInfo {\n// Stats.\nsize_t NumExecutedMutations = 0;\nsize_t NumSuccessfullMutations = 0;\n@@ -16,7 +26,53 @@ index 54d1e09ec6d..5c687013c59 100644\nbool MayDeleteFile = false;\nbool Reduced = false;\nbool HasFocusFunction = false;\n-@@ -131,9 +132,12 @@ class InputCorpus {\n+@@ -65,7 +68,8 @@ struct InputInfo {\n+ // of the seed. Since we do not know the entropy of a seed that has\n+ // never been executed we assign fresh seeds maximum entropy and\n+ // let II->Energy approach the true entropy from above.\n+- void UpdateEnergy(size_t GlobalNumberOfFeatures) {\n++ void UpdateEnergy(size_t GlobalNumberOfFeatures, bool ScalePerExecTime,\n++ uint64_t AverageTimeOfUnit) {\n+ Energy = 0.0;\n+ SumIncidence = 0;\n+\n+@@ -88,6 +92,27 @@ struct InputInfo {\n+ // Normalize.\n+ if (SumIncidence != 0)\n+ Energy = (Energy / SumIncidence) + logl(SumIncidence);\n++\n++ if (ScalePerExecTime) {\n++ // Scaling to favor inputs with lower execution time.\n++ uint32_t PerfScore = 100;\n++ if (TimeOfUnit.count() * 0.1 > AverageTimeOfUnit)\n++ PerfScore = 10;\n++ else if (TimeOfUnit.count() * 0.25 > AverageTimeOfUnit)\n++ PerfScore = 25;\n++ else if (TimeOfUnit.count() * 0.5 > AverageTimeOfUnit)\n++ PerfScore = 50;\n++ else if (TimeOfUnit.count() * 0.75 > AverageTimeOfUnit)\n++ PerfScore = 75;\n++ else if (TimeOfUnit.count() * 4 < AverageTimeOfUnit)\n++ PerfScore = 300;\n++ else if (TimeOfUnit.count() * 3 < AverageTimeOfUnit)\n++ PerfScore = 200;\n++ else if (TimeOfUnit.count() * 2 < AverageTimeOfUnit)\n++ PerfScore = 150;\n++\n++ Energy *= PerfScore;\n++ }\n+ }\n+\n+ // Increment the frequency of the feature Idx.\n+@@ -120,6 +145,7 @@ struct EntropicOptions {\n+ bool Enabled;\n+ size_t NumberOfRarestFeatures;\n+ size_t FeatureFrequencyThreshold;\n++ bool ScalePerExecTime;\n+ };\n+\n+ class InputCorpus {\n+@@ -131,9 +157,12 @@ class InputCorpus {\nEntropicOptions Entropic;\n@@ -31,24 +87,26 @@ index 54d1e09ec6d..5c687013c59 100644\nmemset(InputSizesPerFeature, 0, sizeof(InputSizesPerFeature));\nmemset(SmallestElementPerFeature, 0, sizeof(SmallestElementPerFeature));\n}\n-@@ -177,7 +181,7 @@ public:\n+@@ -177,7 +206,8 @@ public:\nbool empty() const { return Inputs.empty(); }\nconst Unit &operator[] (size_t Idx) const { return Inputs[Idx]->U; }\nInputInfo *AddToCorpus(const Unit &U, size_t NumFeatures, bool MayDeleteFile,\n- bool HasFocusFunction,\n+ bool HasFocusFunction, bool SeedInput,\n++ std::chrono::microseconds TimeOfUnit,\nconst Vector<uint32_t> &FeatureSet,\nconst DataFlowTrace &DFT, const InputInfo *BaseII) {\nassert(!U.empty());\n-@@ -187,6 +191,7 @@ public:\n+@@ -187,6 +217,8 @@ public:\nInputInfo &II = *Inputs.back();\nII.U = U;\nII.NumFeatures = NumFeatures;\n+ II.SeedInput = SeedInput;\n++ II.TimeOfUnit = TimeOfUnit;\nII.MayDeleteFile = MayDeleteFile;\nII.UniqFeatureSet = FeatureSet;\nII.HasFocusFunction = HasFocusFunction;\n-@@ -276,6 +281,11 @@ public:\n+@@ -276,6 +308,11 @@ public:\nreturn Idx;\n}\n@@ -60,8 +118,28 @@ index 54d1e09ec6d..5c687013c59 100644\nvoid PrintStats() {\nfor (size_t i = 0; i < Inputs.size(); i++) {\nconst auto &II = *Inputs[i];\n+@@ -460,12 +497,18 @@ private:\n+ Weights.resize(N);\n+ std::iota(Intervals.begin(), Intervals.end(), 0);\n+\n++ std::chrono::microseconds TotalTimeOfUnit(0);\n++ for (auto II : Inputs) {\n++ TotalTimeOfUnit += II->TimeOfUnit;\n++ }\n++\n+ bool VanillaSchedule = true;\n+ if (Entropic.Enabled) {\n+ for (auto II : Inputs) {\n+ if (II->NeedsEnergyUpdate && II->Energy != 0.0) {\n+ II->NeedsEnergyUpdate = false;\n+- II->UpdateEnergy(RareFeatures.size());\n++ II->UpdateEnergy(RareFeatures.size(), Entropic.ScalePerExecTime,\n++ TotalTimeOfUnit.count() / N);\n+ }\n+ }\n+\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerDriver.cpp b/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n-index 8339697396c..0933af56804 100644\n+index bed9e84de67..c11e6a0084a 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n+++ b/compiler-rt/lib/fuzzer/FuzzerDriver.cpp\n@@ -649,6 +649,7 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n@@ -81,7 +159,23 @@ index 8339697396c..0933af56804 100644\nOptions.MutateDepth = Flags.mutate_depth;\nOptions.ReduceDepth = Flags.reduce_depth;\nOptions.UseCounters = Flags.use_counters;\n-@@ -753,7 +756,8 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n+@@ -718,6 +721,7 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n+ (size_t)Flags.entropic_feature_frequency_threshold;\n+ Options.EntropicNumberOfRarestFeatures =\n+ (size_t)Flags.entropic_number_of_rarest_features;\n++ Options.EntropicScalePerExecTime = Flags.entropic_scale_per_exec_time;\n+ if (Options.Entropic) {\n+ if (!Options.FocusFunction.empty()) {\n+ Printf(\"ERROR: The parameters `--entropic` and `--focus_function` cannot \"\n+@@ -733,6 +737,7 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\n+ Entropic.FeatureFrequencyThreshold =\n+ Options.EntropicFeatureFrequencyThreshold;\n+ Entropic.NumberOfRarestFeatures = Options.EntropicNumberOfRarestFeatures;\n++ Entropic.ScalePerExecTime = Options.EntropicScalePerExecTime;\n+\n+ unsigned Seed = Flags.seed;\n+ // Initialize Seed.\n+@@ -753,7 +758,8 @@ int FuzzerDriver(int *argc, char ***argv, UserCallback Callback) {\nRandom Rand(Seed);\nauto *MD = new MutationDispatcher(Rand, Options);\n@@ -92,7 +186,7 @@ index 8339697396c..0933af56804 100644\nfor (auto &U: Dictionary)\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerFlags.def b/compiler-rt/lib/fuzzer/FuzzerFlags.def\n-index 832224a705d..10b1f5f539a 100644\n+index 832224a705d..3f66242985d 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerFlags.def\n+++ b/compiler-rt/lib/fuzzer/FuzzerFlags.def\n@@ -23,7 +23,14 @@ FUZZER_FLAG_INT(len_control, 100, \"Try generating small inputs first, \"\n@@ -110,6 +204,14 @@ index 832224a705d..10b1f5f539a 100644\nFUZZER_FLAG_INT(mutate_depth, 5,\n\"Apply this number of consecutive mutations to each input.\")\nFUZZER_FLAG_INT(reduce_depth, 0, \"Experimental/internal. \"\n+@@ -161,6 +168,7 @@ FUZZER_FLAG_INT(entropic_number_of_rarest_features, 100, \"Experimental. If \"\n+ \"entropic is enabled, we keep track of the frequencies only for the \"\n+ \"Top-X least abundant features (union features that are considered as \"\n+ \"rare).\")\n++FUZZER_FLAG_INT(entropic_scale_per_exec_time, 0, \"Experimental.\")\n+\n+ FUZZER_FLAG_INT(analyze_dict, 0, \"Experimental\")\n+ FUZZER_DEPRECATED_FLAG(use_clang_coverage)\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerFork.cpp b/compiler-rt/lib/fuzzer/FuzzerFork.cpp\nindex d9e6b79443e..97e91cbe869 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerFork.cpp\n@@ -154,10 +256,18 @@ index 31096ce804b..e75807209f5 100644\nsize_t NumberOfLeakDetectionAttempts = 0;\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerLoop.cpp b/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n-index 02db6d27b0a..28d5f32c0d6 100644\n+index 02db6d27b0a..65a02c35758 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n+++ b/compiler-rt/lib/fuzzer/FuzzerLoop.cpp\n-@@ -478,7 +478,7 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\n+@@ -469,6 +469,7 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\n+ return false;\n+\n+ ExecuteCallback(Data, Size);\n++ auto TimeOfUnit = duration_cast<microseconds>(UnitStopTime - UnitStartTime);\n+\n+ UniqFeatureSetTmp.clear();\n+ size_t FoundUniqFeaturesOfII = 0;\n+@@ -478,7 +479,7 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\nUniqFeatureSetTmp.push_back(Feature);\nif (Options.Entropic)\nCorpus.UpdateFeatureFrequency(II, Feature);\n@@ -166,24 +276,22 @@ index 02db6d27b0a..28d5f32c0d6 100644\nif (std::binary_search(II->UniqFeatureSet.begin(),\nII->UniqFeatureSet.end(), Feature))\nFoundUniqFeaturesOfII++;\n-@@ -487,11 +487,12 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\n+@@ -487,11 +488,12 @@ bool Fuzzer::RunOne(const uint8_t *Data, size_t Size, bool MayDeleteFile,\n*FoundUniqFeatures = FoundUniqFeaturesOfII;\nPrintPulseAndReportSlowInput(Data, Size);\nsize_t NumNewFeatures = Corpus.NumFeatureUpdates() - NumUpdatesBefore;\n- if (NumNewFeatures) {\n+ if (NumNewFeatures || (Options.KeepSeed && IsExecutingSeedCorpora)) {\nTPC.UpdateObservedPCs();\n-- auto NewII = Corpus.AddToCorpus({Data, Data + Size}, NumNewFeatures,\n-- MayDeleteFile, TPC.ObservedFocusFunction(),\n+ auto NewII = Corpus.AddToCorpus({Data, Data + Size}, NumNewFeatures,\n+ MayDeleteFile, TPC.ObservedFocusFunction(),\n- UniqFeatureSetTmp, DFT, II);\n-+ auto NewII =\n-+ Corpus.AddToCorpus({Data, Data + Size}, NumNewFeatures, MayDeleteFile,\n-+ TPC.ObservedFocusFunction(),\n-+ IsExecutingSeedCorpora, UniqFeatureSetTmp, DFT, II);\n++ IsExecutingSeedCorpora,\n++ TimeOfUnit, UniqFeatureSetTmp, DFT, II);\nWriteFeatureSetToFile(Options.FeaturesDir, Sha1ToString(NewII->Sha1),\nNewII->UniqFeatureSet);\nreturn true;\n-@@ -664,8 +665,14 @@ void Fuzzer::MutateAndTestOne() {\n+@@ -664,8 +666,14 @@ void Fuzzer::MutateAndTestOne() {\nMD.StartMutationSequence();\nauto &II = Corpus.ChooseUnitToMutate(MD.GetRand());\n@@ -200,7 +308,7 @@ index 02db6d27b0a..28d5f32c0d6 100644\nconst auto &U = II.U;\nmemcpy(BaseSha1, II.Sha1, sizeof(BaseSha1));\nassert(CurrentUnitData);\n-@@ -764,6 +771,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+@@ -764,6 +772,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\nassert(CorporaFiles.front().Size <= CorporaFiles.back().Size);\n}\n@@ -209,7 +317,7 @@ index 02db6d27b0a..28d5f32c0d6 100644\n// Load and execute inputs one by one.\nfor (auto &SF : CorporaFiles) {\nauto U = FileToVector(SF.File, MaxInputLen, /*ExitOnError=*/false);\n-@@ -773,6 +782,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+@@ -773,6 +783,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\nTryDetectingAMemoryLeak(U.data(), U.size(),\n/*DuringInitialCorpusExecution*/ true);\n}\n@@ -218,7 +326,7 @@ index 02db6d27b0a..28d5f32c0d6 100644\n}\nPrintStats(\"INITED\");\n-@@ -785,6 +796,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\n+@@ -785,6 +797,8 @@ void Fuzzer::ReadAndExecuteSeedCorpora(Vector<SizedFile> &CorporaFiles) {\nCorpus.NumInputsThatTouchFocusFunction());\n}\n@@ -228,7 +336,7 @@ index 02db6d27b0a..28d5f32c0d6 100644\nPrintf(\"ERROR: no interesting inputs were found. \"\n\"Is the code instrumented for coverage? Exiting.\\n\");\ndiff --git a/compiler-rt/lib/fuzzer/FuzzerOptions.h b/compiler-rt/lib/fuzzer/FuzzerOptions.h\n-index 9d975bd61fe..bb2c14be47b 100644\n+index b75e7c7af70..9b98383b4aa 100644\n--- a/compiler-rt/lib/fuzzer/FuzzerOptions.h\n+++ b/compiler-rt/lib/fuzzer/FuzzerOptions.h\n@@ -18,6 +18,7 @@ struct FuzzingOptions {\n@@ -248,6 +356,14 @@ index 9d975bd61fe..bb2c14be47b 100644\nint MutateDepth = 5;\nbool ReduceDepth = false;\nbool UseCounters = false;\n+@@ -47,6 +50,7 @@ struct FuzzingOptions {\n+ bool Entropic = false;\n+ size_t EntropicFeatureFrequencyThreshold = 0xFF;\n+ size_t EntropicNumberOfRarestFeatures = 100;\n++ bool EntropicScalePerExecTime = false;\n+ std::string OutputCorpus;\n+ std::string ArtifactPrefix = \"./\";\n+ std::string ExactArtifactPath;\ndiff --git a/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp b/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp\nindex 0e9435ab8fc..dfc642ab6d0 100644\n--- a/compiler-rt/lib/fuzzer/tests/FuzzerUnittest.cpp\n" }, { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# are still testing this feature. You should request an experiment by contacting\n# us as you normally do.\n+- experiment: 2020-08-21\n+ fuzzers:\n+ - entropic_keepseed\n+\n- experiment: 2020-08-20\nfuzzers:\n- aflplusplus\n" } ]

[ { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "#\n# You can run \"make presubmit\" to do basic validation on this file.\n# Please add new experiment requests towards the top of this file.\n-# NOTE: Users of the FuzzBench service should not be editing this file yet, we\n-# are still testing this feature. You should request an experiment by contacting\n-# us as you normally do.\n- experiment: 2020-08-21\nfuzzers:\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/developing-fuzzbench/adding_a_new_benchmark.md", "new_path": "docs/developing-fuzzbench/adding_a_new_benchmark.md", "diff": "@@ -27,8 +27,8 @@ repo, you must create these yourself.\nYou can use most existing OSS-Fuzz projects a benchmark. First decide which project\nand fuzz target you want to use as a benchmark. Next, find out the commit at which\nyou want to use the project for the benchmark. Finally, find out the date and time\n-(UTC) of that commit in ISO format. This can be done in the (benchmark) project\n-repo as follows:\n+(UTC) of that commit in ISO format. You can get the date and time from the project\n+(benchmark) repo with this command:\n```shell\ngit --no-pager log -1 $COMMIT_HASH --format=%cd --date=iso-strict\n```\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/fuzzers.yml", "new_path": ".github/workflows/fuzzers.yml", "diff": "@@ -36,7 +36,9 @@ jobs:\n- weizz\n# Temporary variants.\n- libfuzzer_keepseed\n+ - libfuzzer_magicbytes\n- entropic_keepseed\n+ - entropic_magicbytes\n- entropic_exectime\nbenchmark_type:\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_magicbytes/builder.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+ARG parent_image\n+FROM $parent_image\n+\n+COPY patch.diff /\n+\n+RUN git clone https://github.com/llvm/llvm-project.git /llvm-project && \\\n+ cd /llvm-project && \\\n+ git checkout bb54bcf84970c04c9748004f3a4cf59b0c1832a7 && \\\n+ patch -p1 < /patch.diff && \\\n+ cd /llvm-project/compiler-rt/lib/fuzzer && \\\n+ (for f in *.cpp; do \\\n+ clang++ -stdlib=libc++ -fPIC -O2 -std=c++11 $f -c & \\\n+ done && wait) && \\\n+ ar r /libEntropic.a *.o\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_magicbytes/fuzzer.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Integration code for libFuzzer fuzzer.\"\"\"\n+\n+from fuzzers.entropic import fuzzer as entropic_fuzzer\n+from fuzzers.libfuzzer import fuzzer as libfuzzer_fuzzer\n+\n+\n+def build():\n+ \"\"\"Build benchmark.\"\"\"\n+ entropic_fuzzer.build()\n+\n+\n+def fuzz(input_corpus, output_corpus, target_binary):\n+ \"\"\"Run fuzzer.\"\"\"\n+ libfuzzer_fuzzer.run_fuzzer(input_corpus,\n+ output_corpus,\n+ target_binary,\n+ extra_flags=[\n+ '-entropic=1',\n+ ])\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_magicbytes/patch.diff", "diff": "+commit c4ecc3388200ea614476cece3a7dde6f0c1a7ca8\n+Author: Dokyung Song <[email protected]>\n+Date: Fri Aug 21 00:54:14 2020 +0000\n+\n+ [libFuzzer] Extend ChangeBinaryInteger mutator to support overwriting selected input with predefined integers.\n+\n+ (Experimental - Uploading this to get early feedback before a large-scale experiment.)\n+\n+ This patch extends the ChangeBinaryInteger mutator to support overwriting the\n+ selected input with predefined integers. The rationale for this heuristic is\n+ that certain byte (word, qword, or qword) overwrite at a specific location (with\n+ \"magic\" integers) in a large input may make an invalid input valid, potentially\n+ triggering new neighbor code paths.\n+\n+ Currently, triggering such an overwrite is costly in libFuzzer.\n+ ChangeBinaryInteger mutator may do the same, but only with a low probability,\n+ because the chosen byte (word, dword, or qword) must already be an integer\n+ ranging from -10 to 10.\n+\n+ CopyPart/CrossOver mutator may also effectively do the same, but only if these\n+ predefined integers are found in any of the corpus inputs; even if the corpus\n+ inputs do contain the predefined integers, the chances are much narrower because\n+ a specific location and a specific width have to be selected.\n+\n+ InsertRepeatedBytes combined with EraseBytes mutators (or other combinations of\n+ existing mutators) may eventually trigger the desired change, but still the\n+ probability is low, as the probabilities of different mutators multiply.\n+\n+ This patch allows to find the desired input in a single mutation (as tested by\n+ the accompanying test - overwrite-bytes.test), effectively increasing the\n+ probability of finding the desired input given a corpus input.\n+\n+ Differential Revision: https://reviews.llvm.org/D86358\n+\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerMutate.cpp b/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n+index 29541eac5dc..75527c95aca 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n++++ b/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n+@@ -379,6 +379,67 @@ size_t MutationDispatcher::Mutate_ChangeASCIIInteger(uint8_t *Data, size_t Size,\n+ return Size;\n+ }\n+\n++#define INTERESTING_8 \\\n++ -128, /* Overflow signed 8-bit when decremented */ \\\n++ -1, /* */ \\\n++ 0, /* */ \\\n++ 1, /* */ \\\n++ 16, /* One-off with common buffer size */ \\\n++ 32, /* One-off with common buffer size */ \\\n++ 64, /* One-off with common buffer size */ \\\n++ 100, /* One-off with common buffer size */ \\\n++ 127 /* Overflow signed 8-bit when incremented */\n++\n++#define INTERESTING_16 \\\n++ -32768, /* Overflow signed 16-bit when decremented */ \\\n++ -129, /* Overflow signed 8-bit */ \\\n++ 128, /* Overflow signed 8-bit */ \\\n++ 255, /* Overflow unsig 8-bit when incremented */ \\\n++ 256, /* Overflow unsig 8-bit */ \\\n++ 512, /* One-off with common buffer size */ \\\n++ 1000, /* One-off with common buffer size */ \\\n++ 1024, /* One-off with common buffer size */ \\\n++ 4096, /* One-off with common buffer size */ \\\n++ 32767 /* Overflow signed 16-bit when incremented */\n++\n++#define INTERESTING_32 \\\n++ -2147483648LL, /* Overflow signed 32-bit when decremented */ \\\n++ -100663046, /* Large negative number (endian-agnostic) */ \\\n++ -32769, /* Overflow signed 16-bit */ \\\n++ 32768, /* Overflow signed 16-bit */ \\\n++ 65535, /* Overflow unsig 16-bit when incremented */ \\\n++ 65536, /* Overflow unsig 16 bit */ \\\n++ 100663045, /* Large positive number (endian-agnostic) */ \\\n++ 2147483647 /* Overflow signed 32-bit when incremented */\n++\n++template <class T> class MagicInt8 {\n++public:\n++ static constexpr T Values[] = {INTERESTING_8};\n++};\n++\n++template <class T> class MagicInt16 {\n++public:\n++ static constexpr T Values[] = {INTERESTING_8, INTERESTING_16};\n++};\n++\n++template <class T> class MagicInt32 {\n++public:\n++ static constexpr T Values[] = {INTERESTING_8, INTERESTING_16, INTERESTING_32};\n++};\n++\n++// Definitions\n++template <class T> constexpr T MagicInt8<T>::Values[];\n++template <class T> constexpr T MagicInt16<T>::Values[];\n++template <class T> constexpr T MagicInt32<T>::Values[];\n++\n++template <class T>\n++using MagicInt = typename std::conditional<\n++ sizeof(T) == 1, MagicInt8<int8_t>,\n++ typename std::conditional<\n++ sizeof(T) == 2, MagicInt16<int16_t>,\n++ typename std::conditional<sizeof(T) == 4, MagicInt32<int32_t>,\n++ MagicInt32<int64_t>>::type>::type>::type;\n++\n+ template<class T>\n+ size_t ChangeBinaryInteger(uint8_t *Data, size_t Size, Random &Rand) {\n+ if (Size < sizeof(T)) return 0;\n+@@ -389,6 +450,13 @@ size_t ChangeBinaryInteger(uint8_t *Data, size_t Size, Random &Rand) {\n+ Val = Size;\n+ if (Rand.RandBool())\n+ Val = Bswap(Val);\n++ } else if (Rand.RandBool()) {\n++ auto SignedVal =\n++ MagicInt<T>::Values[Rand(sizeof(MagicInt<T>::Values) / sizeof(T))];\n++ memcpy(&Val, &SignedVal, sizeof(SignedVal));\n++ if (Rand.RandBool()) {\n++ Val = Bswap(Val);\n++ }\n+ } else {\n+ memcpy(&Val, Data + Off, sizeof(Val));\n+ T Add = Rand(21);\n+diff --git a/compiler-rt/test/fuzzer/OverwriteBytesMain.cpp b/compiler-rt/test/fuzzer/OverwriteBytesMain.cpp\n+new file mode 100644\n+index 00000000000..0c5f392cde2\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/OverwriteBytesMain.cpp\n+@@ -0,0 +1,9 @@\n++#include <cstdint>\n++#include <cstdio>\n++\n++#include \"OverwriteBytesTest.h\"\n++\n++int main(int argc, char **argv) {\n++ fwrite(SeedInput, sizeof(SeedInput[0]), sizeof(SeedInput), stdout);\n++ return 0;\n++}\n+diff --git a/compiler-rt/test/fuzzer/OverwriteBytesTest.cpp b/compiler-rt/test/fuzzer/OverwriteBytesTest.cpp\n+new file mode 100644\n+index 00000000000..377a8797b70\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/OverwriteBytesTest.cpp\n+@@ -0,0 +1,33 @@\n++// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.\n++// See https://llvm.org/LICENSE.txt for license information.\n++// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception\n++\n++#include <cassert>\n++#include <cstdint>\n++#include <cstdio>\n++#include <cstdlib>\n++#include <cstring>\n++\n++#include <algorithm>\n++#include <vector>\n++\n++#include \"OverwriteBytesTest.h\"\n++\n++#define MAGIC_BYTE_VALUE 0x1\n++#define MAGIC_BYTE_OFFSET 0xf\n++\n++static volatile int *Nil = nullptr;\n++\n++extern \"C\" int LLVMFuzzerTestOneInput(const uint8_t *Data, size_t Size) {\n++ if (Size != sizeof(SeedInput)) {\n++ return 0;\n++ }\n++\n++ *(uint64_t *)(SeedInput + MAGIC_BYTE_OFFSET) = MAGIC_BYTE_VALUE;\n++\n++ if (memmem(Data, Size, SeedInput, Size) == Data) {\n++ *Nil = 42; // crash.\n++ }\n++\n++ return 0;\n++}\n+diff --git a/compiler-rt/test/fuzzer/OverwriteBytesTest.h b/compiler-rt/test/fuzzer/OverwriteBytesTest.h\n+new file mode 100644\n+index 00000000000..def5705a435\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/OverwriteBytesTest.h\n+@@ -0,0 +1,38 @@\n++uint8_t SeedInput[] = {\n++ 0xba,\n++ 0xe3,\n++ 0x92,\n++ 0x7c,\n++ 0x80,\n++ 0x86,\n++ 0x73,\n++ 0x0f,\n++ 0xf2,\n++ 0x83,\n++ 0x23,\n++ 0x0f,\n++ 0xf5,\n++ 0x17,\n++ 0x4c,\n++ 0x08,\n++ 0xf2,\n++ 0x83,\n++ 0x23,\n++ 0x0f,\n++ 0xd8,\n++ 0x71,\n++ 0x58,\n++ 0x1c,\n++ 0xb9,\n++ 0x8d,\n++ 0xf1,\n++ 0x0e,\n++ 0x80,\n++ 0x86,\n++ 0x73,\n++ 0x0f,\n++ 0xf0,\n++ 0x83,\n++ 0x23,\n++ 0x0f,\n++};\n+diff --git a/compiler-rt/test/fuzzer/overwrite-bytes.test b/compiler-rt/test/fuzzer/overwrite-bytes.test\n+new file mode 100644\n+index 00000000000..1383ff54492\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/overwrite-bytes.test\n+@@ -0,0 +1,9 @@\n++REQUIRES: linux, x86_64\n++RUN: %cpp_compiler %S/OverwriteBytesTest.cpp -o %t-OverwriteBytesTest\n++RUN: %cpp_compiler -fno-sanitize=fuzzer %S/OverwriteBytesMain.cpp -o %t-OverwriteBytesPrintSeed\n++RUN: %t-OverwriteBytesPrintSeed > %t-OverwriteBytesTest.seed\n++\n++RUN: not %run %t-OverwriteBytesTest -seed=1 -use_memmem=0 -mutate_depth=1 -reduce_inputs=0 -runs=10000000 -seed_inputs=%t-OverwriteBytesTest.seed 2>&1 | FileCheck %s\n++\n++CHECK: ABORTING\n++CHECK-NEXT: MS: 1 ChangeBinInt-;\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/entropic_magicbytes/runner.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+FROM gcr.io/fuzzbench/base-runner\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/libfuzzer_magicbytes/builder.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+ARG parent_image\n+FROM $parent_image\n+\n+COPY patch.diff /\n+\n+RUN git clone https://github.com/llvm/llvm-project.git /llvm-project && \\\n+ cd /llvm-project/ && \\\n+ git checkout bb54bcf84970c04c9748004f3a4cf59b0c1832a7 && \\\n+ patch -p1 < /patch.diff && \\\n+ cd compiler-rt/lib/fuzzer && \\\n+ (for f in *.cpp; do \\\n+ clang++ -stdlib=libc++ -fPIC -O2 -std=c++11 $f -c & \\\n+ done && wait) && \\\n+ ar r /usr/lib/libFuzzer.a *.o\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/libfuzzer_magicbytes/fuzzer.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Integration code for libFuzzer fuzzer.\"\"\"\n+\n+from fuzzers.libfuzzer import fuzzer as libfuzzer_fuzzer\n+\n+\n+def build():\n+ \"\"\"Build benchmark.\"\"\"\n+ libfuzzer_fuzzer.build()\n+\n+\n+def fuzz(input_corpus, output_corpus, target_binary):\n+ \"\"\"Run fuzzer.\"\"\"\n+ libfuzzer_fuzzer.run_fuzzer(input_corpus, output_corpus, target_binary)\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/libfuzzer_magicbytes/patch.diff", "diff": "+commit c4ecc3388200ea614476cece3a7dde6f0c1a7ca8\n+Author: Dokyung Song <[email protected]>\n+Date: Fri Aug 21 00:54:14 2020 +0000\n+\n+ [libFuzzer] Extend ChangeBinaryInteger mutator to support overwriting selected input with predefined integers.\n+\n+ (Experimental - Uploading this to get early feedback before a large-scale experiment.)\n+\n+ This patch extends the ChangeBinaryInteger mutator to support overwriting the\n+ selected input with predefined integers. The rationale for this heuristic is\n+ that certain byte (word, qword, or qword) overwrite at a specific location (with\n+ \"magic\" integers) in a large input may make an invalid input valid, potentially\n+ triggering new neighbor code paths.\n+\n+ Currently, triggering such an overwrite is costly in libFuzzer.\n+ ChangeBinaryInteger mutator may do the same, but only with a low probability,\n+ because the chosen byte (word, dword, or qword) must already be an integer\n+ ranging from -10 to 10.\n+\n+ CopyPart/CrossOver mutator may also effectively do the same, but only if these\n+ predefined integers are found in any of the corpus inputs; even if the corpus\n+ inputs do contain the predefined integers, the chances are much narrower because\n+ a specific location and a specific width have to be selected.\n+\n+ InsertRepeatedBytes combined with EraseBytes mutators (or other combinations of\n+ existing mutators) may eventually trigger the desired change, but still the\n+ probability is low, as the probabilities of different mutators multiply.\n+\n+ This patch allows to find the desired input in a single mutation (as tested by\n+ the accompanying test - overwrite-bytes.test), effectively increasing the\n+ probability of finding the desired input given a corpus input.\n+\n+ Differential Revision: https://reviews.llvm.org/D86358\n+\n+diff --git a/compiler-rt/lib/fuzzer/FuzzerMutate.cpp b/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n+index 29541eac5dc..75527c95aca 100644\n+--- a/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n++++ b/compiler-rt/lib/fuzzer/FuzzerMutate.cpp\n+@@ -379,6 +379,67 @@ size_t MutationDispatcher::Mutate_ChangeASCIIInteger(uint8_t *Data, size_t Size,\n+ return Size;\n+ }\n+\n++#define INTERESTING_8 \\\n++ -128, /* Overflow signed 8-bit when decremented */ \\\n++ -1, /* */ \\\n++ 0, /* */ \\\n++ 1, /* */ \\\n++ 16, /* One-off with common buffer size */ \\\n++ 32, /* One-off with common buffer size */ \\\n++ 64, /* One-off with common buffer size */ \\\n++ 100, /* One-off with common buffer size */ \\\n++ 127 /* Overflow signed 8-bit when incremented */\n++\n++#define INTERESTING_16 \\\n++ -32768, /* Overflow signed 16-bit when decremented */ \\\n++ -129, /* Overflow signed 8-bit */ \\\n++ 128, /* Overflow signed 8-bit */ \\\n++ 255, /* Overflow unsig 8-bit when incremented */ \\\n++ 256, /* Overflow unsig 8-bit */ \\\n++ 512, /* One-off with common buffer size */ \\\n++ 1000, /* One-off with common buffer size */ \\\n++ 1024, /* One-off with common buffer size */ \\\n++ 4096, /* One-off with common buffer size */ \\\n++ 32767 /* Overflow signed 16-bit when incremented */\n++\n++#define INTERESTING_32 \\\n++ -2147483648LL, /* Overflow signed 32-bit when decremented */ \\\n++ -100663046, /* Large negative number (endian-agnostic) */ \\\n++ -32769, /* Overflow signed 16-bit */ \\\n++ 32768, /* Overflow signed 16-bit */ \\\n++ 65535, /* Overflow unsig 16-bit when incremented */ \\\n++ 65536, /* Overflow unsig 16 bit */ \\\n++ 100663045, /* Large positive number (endian-agnostic) */ \\\n++ 2147483647 /* Overflow signed 32-bit when incremented */\n++\n++template <class T> class MagicInt8 {\n++public:\n++ static constexpr T Values[] = {INTERESTING_8};\n++};\n++\n++template <class T> class MagicInt16 {\n++public:\n++ static constexpr T Values[] = {INTERESTING_8, INTERESTING_16};\n++};\n++\n++template <class T> class MagicInt32 {\n++public:\n++ static constexpr T Values[] = {INTERESTING_8, INTERESTING_16, INTERESTING_32};\n++};\n++\n++// Definitions\n++template <class T> constexpr T MagicInt8<T>::Values[];\n++template <class T> constexpr T MagicInt16<T>::Values[];\n++template <class T> constexpr T MagicInt32<T>::Values[];\n++\n++template <class T>\n++using MagicInt = typename std::conditional<\n++ sizeof(T) == 1, MagicInt8<int8_t>,\n++ typename std::conditional<\n++ sizeof(T) == 2, MagicInt16<int16_t>,\n++ typename std::conditional<sizeof(T) == 4, MagicInt32<int32_t>,\n++ MagicInt32<int64_t>>::type>::type>::type;\n++\n+ template<class T>\n+ size_t ChangeBinaryInteger(uint8_t *Data, size_t Size, Random &Rand) {\n+ if (Size < sizeof(T)) return 0;\n+@@ -389,6 +450,13 @@ size_t ChangeBinaryInteger(uint8_t *Data, size_t Size, Random &Rand) {\n+ Val = Size;\n+ if (Rand.RandBool())\n+ Val = Bswap(Val);\n++ } else if (Rand.RandBool()) {\n++ auto SignedVal =\n++ MagicInt<T>::Values[Rand(sizeof(MagicInt<T>::Values) / sizeof(T))];\n++ memcpy(&Val, &SignedVal, sizeof(SignedVal));\n++ if (Rand.RandBool()) {\n++ Val = Bswap(Val);\n++ }\n+ } else {\n+ memcpy(&Val, Data + Off, sizeof(Val));\n+ T Add = Rand(21);\n+diff --git a/compiler-rt/test/fuzzer/OverwriteBytesMain.cpp b/compiler-rt/test/fuzzer/OverwriteBytesMain.cpp\n+new file mode 100644\n+index 00000000000..0c5f392cde2\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/OverwriteBytesMain.cpp\n+@@ -0,0 +1,9 @@\n++#include <cstdint>\n++#include <cstdio>\n++\n++#include \"OverwriteBytesTest.h\"\n++\n++int main(int argc, char **argv) {\n++ fwrite(SeedInput, sizeof(SeedInput[0]), sizeof(SeedInput), stdout);\n++ return 0;\n++}\n+diff --git a/compiler-rt/test/fuzzer/OverwriteBytesTest.cpp b/compiler-rt/test/fuzzer/OverwriteBytesTest.cpp\n+new file mode 100644\n+index 00000000000..377a8797b70\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/OverwriteBytesTest.cpp\n+@@ -0,0 +1,33 @@\n++// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.\n++// See https://llvm.org/LICENSE.txt for license information.\n++// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception\n++\n++#include <cassert>\n++#include <cstdint>\n++#include <cstdio>\n++#include <cstdlib>\n++#include <cstring>\n++\n++#include <algorithm>\n++#include <vector>\n++\n++#include \"OverwriteBytesTest.h\"\n++\n++#define MAGIC_BYTE_VALUE 0x1\n++#define MAGIC_BYTE_OFFSET 0xf\n++\n++static volatile int *Nil = nullptr;\n++\n++extern \"C\" int LLVMFuzzerTestOneInput(const uint8_t *Data, size_t Size) {\n++ if (Size != sizeof(SeedInput)) {\n++ return 0;\n++ }\n++\n++ *(uint64_t *)(SeedInput + MAGIC_BYTE_OFFSET) = MAGIC_BYTE_VALUE;\n++\n++ if (memmem(Data, Size, SeedInput, Size) == Data) {\n++ *Nil = 42; // crash.\n++ }\n++\n++ return 0;\n++}\n+diff --git a/compiler-rt/test/fuzzer/OverwriteBytesTest.h b/compiler-rt/test/fuzzer/OverwriteBytesTest.h\n+new file mode 100644\n+index 00000000000..def5705a435\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/OverwriteBytesTest.h\n+@@ -0,0 +1,38 @@\n++uint8_t SeedInput[] = {\n++ 0xba,\n++ 0xe3,\n++ 0x92,\n++ 0x7c,\n++ 0x80,\n++ 0x86,\n++ 0x73,\n++ 0x0f,\n++ 0xf2,\n++ 0x83,\n++ 0x23,\n++ 0x0f,\n++ 0xf5,\n++ 0x17,\n++ 0x4c,\n++ 0x08,\n++ 0xf2,\n++ 0x83,\n++ 0x23,\n++ 0x0f,\n++ 0xd8,\n++ 0x71,\n++ 0x58,\n++ 0x1c,\n++ 0xb9,\n++ 0x8d,\n++ 0xf1,\n++ 0x0e,\n++ 0x80,\n++ 0x86,\n++ 0x73,\n++ 0x0f,\n++ 0xf0,\n++ 0x83,\n++ 0x23,\n++ 0x0f,\n++};\n+diff --git a/compiler-rt/test/fuzzer/overwrite-bytes.test b/compiler-rt/test/fuzzer/overwrite-bytes.test\n+new file mode 100644\n+index 00000000000..1383ff54492\n+--- /dev/null\n++++ b/compiler-rt/test/fuzzer/overwrite-bytes.test\n+@@ -0,0 +1,9 @@\n++REQUIRES: linux, x86_64\n++RUN: %cpp_compiler %S/OverwriteBytesTest.cpp -o %t-OverwriteBytesTest\n++RUN: %cpp_compiler -fno-sanitize=fuzzer %S/OverwriteBytesMain.cpp -o %t-OverwriteBytesPrintSeed\n++RUN: %t-OverwriteBytesPrintSeed > %t-OverwriteBytesTest.seed\n++\n++RUN: not %run %t-OverwriteBytesTest -seed=1 -use_memmem=0 -mutate_depth=1 -reduce_inputs=0 -runs=10000000 -seed_inputs=%t-OverwriteBytesTest.seed 2>&1 | FileCheck %s\n++\n++CHECK: ABORTING\n++CHECK-NEXT: MS: 1 ChangeBinInt-;\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/libfuzzer_magicbytes/runner.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+FROM gcr.io/fuzzbench/base-runner\n" }, { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "# You can run \"make presubmit\" to do basic validation on this file.\n# Please add new experiment requests towards the top of this file.\n+- experiment: 2020-08-25\n+ fuzzers:\n+ - libfuzzer_magicbytes\n+ - entropic_magicbytes\n+\n- experiment: 2020-08-21\nfuzzers:\n- entropic_keepseed\n" } ]

[ { "change_type": "MODIFY", "old_path": "common/benchmark_utils.py", "new_path": "common/benchmark_utils.py", "diff": "import os\nimport re\n+import yaml\n+\nfrom common import environment\nfrom common import logs\nfrom common import benchmark_config\nfrom common import utils\n-VALID_BENCHMARK_REGEX = re.compile(r'^[A-Za-z0-9\\._\\-]+$')\n+# Must be valid in a docker tag.\n+VALID_BENCHMARK_REGEX = re.compile(r'^[a-z0-9\\._\\-]+$')\nBENCHMARKS_DIR = os.path.join(utils.ROOT_DIR, 'benchmarks')\n-def get_project(benchmark):\n- \"\"\"Returns the OSS-Fuzz project of |benchmark| if it is based on an\n- OSS-Fuzz project, otherwise raises ValueError.\"\"\"\n- return benchmark_config.get_config(benchmark)['project']\n-\n-\ndef get_fuzz_target(benchmark):\n\"\"\"Returns the fuzz target of |benchmark|\"\"\"\nreturn benchmark_config.get_config(benchmark)['fuzz_target']\n@@ -53,17 +50,37 @@ def get_builder_image_url(benchmark, fuzzer, docker_registry):\ndocker_registry=docker_registry, fuzzer=fuzzer, benchmark=benchmark)\n-def validate(benchmark):\n- \"\"\"Return True if |benchmark| is a valid fuzzbench fuzzer.\"\"\"\n+def validate_name(benchmark):\n+ \"\"\"Returns True if |benchmark| is a valid fuzzbench benchmark name.\"\"\"\nif VALID_BENCHMARK_REGEX.match(benchmark) is None:\nlogs.error('%s does not conform to %s pattern.', benchmark,\nVALID_BENCHMARK_REGEX.pattern)\nreturn False\n- if benchmark in get_all_benchmarks():\nreturn True\n+\n+\n+def validate(benchmark):\n+ \"\"\"Returns True if |benchmark| is a valid fuzzbench benchmark.\"\"\"\n+ if not validate_name(benchmark):\n+ return False\n+\n+ if benchmark not in get_all_benchmarks():\nlogs.error('%s must have a benchmark.yaml.', benchmark)\nreturn False\n+ try:\n+ get_fuzz_target(benchmark)\n+ except yaml.parser.ParserError:\n+ logs.error('%s must have a valid benchmark.yaml file. Failed to parse.',\n+ benchmark)\n+ return False\n+ except KeyError:\n+ logs.error('%s\\'s benchmark.yaml does not define \"fuzz_target\".',\n+ benchmark)\n+ return False\n+\n+ return True\n+\ndef get_all_benchmarks():\n\"\"\"Returns the list of all benchmarks.\"\"\"\n" }, { "change_type": "MODIFY", "old_path": "common/fuzzer_utils.py", "new_path": "common/fuzzer_utils.py", "diff": "@@ -23,7 +23,10 @@ from common import utils\nDEFAULT_FUZZ_TARGET_NAME = 'fuzz-target'\nFUZZ_TARGET_SEARCH_STRING = b'LLVMFuzzerTestOneInput'\n-VALID_FUZZER_REGEX = re.compile(r'^[A-Za-z0-9_]+$')\n+\n+# Must be a valid python module and docker tag.\n+VALID_FUZZER_REGEX = re.compile(r'^[a-z0-9_]+$')\n+\nFUZZERS_DIR = os.path.join(utils.ROOT_DIR, 'fuzzers')\nCOVERAGE_TOOLS = {'coverage', 'coverage_source_based'}\n@@ -95,16 +98,24 @@ def get_fuzz_target_binary(search_directory: str,\nreturn None\n-def validate(fuzzer):\n- \"\"\"Return True if |fuzzer| is a valid fuzzbench fuzzer.\"\"\"\n+def validate_name(fuzzer):\n+ \"\"\"Return True if |fuzzer| is a valid fuzzbench fuzzer name.\"\"\"\n# Although importing probably allows a subset of what the regex allows, use\n# the regex anyway to be safe. The regex is enforcing that the fuzzer is a\n# valid path for GCS or a linux system.\nif VALID_FUZZER_REGEX.match(fuzzer) is None:\n- logs.error('%s does not conform to %s pattern.', fuzzer,\n+ logs.error('Fuzzer: %s does not conform to pattern: %s.', fuzzer,\nVALID_FUZZER_REGEX.pattern)\nreturn False\n+ return True\n+\n+\n+def validate(fuzzer):\n+ \"\"\"Return True if |fuzzer| is a valid fuzzbench fuzzer.\"\"\"\n+ if not validate_name(fuzzer):\n+ return False\n+\n# Try importing the fuzzer module.\nmodule_name = 'fuzzers.{}.fuzzer'.format(fuzzer)\ntry:\n" }, { "change_type": "MODIFY", "old_path": "common/test_benchmark_utils.py", "new_path": "common/test_benchmark_utils.py", "diff": "@@ -40,3 +40,19 @@ def test_get_runner_image_url(benchmark, expected_url, oss_fuzz_benchmark):\nassert benchmark_utils.get_runner_image_url('experiment', benchmark,\n'fuzzer',\nDOCKER_REGISTRY) == expected_url\n+\n+\[email protected](('benchmark_name',), [\n+ ('libPNG',),\n+ ('libpng!',),\n+])\n+def test_validate_name_invalid(benchmark_name):\n+ \"\"\"Tests that validate_name returns False for am invalid benchmark name.\"\"\"\n+ assert not benchmark_utils.validate_name(benchmark_name)\n+\n+\[email protected](('benchmark_name',), [('libpng',), ('libpng_1',),\n+ ('libpng-1',), ('libpng.1',)])\n+def test_validate_name_valid(benchmark_name):\n+ \"\"\"Tests that validate_name returns True for a valid benchmark name.\"\"\"\n+ assert benchmark_utils.validate_name(benchmark_name)\n" }, { "change_type": "MODIFY", "old_path": "common/test_fuzzer_utils.py", "new_path": "common/test_fuzzer_utils.py", "diff": "# See the License for the specific language governing permissions and\n# limitations under the License.\n\"\"\"Tests for fuzzer_utils.py.\"\"\"\n+import pytest\n+\nfrom common import fuzzer_utils\n# pylint: disable=invalid-name,unused-argument\n@@ -53,3 +55,16 @@ def test_found_fuzzer_containing_string_without_fuzzer_name_arg(fs, environ):\nfs.create_file('/out/custom-target', contents='\\n\\nLLVMFuzzerTestOneInput')\nassert fuzzer_utils.get_fuzz_target_binary('/out',\nNone) == ('/out/custom-target')\n+\n+\[email protected](('fuzzer_name',), [('afl++',), ('mopt-afl',),\n+ ('mOptAFL',), ('AFL',)])\n+def test_validate_name_invalid(fuzzer_name):\n+ \"\"\"Tests that validate_name returns False for an invalid fuzzer name.\"\"\"\n+ assert not fuzzer_utils.validate_name(fuzzer_name)\n+\n+\[email protected](('fuzzer_name',), [('afl',), ('mopt_afl',), ('afl2',)])\n+def test_validate_name_valid(fuzzer_name):\n+ \"\"\"Tests that validate_name returns True for a valid fuzzer name.\"\"\"\n+ assert fuzzer_utils.validate_name(fuzzer_name)\n" } ]

[ { "change_type": "MODIFY", "old_path": "docs/getting-started/prerequisites.md", "new_path": "docs/getting-started/prerequisites.md", "diff": "@@ -90,3 +90,11 @@ You can format your changes using the following command:\n```bash\nmake format\n```\n+\n+### Local Support\n+\n+If you want to run FuzzBench [locally]({{ site.baseurl }}/running-a-local-experiment/#Running a local experiment)\n+on your own machine or servers, it needs `rsync` installed:\n+```bash\n+sudo apt-get install rsync\n+```\n" }, { "change_type": "ADD", "old_path": null, "new_path": "docs/running-a-local-experiment/running_a_local_experiment.md", "diff": "+---\n+layout: default\n+title: Running a local experiment\n+nav_order: 5\n+permalink: /running-a-local-experiment\n+---\n+\n+# Running a local experiment\n+\n+This page explains how to run a local [experiment]({{ site.baseurl }}/reference/glossary/#Experiment) on\n+your own.\n+\n+- TOC\n+{:toc}\n+\n+This page will walk you through on how to use `run_experiment.py` to start\n+a local experiment. The `run_experiment.py` script will\n+create and run a dispatcher docker container which runs the experiment,\n+including:\n+1. Building desired fuzzer-benchmark combinations.\n+1. Starting instances to run fuzzing trials with the fuzzer-benchmark\n+ builds and stopping them when they are done.\n+1. Measuring the coverage from these trials.\n+1. Generating reports based on these measurements.\n+\n+The rest of this page will assume all commands are run from the root of\n+FuzzBench checkout.\n+\n+**NOTE**: Currently, there is no resource control in experiment trials (e.g. allocated cpus, memory),\n+but we do plan to add it in the near future.\n+\n+# run_experiment.py\n+\n+## Experiment configuration file\n+\n+You need to create an experiment configuration yaml file.\n+This file contains the configuration parameters for experiments that do not\n+change very often.\n+Below is an example configuration file with explanations of each required\n+parameter.\n+\n+```yaml\n+# The number of trials of a fuzzer-benchmark pair.\n+trials: 5\n+\n+# The amount of time in seconds that each trial is run for.\n+# 1 day = 24 * 60 * 60 = 86400\n+max_total_time: 86400\n+\n+# The location of your docker registry.\n+docker_registry: lab-server:5000\n+\n+# The local experiment folder that will store most of the experiment data.\n+# Please use an absolute path.\n+experiment_filestore: /tmp/experiment-data\n+\n+# The local report folder where HTML reports and summary data will be stored.\n+# Please use an absolute path.\n+report_filestore: /tmp/report-data\n+```\n+\n+## Benchmarks\n+\n+Pick the benchmarks you want to use from the `benchmarks/` directory.\n+\n+For example: `freetype2-2017` and `bloaty_fuzz_target`.\n+\n+## Fuzzers\n+\n+Pick the fuzzers you want to use from the `fuzzers/` directory.\n+For example: `libfuzzer` and `afl`.\n+\n+## Executing run_experiment.py\n+\n+Now that everything is ready, execute `run_experiment.py`:\n+\n+```bash\n+PYTHONPATH=. python3 experiment/run_experiment.py \\\n+--experiment-config experiment-config.yaml \\\n+--benchmarks freetype2-2017 bloaty_fuzz_target \\\n+--experiment-name $EXPERIMENT_NAME \\\n+--fuzzers afl libfuzzer\n+```\n+\n+where `$EXPERIMENT_NAME` is the name you want to give the experiment.\n+\n+## Viewing reports\n+\n+You should eventually be able to see reports from your experiment, that are\n+update at some interval throughout the experiment. However, you may have to wait\n+a while until they first appear since a lot must happen before there is data to\n+generate report. Once they are available, you should be able to view them at:\n+`/tmp/report-data/$EXPERIMENT_NAME/index.html`.\n" } ]

[ { "change_type": "MODIFY", "old_path": "docker/generate_makefile.py", "new_path": "docker/generate_makefile.py", "diff": "@@ -137,6 +137,12 @@ def main():\n])\nprint('build-{fuzzer}-all: {all_targets}'.format(\nfuzzer=fuzzer, all_targets=all_build_targets))\n+ all_test_run_targets = ' '.join([\n+ 'test-run-{0}-{1}'.format(fuzzer, benchmark)\n+ for benchmark in benchmarks\n+ ])\n+ print('test-run-{fuzzer}-all: {all_targets}'.format(\n+ fuzzer=fuzzer, all_targets=all_test_run_targets))\n# Print all targets build target.\nall_build_targets = ' '.join(\n" }, { "change_type": "MODIFY", "old_path": "docs/faq.md", "new_path": "docs/faq.md", "diff": "@@ -75,13 +75,28 @@ to misunderstand configuration details that can have an impact on the results.\nIf you can, please reach out to the authors to confirm your configuration looks\ngood to them.\n-## I'd like to get my fuzzer evaluated on the free FuzzBench service, but I don't want the results to be public yet.\n-\n-Please reach out to us at [email protected].\n-\n-## I'm working on a new tool, but I'm not ready to make my code public yet. Can I use the FuzzBench service?\n+## I'd like to get my fuzzer evaluated, but I don't want the results and/or code to be public yet. Can I use the FuzzBench service?\n+\n+Probably yes. We run private experiments for this purpose.\n+Please reach out to us at [email protected]. If we agree to benchmark your\n+fuzzer, please follow the guide on\n+[adding a new fuzzer]({{ site.baseurl }}/getting-started/adding-a-new-fuzzer/)\n+on how to integrate your fuzzer with FuzzBench.\n+\n+You can ignore the sections on [Requesting an experiment]({{ site.baseurl }}/getting-started/adding-a-new-fuzzer/#requesting-an-experiment) and\n+[Submitting your integration]({{ site.baseurl }}/getting-started/adding-a-new-fuzzer/#submitting-your-integration).\n+Please test your fuzzer works with our benchmarks, we don't have CI to verify\n+this for private experiments.\n+Ideally, you should test all benchmarks using `make -j test-run-$FUZZER-all`.\n+This takes too long on most machines so you should at least test a few of them:\n+```\n+make test-run-$FUZZER-zlib_zlib_uncompress_fuzzer test-run-$FUZZER-libpng-1.2.56\n+```\n-Yes, please reach out to us at [email protected].\n+You should also run `make presubmit` to validate the fuzzer's name and\n+integration code.\n+When your fuzzer is ready, send us a patch file that applies cleanly to\n+FuzzBench with `git apply <patch_file>`.\n## How can you prevent researchers from optimizing their tools only for these benchmarks?\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/fuzzers.yml", "new_path": ".github/workflows/fuzzers.yml", "diff": "@@ -60,10 +60,6 @@ jobs:\n- oss-fuzz\n- standard\n- env:\n- FUZZER: ${{ matrix.fuzzer }}\n- BENCHMARK_TYPE: ${{ matrix.benchmark_type }}\n-\nsteps:\n- uses: actions/checkout@v2\n- run: | # Needed for git diff to work.\n@@ -79,7 +75,7 @@ jobs:\ndf -h\n- name: Setup Python environment\n- uses: actions/[email protected]\n+ uses: actions/setup-python@v2\nwith:\npython-version: 3.7\n@@ -103,4 +99,4 @@ jobs:\n- name: Build Benchmarks\nrun: |\n- PATH=.venv/bin/:$PATH PYTHONPATH=. python3 .github/workflows/build_and_test_run_fuzzer_benchmarks.py $BENCHMARK_TYPE $FUZZER\n+ PATH=.venv/bin/:$PATH PYTHONPATH=. python3 .github/workflows/build_and_test_run_fuzzer_benchmarks.py ${{ matrix.benchmark_type }} ${{ matrix.fuzzer }}\n" }, { "change_type": "MODIFY", "old_path": ".github/workflows/presubmit.yml", "new_path": ".github/workflows/presubmit.yml", "diff": "@@ -12,7 +12,7 @@ jobs:\ngit symbolic-ref refs/remotes/origin/HEAD refs/remotes/origin/master\n- name: Setup Python environment\n- uses: actions/[email protected]\n+ uses: actions/setup-python@v2\nwith:\npython-version: 3.7\n" } ]

[ { "change_type": "MODIFY", "old_path": "analysis/generate_report.py", "new_path": "analysis/generate_report.py", "diff": "@@ -147,7 +147,8 @@ def generate_report(experiment_names,\nqueries.add_nonprivate_experiments_for_merge_with_clobber(\nexperiment_names))\n- report_name = report_name or experiment_names[0]\n+ main_experiment_name = experiment_names[0]\n+ report_name = report_name or main_experiment_name\nfilesystem.create_directory(report_directory)\n@@ -157,6 +158,8 @@ def generate_report(experiment_names,\nelse:\nexperiment_df = queries.get_experiment_data(experiment_names)\n+ description = queries.get_experiment_description(main_experiment_name)\n+\ndata_utils.validate_data(experiment_df)\nif benchmarks is not None:\n@@ -197,7 +200,8 @@ def generate_report(experiment_names,\ntemplate = report_type + '.html'\ndetailed_report = rendering.render_report(experiment_ctx, template,\n- in_progress, coverage_report)\n+ in_progress, coverage_report,\n+ description)\nfilesystem.write(os.path.join(report_directory, 'index.html'),\ndetailed_report)\n" }, { "change_type": "MODIFY", "old_path": "analysis/queries.py", "new_path": "analysis/queries.py", "diff": "@@ -36,6 +36,15 @@ def get_experiment_data(experiment_names):\nreturn pd.read_sql_query(snapshots_query.statement, db_utils.engine)\n+def get_experiment_description(experiment_name):\n+ \"\"\"Get the description of the experiment named by |experiment_name|.\"\"\"\n+ # Do another query for the description so we don't explode the size of the\n+ # results from get_experiment_data.\n+ return db_utils.query(Experiment.description)\\\n+ .select_from(Experiment)\\\n+ .filter(Experiment.name == experiment_name).one()\n+\n+\ndef add_nonprivate_experiments_for_merge_with_clobber(experiment_names):\n\"\"\"Returns a new list containing experiment names preeceeded by a list of\nnonprivate experiments in the order in which they were run, such that\n" }, { "change_type": "MODIFY", "old_path": "analysis/rendering.py", "new_path": "analysis/rendering.py", "diff": "@@ -20,13 +20,17 @@ import jinja2\nfrom common import utils\n-def render_report(experiment_results, template, in_progress, coverage_report):\n+def render_report(experiment_results, template, in_progress, coverage_report,\n+ description):\n\"\"\"Renders report with |template| using data provided by the\n|experiment_results| context.\nArguments:\ntemplate: filename of the report template. E.g., 'default.html'.\nexperiment_results: an ExperimentResults object.\n+ in_progress: Whether the experiment is still in progress.\n+ coverage_report: Whether to report detailed info about coverage.\n+ description: A description of the experiment.\nReturns the rendered template.\n\"\"\"\n@@ -40,4 +44,5 @@ def render_report(experiment_results, template, in_progress, coverage_report):\nreturn template.render(experiment=experiment_results,\nin_progress=in_progress,\n- coverage_report=coverage_report)\n+ coverage_report=coverage_report,\n+ description=description)\n" }, { "change_type": "MODIFY", "old_path": "analysis/report_templates/default.html", "new_path": "analysis/report_templates/default.html", "diff": "The experiment was conducted using this FuzzBench commit:\n<a href=\"https://github.com/google/fuzzbench/commits/{{ experiment.git_hash }}\">{{ experiment.git_hash }}</a>\n{% endif %}\n+\n+ {% if description %}\n+ <br><br>\n+ Experiment Description:<br><br>\n+ {{ description }}\n+ {% endif %}\n</div> <!-- id=\"data\" -->\n</div> <!-- class=\"col\" -->\n" }, { "change_type": "ADD", "old_path": null, "new_path": "database/alembic/versions/26dcc0e12872_add_experiment_description.py", "diff": "+\"\"\"Add experiment description\n+\n+Revision ID: 26dcc0e12872\n+Revises: c83ac04855b4\n+Create Date: 2020-10-13 09:04:25.881798\n+\n+\"\"\"\n+from alembic import op\n+import sqlalchemy as sa\n+\n+\n+# revision identifiers, used by Alembic.\n+revision = '26dcc0e12872'\n+down_revision = 'c83ac04855b4'\n+branch_labels = None\n+depends_on = None\n+\n+\n+def upgrade():\n+ op.add_column('experiment', sa.Column(\n+ 'description', sa.UnicodeText(), nullable=True))\n+\n+\n+def downgrade():\n+ op.drop_column('experiment', 'description')\n" }, { "change_type": "MODIFY", "old_path": "database/models.py", "new_path": "database/models.py", "diff": "\"\"\"SQLAlchemy Database Models.\"\"\"\nimport sqlalchemy\nfrom sqlalchemy.ext import declarative\n-from sqlalchemy import Boolean, Column, Integer, String, ForeignKey, DateTime\n+from sqlalchemy import Boolean\n+from sqlalchemy import Column\n+from sqlalchemy import DateTime\n+from sqlalchemy import ForeignKey\n+from sqlalchemy import Integer\n+from sqlalchemy import String\n+from sqlalchemy import UnicodeText\nBase = declarative.declarative_base() # pylint: disable=invalid-name\n@@ -29,6 +35,7 @@ class Experiment(Base):\ngit_hash = Column(String, nullable=True)\nprivate = Column(Boolean, nullable=False, default=False)\nexperiment_filestore = Column(String, nullable=True)\n+ description = Column(UnicodeText, nullable=True)\nclass Trial(Base):\n" }, { "change_type": "MODIFY", "old_path": "experiment/dispatcher.py", "new_path": "experiment/dispatcher.py", "diff": "@@ -66,7 +66,8 @@ def _initialize_experiment_in_db(experiment_config: dict):\nname=experiment_config['experiment'],\ngit_hash=experiment_config['git_hash'],\nprivate=experiment_config.get('private', True),\n- experiment_filestore=experiment_config['experiment_filestore'])\n+ experiment_filestore=experiment_config['experiment_filestore'],\n+ description=experiment_config['description']),\n])\n" }, { "change_type": "MODIFY", "old_path": "experiment/run_experiment.py", "new_path": "experiment/run_experiment.py", "diff": "@@ -205,6 +205,7 @@ def start_experiment( # pylint: disable=too-many-arguments\nconfig_filename: str,\nbenchmarks: List[str],\nfuzzers: List[str],\n+ description: str = None,\nno_seeds=False,\nno_dictionaries=False,\noss_fuzz_corpus=False):\n@@ -222,6 +223,7 @@ def start_experiment( # pylint: disable=too-many-arguments\nconfig['no_seeds'] = no_seeds\nconfig['no_dictionaries'] = no_dictionaries\nconfig['oss_fuzz_corpus'] = oss_fuzz_corpus\n+ config['description'] = description\nset_up_experiment_config_file(config)\n@@ -476,6 +478,10 @@ def main():\n'--experiment-name',\nhelp='Experiment name.',\nrequired=True)\n+ parser.add_argument('-d',\n+ '--description',\n+ help='Description of the experiment.',\n+ required=False)\nparser.add_argument('-f',\n'--fuzzers',\nhelp='Fuzzers to use.',\n@@ -509,6 +515,7 @@ def main():\nargs.experiment_config,\nargs.benchmarks,\nfuzzers,\n+ description=args.description,\nno_seeds=args.no_seeds,\nno_dictionaries=args.no_dictionaries,\noss_fuzz_corpus=args.oss_fuzz_corpus)\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_data/experiment-config.yaml", "new_path": "experiment/test_data/experiment-config.yaml", "diff": "@@ -31,3 +31,4 @@ git_hash: \"git-hash\"\nno_seeds: false\nno_dictionaries: false\noss_fuzz_corpus: false\n+description: \"Test experiment\"\n" }, { "change_type": "MODIFY", "old_path": "service/automatic_run_experiment.py", "new_path": "service/automatic_run_experiment.py", "diff": "@@ -21,6 +21,7 @@ import collections\nimport os\nimport re\nimport sys\n+from typing import Optional\nfrom common import logs\nfrom common import fuzzer_utils\n@@ -77,13 +78,19 @@ BENCHMARKS = [\ndef _get_experiment_name(experiment_config: dict) -> str:\n- \"\"\"Returns the name of the experiment described by experiment_config as a\n+ \"\"\"Returns the name of the experiment described by |experiment_config| as a\nstring.\"\"\"\n# Use str because the yaml parser will parse things like `2020-05-06` as\n# a datetime if not included in quotes.\nreturn str(experiment_config['experiment'])\n+def _get_description(experiment_config: dict) -> Optional[str]:\n+ \"\"\"Returns the description of the experiment described by\n+ |experiment_config| as a string.\"\"\"\n+ return experiment_config.get('description')\n+\n+\ndef _get_requested_experiments():\n\"\"\"Return requested experiments.\"\"\"\nreturn yaml_utils.read(REQUESTED_EXPERIMENTS_PATH)\n@@ -203,10 +210,11 @@ def run_requested_experiment(dry_run):\nlogs.info('Running experiment: %s with fuzzers: %s.', experiment_name,\n' '.join(fuzzers))\n- return _run_experiment(experiment_name, fuzzers, dry_run)\n+ description = _get_description(requested_experiment)\n+ return _run_experiment(experiment_name, fuzzers, description, dry_run)\n-def _run_experiment(experiment_name, fuzzers, dry_run=False):\n+def _run_experiment(experiment_name, fuzzers, description, dry_run=False):\n\"\"\"Run an experiment named |experiment_name| on |fuzzer_configs| and shut it\ndown once it terminates.\"\"\"\nlogs.info('Starting experiment: %s.', experiment_name)\n@@ -214,7 +222,7 @@ def _run_experiment(experiment_name, fuzzers, dry_run=False):\nlogs.info('Dry run. Not actually running experiment.')\nreturn\nrun_experiment.start_experiment(experiment_name, EXPERIMENT_CONFIG_FILE,\n- BENCHMARKS, fuzzers)\n+ BENCHMARKS, fuzzers, description)\ndef main():\n" }, { "change_type": "MODIFY", "old_path": "service/test_automatic_run_experiment.py", "new_path": "service/test_automatic_run_experiment.py", "diff": "@@ -28,10 +28,11 @@ EXPERIMENT = '2020-01-01'\nEXPERIMENT_REQUESTS = [{\n'experiment': datetime.date(2020, 6, 8),\n- 'fuzzers': ['aflplusplus', 'libfuzzer']\n+ 'fuzzers': ['aflplusplus', 'libfuzzer'],\n}, {\n'experiment': datetime.date(2020, 6, 5),\n- 'fuzzers': ['honggfuzz', 'afl']\n+ 'fuzzers': ['honggfuzz', 'afl'],\n+ 'description': 'Test experiment',\n}]\n@@ -94,7 +95,7 @@ def test_run_requested_experiment(mocked_get_requested_experiments,\n]\nexpected_calls = [\nmock.call(expected_experiment_name, expected_config_file,\n- expected_benchmarks, expected_fuzzers)\n+ expected_benchmarks, expected_fuzzers, 'Test experiment')\n]\nstart_experiment_call_args = mocked_start_experiment.call_args_list\nassert len(start_experiment_call_args) == 1\n" } ]

[ { "change_type": "MODIFY", "old_path": "docker/gcb/base-images.yaml", "new_path": "docker/gcb/base-images.yaml", "diff": "@@ -18,6 +18,13 @@ images:\n- gcr.io/fuzzbench/base-image:test-experiment\n- gcr.io/fuzzbench/base-image\nsteps:\n+- args:\n+ - pull\n+ - ubuntu:xenial\n+ env:\n+ - DOCKER_BUILDKIT=1\n+ id: pull-ubuntu-xenial\n+ name: docker:19.03.12\n- args:\n- build\n- --tag\n@@ -31,6 +38,8 @@ steps:\n- --file\n- docker/base-image/Dockerfile\n- docker/base-image\n- env: DOCKER_BUILDKIT=1\n+ env:\n+ - DOCKER_BUILDKIT=1\nid: base-image\nname: docker:19.03.12\n+ wait_for: []\n" }, { "change_type": "MODIFY", "old_path": "experiment/build/generate_cloudbuild.py", "new_path": "experiment/build/generate_cloudbuild.py", "diff": "@@ -80,7 +80,7 @@ def create_cloud_build_spec(image_templates,\nif build_base_images:\ncloud_build_spec['steps'].append({\n'id': 'pull-ubuntu-xenial',\n- 'env': 'DOCKER_BUILDKIT=1',\n+ 'env': ['DOCKER_BUILDKIT=1'],\n'name': DOCKER_IMAGE,\n'args': ['pull', 'ubuntu:xenial'],\n})\n@@ -88,7 +88,7 @@ def create_cloud_build_spec(image_templates,\nfor image_name, image_specs in image_templates.items():\nstep = {\n'id': image_name,\n- 'env': 'DOCKER_BUILDKIT=1',\n+ 'env': ['DOCKER_BUILDKIT=1'],\n'name': DOCKER_IMAGE,\n}\nstep['args'] = [\n" }, { "change_type": "MODIFY", "old_path": "experiment/build/test_generate_cloudbuild.py", "new_path": "experiment/build/test_generate_cloudbuild.py", "diff": "@@ -40,12 +40,12 @@ def test_generate_cloud_build_spec_build_base_image():\nexpected_spec = {\n'steps': [{\n'id': 'pull-ubuntu-xenial',\n- 'env': 'DOCKER_BUILDKIT=1',\n+ 'env': ['DOCKER_BUILDKIT=1'],\n'name': 'docker:19.03.12',\n'args': ['pull', 'ubuntu:xenial']\n}, {\n'id': 'base-image',\n- 'env': 'DOCKER_BUILDKIT=1',\n+ 'env': ['DOCKER_BUILDKIT=1'],\n'name': 'docker:19.03.12',\n'args': [\n'build', '--tag', 'gcr.io/fuzzbench/base-image', '--tag',\n@@ -90,7 +90,7 @@ def test_generate_cloud_build_spec_build_fuzzer_benchmark():\nexpected_spec = {\n'steps': [{\n'id': 'afl-zlib-builder-intermediate',\n- 'env': 'DOCKER_BUILDKIT=1',\n+ 'env': ['DOCKER_BUILDKIT=1'],\n'name': 'docker:19.03.12',\n'args': [\n'build', '--tag',\n@@ -159,7 +159,7 @@ def test_generate_cloud_build_spec_build_benchmark_coverage():\nexpected_spec = {\n'steps': [{\n'id': 'zlib-project-builder',\n- 'env': 'DOCKER_BUILDKIT=1',\n+ 'env': ['DOCKER_BUILDKIT=1'],\n'name': 'docker:19.03.12',\n'args': [\n'build', '--tag', 'gcr.io/fuzzbench/builders/benchmark/zlib',\n@@ -172,7 +172,7 @@ def test_generate_cloud_build_spec_build_benchmark_coverage():\n'wait_for': []\n}, {\n'id': 'coverage-zlib-builder-intermediate',\n- 'env': 'DOCKER_BUILDKIT=1',\n+ 'env': ['DOCKER_BUILDKIT=1'],\n'name': 'docker:19.03.12',\n'args': [\n'build', '--tag',\n@@ -188,7 +188,7 @@ def test_generate_cloud_build_spec_build_benchmark_coverage():\n'wait_for': ['zlib-project-builder']\n}, {\n'id': 'coverage-zlib-builder',\n- 'env': 'DOCKER_BUILDKIT=1',\n+ 'env': ['DOCKER_BUILDKIT=1'],\n'name': 'docker:19.03.12',\n'args': [\n'build', '--tag', 'gcr.io/fuzzbench/builders/coverage/zlib',\n" } ]

[ { "change_type": "MODIFY", "old_path": "conftest.py", "new_path": "conftest.py", "diff": "@@ -80,6 +80,7 @@ def experiment(environ): # pylint: disable=redefined-outer-name,unused-argument\nos.environ['EXPERIMENT_FILESTORE'] = 'gs://experiment-data'\nos.environ['REPORT_FILESTORE'] = 'gs://web-bucket'\nos.environ['CLOUD_PROJECT'] = 'fuzzbench'\n+ os.environ['DOCKER_REGISTRY'] = 'gcr.io/fuzzbench'\[email protected]\n" }, { "change_type": "MODIFY", "old_path": "experiment/build/gcb_build.py", "new_path": "experiment/build/gcb_build.py", "diff": "@@ -47,7 +47,7 @@ def _get_buildable_images(fuzzer=None, benchmark=None):\ndef build_base_images():\n\"\"\"Build base images on GCB.\"\"\"\nimage_templates = {'base-image': _get_buildable_images()['base-image']}\n- config = generate_cloudbuild.create_cloud_build_spec(image_templates,\n+ config = generate_cloudbuild.create_cloudbuild_spec(image_templates,\nbuild_base_images=True)\n_build(config, 'base-images')\n@@ -61,7 +61,7 @@ def build_coverage(benchmark):\nif (image_name == (benchmark + '-project-builder') or\nimage_specs['type'] == 'coverage')\n}\n- config = generate_cloudbuild.create_cloud_build_spec(image_templates,\n+ config = generate_cloudbuild.create_cloudbuild_spec(image_templates,\nbenchmark=benchmark)\nconfig_name = 'benchmark-{benchmark}-coverage'.format(benchmark=benchmark)\n_build(config, config_name)\n@@ -110,7 +110,7 @@ def build_fuzzer_benchmark(fuzzer: str, benchmark: str):\nif image_specs['type'] in ('base', 'coverage', 'dispatcher'):\ncontinue\nimage_templates[image_name] = image_specs\n- config = generate_cloudbuild.create_cloud_build_spec(image_templates)\n+ config = generate_cloudbuild.create_cloudbuild_spec(image_templates)\nconfig_name = 'benchmark-{benchmark}-fuzzer-{fuzzer}'.format(\nbenchmark=benchmark, fuzzer=fuzzer)\n" }, { "change_type": "MODIFY", "old_path": "experiment/build/generate_cloudbuild.py", "new_path": "experiment/build/generate_cloudbuild.py", "diff": "import os\nimport posixpath\n-from common import yaml_utils\n-from common import experiment_utils\nfrom common import experiment_path as exp_path\n+from common import experiment_utils\n+from common import yaml_utils\nfrom common.utils import ROOT_DIR\nfrom experiment.build import build_utils\nDOCKER_IMAGE = 'docker:19.03.12'\n-PROJECT_DOCKER_REGISTRY = 'gcr.io/fuzzbench'\ndef get_experiment_tag_for_image(image_specs, tag_by_experiment=True):\n@@ -60,7 +59,12 @@ def coverage_steps(benchmark):\nreturn steps\n-def create_cloud_build_spec(image_templates,\n+def get_docker_registry():\n+ \"\"\"Returns the docker registry for this experiment.\"\"\"\n+ return os.environ['DOCKER_REGISTRY']\n+\n+\n+def create_cloudbuild_spec(image_templates,\nbenchmark='',\nbuild_base_images=False):\n\"\"\"Generates Cloud Build specification.\n@@ -73,18 +77,20 @@ def create_cloud_build_spec(image_templates,\nReturns:\nGCB build steps.\n\"\"\"\n- cloud_build_spec = {'steps': [], 'images': []}\n+ cloudbuild_spec = {'steps': [], 'images': []}\n# Workaround for bug https://github.com/moby/moby/issues/40262.\n# This is only needed for base-image as it inherits from ubuntu:xenial.\nif build_base_images:\n- cloud_build_spec['steps'].append({\n+ cloudbuild_spec['steps'].append({\n'id': 'pull-ubuntu-xenial',\n'env': ['DOCKER_BUILDKIT=1'],\n'name': DOCKER_IMAGE,\n'args': ['pull', 'ubuntu:xenial'],\n})\n+ docker_registry = get_docker_registry()\n+\nfor image_name, image_specs in image_templates.items():\nstep = {\n'id': image_name,\n@@ -93,8 +99,7 @@ def create_cloud_build_spec(image_templates,\n}\nstep['args'] = [\n'build', '--tag',\n- posixpath.join(PROJECT_DOCKER_REGISTRY,\n- image_specs['tag']), '--tag',\n+ posixpath.join(docker_registry, image_specs['tag']), '--tag',\nget_experiment_tag_for_image(image_specs), '--cache-from',\nget_experiment_tag_for_image(image_specs, tag_by_experiment=False),\n'--build-arg', 'BUILDKIT_INLINE_CACHE=1'\n@@ -113,24 +118,24 @@ def create_cloud_build_spec(image_templates,\ncontinue\nstep['wait_for'] += [dependency]\n- cloud_build_spec['steps'].append(step)\n- cloud_build_spec['images'].append(\n+ cloudbuild_spec['steps'].append(step)\n+ cloudbuild_spec['images'].append(\nget_experiment_tag_for_image(image_specs))\n- cloud_build_spec['images'].append(\n+ cloudbuild_spec['images'].append(\nget_experiment_tag_for_image(image_specs, tag_by_experiment=False))\nif any(image_specs['type'] in 'coverage'\nfor _, image_specs in image_templates.items()):\n- cloud_build_spec['steps'] += coverage_steps(benchmark)\n+ cloudbuild_spec['steps'] += coverage_steps(benchmark)\n- return cloud_build_spec\n+ return cloudbuild_spec\ndef main():\n\"\"\"Write base-images build spec when run from command line.\"\"\"\nimage_templates = yaml_utils.read(\nos.path.join(ROOT_DIR, 'docker', 'image_types.yaml'))\n- base_images_spec = create_cloud_build_spec(\n+ base_images_spec = create_cloudbuild_spec(\n{'base-image': image_templates['base-image']}, build_base_images=True)\nbase_images_spec_file = os.path.join(ROOT_DIR, 'docker', 'gcb',\n'base-images.yaml')\n" }, { "change_type": "MODIFY", "old_path": "experiment/build/test_generate_cloudbuild.py", "new_path": "experiment/build/test_generate_cloudbuild.py", "diff": "# limitations under the License.\n\"\"\"Tests for generate_cloudbuild.py.\"\"\"\n-import os\n-\n-from unittest.mock import patch\n-\nfrom experiment.build import generate_cloudbuild\n+# pylint: disable=unused-argument\n+\[email protected](os.environ, {\n- 'CLOUD_PROJECT': 'fuzzbench',\n- 'EXPERIMENT': 'test-experiment'\n-})\n-def test_generate_cloud_build_spec_build_base_image():\n+def test_generate_cloudbuild_spec_build_base_image(experiment):\n\"\"\"Tests cloud build configuration yaml for the base image.\"\"\"\nimage_templates = {\n'base-image': {\n@@ -34,7 +28,7 @@ def test_generate_cloud_build_spec_build_base_image():\n'type': 'base'\n}\n}\n- generated_spec = generate_cloudbuild.create_cloud_build_spec(\n+ generated_spec = generate_cloudbuild.create_cloudbuild_spec(\nimage_templates, build_base_images=True)\nexpected_spec = {\n@@ -65,11 +59,7 @@ def test_generate_cloud_build_spec_build_base_image():\nassert generated_spec == expected_spec\[email protected](os.environ, {\n- 'CLOUD_PROJECT': 'fuzzbench',\n- 'EXPERIMENT': 'test-experiment'\n-})\n-def test_generate_cloud_build_spec_build_fuzzer_benchmark():\n+def test_generate_cloudbuild_spec_build_fuzzer_benchmark(experiment):\n\"\"\"Tests cloud build configuration yaml for a fuzzer-benchmark build.\"\"\"\nimage_templates = {\n'afl-zlib-builder-intermediate': {\n@@ -84,8 +74,7 @@ def test_generate_cloud_build_spec_build_fuzzer_benchmark():\n}\n}\n- generated_spec = generate_cloudbuild.create_cloud_build_spec(\n- image_templates)\n+ generated_spec = generate_cloudbuild.create_cloudbuild_spec(image_templates)\nexpected_spec = {\n'steps': [{\n@@ -113,14 +102,7 @@ def test_generate_cloud_build_spec_build_fuzzer_benchmark():\nassert generated_spec == expected_spec\[email protected](\n- os.environ, {\n- 'CLOUD_PROJECT': 'fuzzbench',\n- 'EXPERIMENT': 'test-experiment',\n- 'EXPERIMENT_FILESTORE': 'gs://fuzzbench-data',\n- 'WORK': '/work',\n- })\n-def test_generate_cloud_build_spec_build_benchmark_coverage():\n+def test_generate_cloudbuild_spec_build_benchmark_coverage(experiment):\n\"\"\"Tests cloud build configuration yaml for a benchmark coverage build.\"\"\"\nimage_templates = {\n'zlib-project-builder': {\n@@ -153,7 +135,7 @@ def test_generate_cloud_build_spec_build_benchmark_coverage():\n}\n}\n- generated_spec = generate_cloudbuild.create_cloud_build_spec(\n+ generated_spec = generate_cloudbuild.create_cloudbuild_spec(\nimage_templates, benchmark='zlib')\nexpected_spec = {\n@@ -218,7 +200,7 @@ def test_generate_cloud_build_spec_build_benchmark_coverage():\n'gcr.io/cloud-builders/gsutil',\n'args': [\n'-m', 'cp', '/workspace/out/coverage-build-zlib.tar.gz',\n- 'gs://fuzzbench-data/test-experiment/coverage-binaries/'\n+ 'gs://experiment-data/test-experiment/coverage-binaries/'\n]\n}],\n'images': [\n" } ]

[ { "change_type": "MODIFY", "old_path": "experiment/resources/dispatcher-startup-script-template.sh", "new_path": "experiment/resources/dispatcher-startup-script-template.sh", "diff": "# See the License for the specific language governing permissions and\n# limitations under the License.\n+docker-credential-gcr configure-docker -include-artifact-registry\necho 0 | sudo tee /proc/sys/kernel/yama/ptrace_scope\ndocker run --rm \\\n-e INSTANCE_NAME={{instance_name}} -e EXPERIMENT={{experiment}} \\\n" }, { "change_type": "MODIFY", "old_path": "experiment/resources/runner-startup-script-template.sh", "new_path": "experiment/resources/runner-startup-script-template.sh", "diff": "@@ -23,6 +23,7 @@ echo core >/proc/sys/kernel/core_pattern\n# Start docker.\n{% if not local_experiment %}\n+docker-credential-gcr configure-docker -include-artifact-registry\nwhile ! docker pull {{docker_image_url}}\ndo\necho 'Error pulling image, retrying...'\n" }, { "change_type": "MODIFY", "old_path": "experiment/test_scheduler.py", "new_path": "experiment/test_scheduler.py", "diff": "@@ -94,6 +94,7 @@ def test_create_trial_instance(benchmark, expected_image, expected_target,\nand creates a startup script for the instance, as we expect it to.\"\"\"\nexpected_startup_script = '''# Start docker.\n+docker-credential-gcr configure-docker -include-artifact-registry\nwhile ! docker pull {docker_image_url}\ndo\necho 'Error pulling image, retrying...'\n" } ]

[ { "change_type": "MODIFY", "old_path": "common/experiment_utils.py", "new_path": "common/experiment_utils.py", "diff": "@@ -81,16 +81,6 @@ def get_crashes_archive_name(cycle: int) -> str:\nreturn 'crashes-%04d.tar.gz' % cycle\n-def get_base_docker_tag(cloud_project=None):\n- \"\"\"Returns the base docker tag (i.e. Docker repo URL) given cloud_project.\n- If cloud is not provided, then the value of the environment variable\n- CLOUD_PROJECT is used.\"\"\"\n- # Google Cloud Docker repos use the form \"gcr.io/$CLOUD_PROJECT\"\n- if cloud_project is None:\n- cloud_project = get_cloud_project()\n- return posixpath.join('gcr.io', cloud_project)\n-\n-\ndef is_local_experiment():\n\"\"\"Returns True if running a local experiment.\"\"\"\nreturn bool(environment.get('LOCAL_EXPERIMENT'))\n" }, { "change_type": "MODIFY", "old_path": "experiment/build/generate_cloudbuild.py", "new_path": "experiment/build/generate_cloudbuild.py", "diff": "@@ -27,8 +27,7 @@ DOCKER_IMAGE = 'docker:19.03.12'\ndef get_experiment_tag_for_image(image_specs, tag_by_experiment=True):\n\"\"\"Returns the registry with the experiment tag for given image.\"\"\"\n- tag = posixpath.join(experiment_utils.get_base_docker_tag(),\n- image_specs['tag'])\n+ tag = posixpath.join(get_docker_registry(), image_specs['tag'])\nif tag_by_experiment:\ntag += ':' + experiment_utils.get_experiment_name()\nreturn tag\n@@ -43,8 +42,8 @@ def coverage_steps(benchmark):\nDOCKER_IMAGE,\n'args': [\n'run', '-v', '/workspace/out:/host-out',\n- posixpath.join(experiment_utils.get_base_docker_tag(), 'builders',\n- 'coverage', benchmark) + ':' +\n+ posixpath.join(get_docker_registry(), 'builders', 'coverage',\n+ benchmark) + ':' +\nexperiment_utils.get_experiment_name(), '/bin/bash', '-c',\n'cd /out; tar -czvf /host-out/coverage-build-' + benchmark +\n'.tar.gz * /src /work'\n" } ]

[ { "change_type": "MODIFY", "old_path": "Makefile", "new_path": "Makefile", "diff": "@@ -53,7 +53,7 @@ ${VENV_ACTIVATE}: requirements.txt\ninstall-dependencies: ${VENV_ACTIVATE}\n-docker/generated.mk: docker/generate_makefile.py fuzzers benchmarks ${VENV_ACTIVATE}\n+docker/generated.mk: docker/generate_makefile.py docker/image_types.yaml fuzzers benchmarks ${VENV_ACTIVATE}\nsource ${VENV_ACTIVATE} && PYTHONPATH=. python3 $< > $@\npresubmit: install-dependencies\n" }, { "change_type": "MODIFY", "old_path": "docker/image_types.yaml", "new_path": "docker/image_types.yaml", "diff": "- 'parent_image=gcr.io/fuzzbench/builders/coverage/{benchmark}-intermediate'\ndepends_on:\n- 'coverage-{benchmark}-builder-intermediate'\n+ - 'base-image'\ndockerfile: 'docker/benchmark-builder/Dockerfile'\ncontext: '.'\ntag: 'builders/coverage/{benchmark}'\n- 'parent_image=gcr.io/fuzzbench/builders/{fuzzer}/{benchmark}-intermediate'\ndepends_on:\n- '{fuzzer}-{benchmark}-builder-intermediate'\n+ - 'base-image'\ndockerfile: 'docker/benchmark-builder/Dockerfile'\ncontext: '.'\ntag: 'builders/{fuzzer}/{benchmark}'\n'{fuzzer}-{benchmark}-intermediate-runner':\ndepends_on:\n- '{fuzzer}-{benchmark}-builder'\n+ - 'base-image'\ndockerfile: 'fuzzers/{fuzzer}/runner.Dockerfile'\ncontext: 'fuzzers/{fuzzer}'\ntag: 'runners/{fuzzer}/{benchmark}-intermediate'\n" } ]

[ { "change_type": "MODIFY", "old_path": "Makefile", "new_path": "Makefile", "diff": "@@ -54,7 +54,7 @@ ${VENV_ACTIVATE}: requirements.txt\ninstall-dependencies: ${VENV_ACTIVATE}\ndocker/generated.mk: docker/generate_makefile.py docker/image_types.yaml fuzzers benchmarks ${VENV_ACTIVATE}\n- source ${VENV_ACTIVATE} && PYTHONPATH=. python3 $< > $@\n+ source ${VENV_ACTIVATE} && PYTHONPATH=. python3 $< $@\npresubmit: install-dependencies\nsource ${VENV_ACTIVATE} && python3 presubmit.py\n" }, { "change_type": "MODIFY", "old_path": "docker/generate_makefile.py", "new_path": "docker/generate_makefile.py", "diff": "\"\"\"Simple generator for local Makefile rules.\"\"\"\nimport os\n+import sys\nfrom common import yaml_utils\nfrom common import benchmark_utils\n@@ -24,30 +25,33 @@ BASE_TAG = \"gcr.io/fuzzbench\"\nBENCHMARK_DIR = benchmark_utils.BENCHMARKS_DIR\n-def _print_benchmark_fuzz_target(benchmarks):\n- \"\"\"Prints benchmark variables from benchmark.yaml files.\"\"\"\n+def _get_benchmark_fuzz_target(benchmarks):\n+ \"\"\"Returns benchmark variables from benchmark.yaml files.\"\"\"\n+ variables = ''\nfor benchmark in benchmarks:\nbenchmark_vars = yaml_utils.read(\nos.path.join(BENCHMARK_DIR, benchmark, 'benchmark.yaml'))\n- print(benchmark + '-fuzz-target=' + benchmark_vars['fuzz_target'])\n- print()\n+ variables += (benchmark + '-fuzz-target=' +\n+ benchmark_vars['fuzz_target'] + '\\n')\n+ variables += '\\n'\n+ return variables\n-def _print_makefile_run_template(image):\n+def _get_makefile_run_template(image):\nfuzzer = image['fuzzer']\nbenchmark = image['benchmark']\n+ section = ''\n- for run_type in ('run', 'debug', 'test-run', 'repro-bugs'):\n- if run_type == 'repro-bugs':\n- bugs_testcases_dir = os.path.join(BENCHMARK_DIR, benchmark,\n- 'testcases')\n- if not os.path.isdir(bugs_testcases_dir):\n- continue\n- print(('{run_type}-{fuzzer}-{benchmark}: ' +\n- '.{fuzzer}-{benchmark}-runner').format(run_type=run_type,\n- benchmark=benchmark,\n- fuzzer=fuzzer))\n- print('\\\n+ run_types = ['run', 'debug', 'test-run']\n+ testcases_dir = os.path.join(BENCHMARK_DIR, benchmark, 'testcases')\n+ if os.path.exists(testcases_dir):\n+ run_types.append('repro-bugs')\n+\n+ for run_type in run_types:\n+ section += (\n+ f'{run_type}-{fuzzer}-{benchmark}: .{fuzzer}-{benchmark}-runner\\n')\n+\n+ section += f'\\\n\\tdocker run \\\\\\n\\\n\\t--cpus=1 \\\\\\n\\\n\\t--cap-add SYS_NICE \\\\\\n\\\n@@ -58,116 +62,120 @@ def _print_makefile_run_template(image):\n\\t-e FUZZER={fuzzer} \\\\\\n\\\n\\t-e BENCHMARK={benchmark} \\\\\\n\\\n\\t-e FUZZ_TARGET=$({benchmark}-fuzz-target) \\\\\\\n-'.format(fuzzer=fuzzer, benchmark=benchmark))\n+\\n'\n- params = \"\"\nif run_type == 'test-run':\n- print('\\t-e MAX_TOTAL_TIME=20 \\\\\\n\\t-e SNAPSHOT_PERIOD=10 \\\\')\n+ section += '\\t-e MAX_TOTAL_TIME=20 \\\\\\n\\t-e SNAPSHOT_PERIOD=10 \\\\\\n'\nif run_type == 'debug':\n- print('\\t--entrypoint \"/bin/bash\" \\\\\\n\\t-it ', end='')\n+ section += '\\t--entrypoint \"/bin/bash\" \\\\\\n\\t-it '\nelif run_type == 'repro-bugs':\n- print('\\t-v {path}:/testcases \\\\\\n\\\n-\\t--entrypoint /bin/bash '.format(path=bugs_testcases_dir),\n- end='')\n- params = \" -c \\\"for f in /testcases/*; do\\\n- echo _________________________________________;\\\n- echo \\\\$$f:;\\\n- \\\\$$OUT/\\\\$$FUZZ_TARGET -timeout=25\\\n- -rss_limit_mb=2560 \\\\$$f; done;\\\"\"\n-\n+ section += f'\\t-v {testcases_dir}:/testcases \\\\\\n\\t'\n+ section += '--entrypoint /bin/bash '\n+ section += os.path.join(BASE_TAG, image['tag'])\n+ section += ' -c \"for f in /testcases/*; do '\n+ section += 'echo _________________________________________; '\n+ section += 'echo \\\\$$f:; '\n+ section += '\\\\$$OUT/\\\\$$FUZZ_TARGET -timeout=25 -rss_limit_mb=2560 '\n+ section += '\\\\$$f; done;\" '\n+ section += '\\n\\n'\n+ continue\nelif run_type == 'run':\n- print('\\t-it ', end='')\n+ section += '\\t-it '\nelse:\n- print('\\t', end='')\n+ section += '\\t'\n- if params:\n- print(os.path.join(BASE_TAG, image['tag']), end='')\n- print(params)\n- else:\n- print(os.path.join(BASE_TAG, image['tag']))\n- print()\n+ section += os.path.join(BASE_TAG, image['tag'])\n+ section += '\\n\\n'\n+ return section\n-# TODO(Tanq16): Function must return a string as opposed to printing it.\n-def print_rules_for_image(name, image):\n- \"\"\"Print makefile section for given image to stdout.\"\"\"\n+def get_rules_for_image(name, image):\n+ \"\"\"Returns makefile section for |image|.\"\"\"\nif not ('base-' in name or 'dispatcher-' in name):\n- print('.', end='')\n- print(name + ':', end='')\n+ section = '.'\n+ else:\n+ section = ''\n+ section += name + ':'\nif 'depends_on' in image:\nfor dep in image['depends_on']:\nif 'base' in dep:\n- print(' ' + dep, end='')\n+ section += ' ' + dep\nelse:\n- print(' .' + dep, end='')\n- print()\n+ section += ' .' + dep\n+ section += '\\n'\nif 'base-' in name:\n- print('\\tdocker pull ubuntu:xenial')\n- print('\\tdocker build \\\\')\n- print('\\t--tag ' + os.path.join(BASE_TAG, image['tag']) + ' \\\\')\n- print('\\t--build-arg BUILDKIT_INLINE_CACHE=1 \\\\')\n- print('\\t--cache-from ' + os.path.join(BASE_TAG, image['tag']) + ' \\\\')\n+ section += '\\tdocker pull ubuntu:xenial\\n'\n+ section += '\\tdocker build \\\\\\n'\n+ section += '\\t--tag ' + os.path.join(BASE_TAG, image['tag']) + ' \\\\\\n'\n+ section += '\\t--build-arg BUILDKIT_INLINE_CACHE=1 \\\\\\n'\n+ section += ('\\t--cache-from ' + os.path.join(BASE_TAG, image['tag']) +\n+ ' \\\\\\n')\n+\nif 'build_arg' in image:\nfor arg in image['build_arg']:\n- print('\\t--build-arg ' + arg + ' \\\\')\n+ section += '\\t--build-arg ' + arg + ' \\\\\\n'\nif 'dockerfile' in image:\n- print('\\t--file ' + image['dockerfile'] + ' \\\\')\n- print('\\t' + image['context'])\n- print()\n+ section += '\\t--file ' + image['dockerfile'] + ' \\\\\\n'\n+ section += '\\t' + image['context'] + '\\n'\n+ section += '\\n'\n# Print run, debug, test-run rules if image is a runner.\nif 'runner' in name and not ('intermediate' in name or 'base' in name):\n- _print_makefile_run_template(image)\n+ section += _get_makefile_run_template(image)\n+ return section\ndef main():\n- \"\"\"Generates Makefile with docker image build rules.\"\"\"\n+ \"\"\"Writes Makefile with docker image build rules to sys.argv[1].\"\"\"\n+ if len(sys.argv) != 2:\n+ print(f'Usage: {sys.argv[0]} <makefile>')\n+ return 1\n+ makefile_path = sys.argv[1]\n+ makefile_contents = generate_makefile()\n+ with open(makefile_path, 'w') as file_handle:\n+ file_handle.write(makefile_contents)\n+ return 0\n+\n+\n+def generate_makefile():\n+ \"\"\"Generates the contents of the makefile and returns it.\"\"\"\nfuzzers = fuzzer_utils.get_fuzzer_names()\nbenchmarks = benchmark_utils.get_all_benchmarks()\nbuildable_images = docker_images.get_images_to_build(fuzzers, benchmarks)\n- print('export DOCKER_BUILDKIT := 1')\n+ makefile = 'export DOCKER_BUILDKIT := 1\\n\\n'\n# Print oss-fuzz benchmarks property variables.\n- _print_benchmark_fuzz_target(benchmarks)\n+ makefile += _get_benchmark_fuzz_target(benchmarks)\nfor name, image in buildable_images.items():\n- print_rules_for_image(name, image)\n+ makefile += get_rules_for_image(name, image)\n# Print build targets for all fuzzer-benchmark pairs (including coverage).\nfuzzers.append('coverage')\nfor fuzzer in fuzzers:\n- image_type = \"runner\"\n+ image_type = 'runner'\nif 'coverage' in fuzzer:\n- image_type = \"builder\"\n+ image_type = 'builder'\nfor benchmark in benchmarks:\n- print(('build-{fuzzer}-{benchmark}: ' +\n- '.{fuzzer}-{benchmark}-{image_type}\\n').format(\n- fuzzer=fuzzer,\n- benchmark=benchmark,\n- image_type=image_type))\n- print()\n+ makefile += (f'build-{fuzzer}-{benchmark}: ' +\n+ f'.{fuzzer}-{benchmark}-{image_type}\\n')\n+ makefile += '\\n'\n# Print fuzzer-all benchmarks build targets.\nfor fuzzer in fuzzers:\n- all_build_targets = ' '.join([\n- 'build-{0}-{1}'.format(fuzzer, benchmark)\n- for benchmark in benchmarks\n- ])\n- print('build-{fuzzer}-all: {all_targets}'.format(\n- fuzzer=fuzzer, all_targets=all_build_targets))\n- all_test_run_targets = ' '.join([\n- 'test-run-{0}-{1}'.format(fuzzer, benchmark)\n- for benchmark in benchmarks\n- ])\n- print('test-run-{fuzzer}-all: {all_targets}'.format(\n- fuzzer=fuzzer, all_targets=all_test_run_targets))\n+ all_build_targets = ' '.join(\n+ [f'build-{fuzzer}-{benchmark}' for benchmark in benchmarks])\n+ makefile += f'build-{fuzzer}-all: {all_build_targets}\\n'\n+ all_test_run_targets = ' '.join(\n+ [f'test-run-{fuzzer}-{benchmark}' for benchmark in benchmarks])\n+ makefile += f'test-run-{fuzzer}-all: {all_test_run_targets}\\n'\n# Print all targets build target.\n- all_build_targets = ' '.join(\n- ['build-{0}-all'.format(name) for name in fuzzers])\n- print('build-all: {all_targets}'.format(all_targets=all_build_targets))\n+ all_build_targets = ' '.join([f'build-{fuzzer}-all' for fuzzer in fuzzers])\n+ makefile += f'build-all: {all_build_targets}'\n+ return makefile\nif __name__ == '__main__':\n- main()\n+ sys.exit(main())\n" }, { "change_type": "MODIFY", "old_path": "docker/test_generate_makefile.py", "new_path": "docker/test_generate_makefile.py", "diff": "# limitations under the License.\n\"\"\"Tests for generate_makefile.py.\"\"\"\n-from unittest.mock import call\n-from unittest.mock import patch\n-\nfrom docker import generate_makefile\n-@patch('builtins.print')\n-def test_print_makefile_build(mocked_print):\n+def test_get_rules_for_image():\n\"\"\"Tests result of a makefile generation for an image.\"\"\"\nname = 'afl-zlib-builder-intermediate'\n@@ -32,27 +28,21 @@ def test_print_makefile_build(mocked_print):\n'build_arg': ['parent_image=gcr.io/fuzzbench/builders/benchmark/zlib']\n}\n- generate_makefile.print_rules_for_image(name, image)\n- assert mocked_print.mock_calls == [\n- call('.', end=''),\n- call('afl-zlib-builder-intermediate:', end=''),\n- call(' .zlib-project-builder', end=''),\n- call(),\n- call('\\tdocker build \\\\'),\n- call('\\t--tag gcr.io/fuzzbench/builders/afl/zlib-intermediate \\\\'),\n- call('\\t--build-arg BUILDKIT_INLINE_CACHE=1 \\\\'),\n- call('\\t--cache-from gcr.io/fuzzbench/builders/afl/zlib-intermediate \\\\'\n- ),\n- call('\\t--build-arg parent_image=gcr.io/' +\n- 'fuzzbench/builders/benchmark/zlib \\\\'),\n- call('\\t--file fuzzers/afl/builder.Dockerfile \\\\'),\n- call('\\tfuzzers/afl'),\n- call()\n- ]\n+ rules_for_image = generate_makefile.get_rules_for_image(name, image)\n+ assert rules_for_image == (\n+ '.afl-zlib-builder-intermediate: .zlib-project-builder\\n'\n+ '\\tdocker build \\\\\\n'\n+ '\\t--tag gcr.io/fuzzbench/builders/afl/zlib-intermediate \\\\\\n'\n+ '\\t--build-arg BUILDKIT_INLINE_CACHE=1 \\\\\\n'\n+ '\\t--cache-from gcr.io/fuzzbench/builders/afl/zlib-intermediate \\\\\\n'\n+ '\\t--build-arg parent_image=gcr.io/fuzzbench/builders/benchmark/zlib \\\\'\n+ '\\n'\n+ '\\t--file fuzzers/afl/builder.Dockerfile \\\\\\n'\n+ '\\tfuzzers/afl\\n'\n+ '\\n')\n-@patch('builtins.print')\n-def test_print_makefile_runner_image(mocked_print):\n+def test_get_rules_for_runner_image():\n\"\"\"Tests result of a makefile generation for a runner image.\"\"\"\nname = 'afl-zlib-runner'\n@@ -65,26 +55,19 @@ def test_print_makefile_runner_image(mocked_print):\n'build_arg': ['fuzzer=afl', 'benchmark=zlib'],\n'depends_on': ['afl-zlib-builder', 'afl-zlib-intermediate-runner']\n}\n-\n- generate_makefile.print_rules_for_image(name, image)\n-\n- assert mocked_print.mock_calls == [\n- call('.', end=''),\n- call('afl-zlib-runner:', end=''),\n- call(' .afl-zlib-builder', end=''),\n- call(' .afl-zlib-intermediate-runner', end=''),\n- call(),\n- call('\\tdocker build \\\\'),\n- call('\\t--tag gcr.io/fuzzbench/runners/afl/zlib \\\\'),\n- call('\\t--build-arg BUILDKIT_INLINE_CACHE=1 \\\\'),\n- call('\\t--cache-from gcr.io/fuzzbench/runners/afl/zlib \\\\'),\n- call('\\t--build-arg fuzzer=afl \\\\'),\n- call('\\t--build-arg benchmark=zlib \\\\'),\n- call('\\t--file docker/benchmark-runner/Dockerfile \\\\'),\n- call('\\t.'),\n- call(),\n- call('run-afl-zlib: .afl-zlib-runner'),\n- call('\\tdocker run \\\\\\n\\\n+ rules_for_image = generate_makefile.get_rules_for_image(name, image)\n+ assert rules_for_image == (\n+ '.afl-zlib-runner: .afl-zlib-builder .afl-zlib-intermediate-runner\\n'\n+ '\\tdocker build \\\\\\n'\n+ '\\t--tag gcr.io/fuzzbench/runners/afl/zlib \\\\\\n'\n+ '\\t--build-arg BUILDKIT_INLINE_CACHE=1 \\\\\\n'\n+ '\\t--cache-from gcr.io/fuzzbench/runners/afl/zlib \\\\\\n'\n+ '\\t--build-arg fuzzer=afl \\\\\\n'\n+ '\\t--build-arg benchmark=zlib \\\\\\n'\n+ '\\t--file docker/benchmark-runner/Dockerfile \\\\\\n'\n+ '\\t.\\n\\n'\n+ 'run-afl-zlib: .afl-zlib-runner\\n' + ('\\\n+\\tdocker run \\\\\\n\\\n\\t--cpus=1 \\\\\\n\\\n\\t--cap-add SYS_NICE \\\\\\n\\\n\\t--cap-add SYS_PTRACE \\\\\\n\\\n@@ -93,12 +76,10 @@ def test_print_makefile_runner_image(mocked_print):\n\\t-e TRIAL_ID=1 \\\\\\n\\\n\\t-e FUZZER=afl \\\\\\n\\\n\\t-e BENCHMARK=zlib \\\\\\n\\\n-\\t-e FUZZ_TARGET=$(zlib-fuzz-target) \\\\'),\n- call('\\t-it ', end=''),\n- call('gcr.io/fuzzbench/runners/afl/zlib'),\n- call(),\n- call('debug-afl-zlib: .afl-zlib-runner'),\n- call('\\tdocker run \\\\\\n\\\n+\\t-e FUZZ_TARGET=$(zlib-fuzz-target) \\\\\\\n+\\n') + '\\t-it gcr.io/fuzzbench/runners/afl/zlib\\n\\n'\n+ 'debug-afl-zlib: .afl-zlib-runner\\n' + ('\\\n+\\tdocker run \\\\\\n\\\n\\t--cpus=1 \\\\\\n\\\n\\t--cap-add SYS_NICE \\\\\\n\\\n\\t--cap-add SYS_PTRACE \\\\\\n\\\n@@ -107,12 +88,11 @@ def test_print_makefile_runner_image(mocked_print):\n\\t-e TRIAL_ID=1 \\\\\\n\\\n\\t-e FUZZER=afl \\\\\\n\\\n\\t-e BENCHMARK=zlib \\\\\\n\\\n-\\t-e FUZZ_TARGET=$(zlib-fuzz-target) \\\\'),\n- call('\\t--entrypoint \"/bin/bash\" \\\\\\n\\t-it ', end=''),\n- call('gcr.io/fuzzbench/runners/afl/zlib'),\n- call(),\n- call('test-run-afl-zlib: .afl-zlib-runner'),\n- call('\\tdocker run \\\\\\n\\\n+\\t-e FUZZ_TARGET=$(zlib-fuzz-target) \\\\\\\n+\\n') + '\\t--entrypoint \"/bin/bash\" \\\\\\n\\t-it gcr.io/fuzzbench/runners/afl/zlib'\n+ '\\n\\n'\n+ 'test-run-afl-zlib: .afl-zlib-runner\\n' + ('\\\n+\\tdocker run \\\\\\n\\\n\\t--cpus=1 \\\\\\n\\\n\\t--cap-add SYS_NICE \\\\\\n\\\n\\t--cap-add SYS_PTRACE \\\\\\n\\\n@@ -121,9 +101,6 @@ def test_print_makefile_runner_image(mocked_print):\n\\t-e TRIAL_ID=1 \\\\\\n\\\n\\t-e FUZZER=afl \\\\\\n\\\n\\t-e BENCHMARK=zlib \\\\\\n\\\n-\\t-e FUZZ_TARGET=$(zlib-fuzz-target) \\\\'),\n- call('\\t-e MAX_TOTAL_TIME=20 \\\\\\n\\t-e SNAPSHOT_PERIOD=10 \\\\'),\n- call('\\t', end=''),\n- call('gcr.io/fuzzbench/runners/afl/zlib'),\n- call(),\n- ]\n+\\t-e FUZZ_TARGET=$(zlib-fuzz-target) \\\\\\\n+\\n') + '\\t-e MAX_TOTAL_TIME=20 \\\\\\n\\t-e SNAPSHOT_PERIOD=10 \\\\\\n'\n+ '\\tgcr.io/fuzzbench/runners/afl/zlib\\n\\n')\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/build_and_test_run_fuzzer_benchmarks.py", "new_path": ".github/workflows/build_and_test_run_fuzzer_benchmarks.py", "diff": "@@ -96,6 +96,9 @@ def make_builds(benchmarks, fuzzer):\n\"\"\"Use make to test the fuzzer on each benchmark in |benchmarks|.\"\"\"\nfuzzer_benchmark_pairs = builder.get_fuzzer_benchmark_pairs([fuzzer],\nbenchmarks)\n+ # Sort benchmarks so that they get built in a deterministic order.\n+ fuzzer_benchmark_pairs = sorted(fuzzer_benchmark_pairs,\n+ key=lambda pair: pair[1])\nprint('Building fuzzer-benchmark pairs: {}'.format(fuzzer_benchmark_pairs))\nfor _, benchmark in fuzzer_benchmark_pairs:\nmake_target = get_make_target(fuzzer, benchmark)\n" } ]

[ { "change_type": "MODIFY", "old_path": "benchmarks/oss_fuzz_benchmark_integration.py", "new_path": "benchmarks/oss_fuzz_benchmark_integration.py", "diff": "@@ -212,7 +212,7 @@ def main():\nhelp='Date of the commit. Example: 2019-10-19T09:07:25+01:00')\nargs = parser.parse_args()\nintegrate_benchmark(args.project, args.fuzz_target, args.benchmark_name,\n- args.commit, args.date, args.oss_fuzz_corpus)\n+ args.commit, args.date)\nreturn 0\n" } ]

[ { "change_type": "MODIFY", "old_path": "common/new_process.py", "new_path": "common/new_process.py", "diff": "@@ -109,8 +109,8 @@ def execute( # pylint: disable=too-many-locals,too-many-branches\nretcode = process.returncode\n- log_message = ('Executed command: \"{command}\" returned: {retcode}.'.format(\n- command=(' '.join(command))[:LOG_LIMIT_FIELD], retcode=retcode))\n+ command_log_str = ' '.join(command)[:LOG_LIMIT_FIELD]\n+ log_message = 'Executed command: \"%s\" returned: %d.'\nif output is not None:\noutput = output.decode('utf-8', errors='ignore')\n@@ -120,8 +120,8 @@ def execute( # pylint: disable=too-many-locals,too-many-branches\nlog_extras = None\nif expect_zero and retcode != 0 and not wrapped_process.timed_out:\n- logs.error(log_message, extras=log_extras)\n+ logs.error(log_message, command_log_str, retcode, extras=log_extras)\nraise subprocess.CalledProcessError(retcode, command)\n- logs.debug(log_message, extras=log_extras)\n+ logs.debug(log_message, command_log_str, retcode, extras=log_extras)\nreturn ProcessResult(retcode, output, wrapped_process.timed_out)\n" } ]

[ { "change_type": "MODIFY", "old_path": "benchmarks/oss_fuzz_benchmark_integration.py", "new_path": "benchmarks/oss_fuzz_benchmark_integration.py", "diff": "@@ -149,11 +149,11 @@ def replace_base_builder(benchmark_dir, commit_date):\n\"\"\"Replace the parent image of the Dockerfile in |benchmark_dir|,\nbase-builder (latest), with a version of base-builder that is likely to\nbuild the project as it was on |commit_date| without issue.\"\"\"\n- base_builder_repo = _load_base_builder_docker_repo() # pylint: disable=protected-access\n+ base_builder_repo = _load_base_builder_docker_repo()\nif base_builder_repo:\nbase_builder_digest = base_builder_repo.find_digest(commit_date)\nlogs.info('Using base-builder with digest %s.', base_builder_digest)\n- _replace_base_builder_digest( # pylint: disable=protected-access\n+ _replace_base_builder_digest(\nos.path.join(benchmark_dir, 'Dockerfile'), base_builder_digest)\n" } ]

[ { "change_type": "MODIFY", "old_path": "benchmarks/oss_fuzz_benchmark_integration.py", "new_path": "benchmarks/oss_fuzz_benchmark_integration.py", "diff": "@@ -36,7 +36,7 @@ OSS_FUZZ_REPO_PATH = os.path.join(OSS_FUZZ_DIR, 'infra')\nclass GitRepoManager:\n- \"\"\"Base repo manager.\"\"\"\n+ \"\"\"Git repo manager.\"\"\"\ndef __init__(self, repo_dir):\nself.repo_dir = repo_dir\n@@ -76,6 +76,11 @@ class BaseBuilderDockerRepo:\ndef copy_oss_fuzz_files(project, commit_date, benchmark_dir):\n\"\"\"Checkout the right files from OSS-Fuzz to build the benchmark based on\n|project| and |commit_date|. Then copy them to |benchmark_dir|.\"\"\"\n+ if not os.path.exists(os.path.join(OSS_FUZZ_DIR, '.git')):\n+ logs.error(\n+ '%s is not a git repo. Try running git submodule update --init',\n+ OSS_FUZZ_DIR)\n+ raise RuntimeError('%s is not a git repo.' % OSS_FUZZ_DIR)\noss_fuzz_repo_manager = GitRepoManager(OSS_FUZZ_DIR)\nprojects_dir = os.path.join(OSS_FUZZ_DIR, 'projects', project)\ntry:\n" } ]

[ { "change_type": "MODIFY", "old_path": "common/experiment_utils.py", "new_path": "common/experiment_utils.py", "diff": "@@ -71,19 +71,26 @@ def get_trial_instance_name(experiment: str, trial_id: int) -> str:\nreturn 'r-%s-%d' % (experiment, trial_id)\n+def get_cycle_filename(basename: str, cycle: int) -> str:\n+ \"\"\"Returns a filename for a file that is relevant to a particular snapshot\n+ cycle.\"\"\"\n+ filename = basename + '-' + ('%04d' % cycle)\n+ return filename\n+\n+\ndef get_corpus_archive_name(cycle: int) -> str:\n\"\"\"Returns a corpus archive name given a cycle.\"\"\"\n- return 'corpus-archive-%04d.tar.gz' % cycle\n+ return get_cycle_filename('corpus-archive', cycle) + '.tar.gz'\ndef get_stats_filename(cycle: int) -> str:\n\"\"\"Returns a corpus archive name given a cycle.\"\"\"\n- return 'stats-%04d.json' % cycle\n+ return get_cycle_filename('stats', cycle) + '.json'\ndef get_crashes_archive_name(cycle: int) -> str:\n\"\"\"Return as crashes archive name given a cycle.\"\"\"\n- return 'crashes-%04d.tar.gz' % cycle\n+ return get_cycle_filename('crashes', cycle) + '.tar.gz'\ndef is_local_experiment():\n" } ]

[ { "change_type": "MODIFY", "old_path": ".gitignore", "new_path": ".gitignore", "diff": "@@ -19,12 +19,8 @@ __pycache__/\n.pytype/\n-# Virtual environments.\n-.env\n+# Virtualenv\n.venv\n-env/\n-venv/\n-ENV/\n# Reports generated by FuzzBench.\nreport/\n@@ -36,3 +32,7 @@ docs/vendor/\n# Auto-generated build files.\ndocker/generated.mk\n+\n+# Emacs backup files.\n+*~\n+\\#*\\#\n\\ No newline at end of file\n" } ]

[ { "change_type": "MODIFY", "old_path": "common/filestore_utils.py", "new_path": "common/filestore_utils.py", "diff": "@@ -37,15 +37,15 @@ def get_impl():\nreturn local_filestore\n-def cp(source, destination, recursive=False, parallel=False, expect_zero=True): # pylint: disable=invalid-name\n+def cp(source, destination, recursive=False, expect_zero=True, parallel=False): # pylint: disable=invalid-name\n\"\"\"Copies |source| to |destination|. If |expect_zero| is True then it can\nraise subprocess.CalledProcessError. |parallel| is only used by the gsutil\nimplementation.\"\"\"\nreturn get_impl().cp(source,\ndestination,\nrecursive=recursive,\n- parallel=parallel,\n- expect_zero=expect_zero)\n+ expect_zero=expect_zero,\n+ parallel=parallel)\ndef ls(path, must_exist=True): # pylint: disable=invalid-name\n@@ -79,3 +79,8 @@ def rsync( # pylint: disable=too-many-arguments\ngsutil_options,\noptions,\nparallel=parallel)\n+\n+\n+def cat(file_path, expect_zero=True):\n+ \"\"\"Reads the file at |file_path| and returns the result.\"\"\"\n+ return get_impl().cat(file_path, expect_zero=expect_zero)\n" }, { "change_type": "MODIFY", "old_path": "common/gsutil.py", "new_path": "common/gsutil.py", "diff": "@@ -28,7 +28,7 @@ def gsutil_command(arguments, expect_zero=True, parallel=False):\nreturn new_process.execute(command + arguments, expect_zero=expect_zero)\n-def cp(source, destination, recursive=False, parallel=False, expect_zero=True): # pylint: disable=invalid-name\n+def cp(source, destination, recursive=False, expect_zero=True, parallel=False): # pylint: disable=invalid-name\n\"\"\"Executes gsutil's \"cp\" command to copy |source| to |destination|. Uses -r\nif |recursive|. If |expect_zero| is True and the command fails then this\nfunction will raise a subprocess.CalledError.\"\"\"\n@@ -37,7 +37,7 @@ def cp(source, destination, recursive=False, parallel=False, expect_zero=True):\ncommand.append('-r')\ncommand.extend([source, destination])\n- return gsutil_command(command, parallel=parallel, expect_zero=expect_zero)\n+ return gsutil_command(command, expect_zero=expect_zero, parallel=parallel)\ndef ls(path, must_exist=True): # pylint: disable=invalid-name\n@@ -84,3 +84,12 @@ def rsync( # pylint: disable=too-many-arguments\ncommand.extend(options)\ncommand.extend([source, destination])\nreturn gsutil_command(command, parallel=parallel)\n+\n+\n+def cat(file_path, expect_zero=True):\n+ \"\"\"Does gsutil cat on |file_path| and returns the result.\"\"\"\n+ command = ['cat', file_path]\n+ # TODO(metzman): Consider replacing this technique with cp to temp file\n+ # and a local `cat`. The problem with this technique is stderr output\n+ # from gsutil can be included.\n+ return gsutil_command(command, expect_zero=expect_zero)\n" }, { "change_type": "MODIFY", "old_path": "common/local_filestore.py", "new_path": "common/local_filestore.py", "diff": "@@ -23,8 +23,8 @@ def cp( # pylint: disable=invalid-name\nsource,\ndestination,\nrecursive=False,\n- parallel=False, # pylint: disable=unused-argument\n- expect_zero=True):\n+ expect_zero=True,\n+ parallel=False): # pylint: disable=unused-argument\n\"\"\"Executes \"cp\" command from |source| to |destination|.\"\"\"\n# Create intermediate folders for `cp` command to behave like `gsutil.cp`.\nfilesystem.create_directory(os.path.dirname(destination))\n@@ -91,3 +91,9 @@ def rsync( # pylint: disable=too-many-arguments\nsource = source + '/'\ncommand.extend([source, destination])\nreturn new_process.execute(command, expect_zero=True)\n+\n+\n+def cat(file_path, expect_zero=True):\n+ \"\"\"Does cat on |file_path| and returns the result.\"\"\"\n+ command = ['cat', file_path]\n+ return new_process.execute(command, expect_zero=expect_zero)\n" } ]

[ { "change_type": "MODIFY", "old_path": "presubmit.py", "new_path": "presubmit.py", "diff": "@@ -60,9 +60,9 @@ _LICENSE_CHECK_EXTENSIONS = [\n_LICENSE_CHECK_STRING = 'http://www.apache.org/licenses/LICENSE-2.0'\n_SRC_ROOT = Path(__file__).absolute().parent\n+THIRD_PARTY_DIR_NAME = 'third_party'\n_IGNORE_DIRECTORIES = [\nos.path.join(_SRC_ROOT, 'database', 'alembic'),\n- os.path.join(_SRC_ROOT, 'third_party'),\nos.path.join(_SRC_ROOT, 'benchmarks'),\n]\n@@ -306,11 +306,18 @@ def validate_experiment_requests(paths: List[Path]):\nreturn result\n-def is_path_in_ignore_directory(path: Path) -> bool:\n- \"\"\"Returns True if |path| is a subpath of an ignored directory.\"\"\"\n+def is_path_ignored(path: Path) -> bool:\n+ \"\"\"Returns True if |path| is a subpath of an ignored directory or is a\n+ third_party directory.\"\"\"\nfor ignore_directory in _IGNORE_DIRECTORIES:\nif filesystem.is_subpath(ignore_directory, path):\nreturn True\n+\n+ # Third party directories can be anywhere.\n+ path_parts = str(path).split(os.sep)\n+ if any(path_part == THIRD_PARTY_DIR_NAME for path_part in path_parts):\n+ return True\n+\nreturn False\n@@ -327,7 +334,7 @@ def license_check(paths: List[Path]) -> bool:\nextension not in _LICENSE_CHECK_EXTENSIONS):\ncontinue\n- if is_path_in_ignore_directory(path):\n+ if is_path_ignored(path):\ncontinue\nwith open(path) as file_handle:\n@@ -349,7 +356,7 @@ def get_all_files() -> List[Path]:\ndef filter_ignored_files(paths: List[Path]) -> List[Path]:\n\"\"\"Returns a list of absolute paths of files in this repo that can be\nchecked statically.\"\"\"\n- return [path for path in paths if not is_path_in_ignore_directory(path)]\n+ return [path for path in paths if not is_path_ignored(path)]\ndef do_tests() -> bool:\n" } ]

[ { "change_type": "ADD", "old_path": null, "new_path": "common/queue_utils.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Code for setting up a work queue with rq.\"\"\"\n+import redis\n+import rq\n+import rq.job\n+\n+from common import experiment_utils\n+\n+\n+def initialize_queue(redis_host):\n+ \"\"\"Returns a redis-backed rq queue.\"\"\"\n+ queue_name = experiment_utils.get_experiment_name()\n+ redis_connection = redis.Redis(host=redis_host)\n+ queue = rq.Queue(queue_name, connection=redis_connection)\n+ return queue\n+\n+\n+def get_all_jobs(queue):\n+ \"\"\"Returns all the jobs in queue.\"\"\"\n+ job_ids = queue.get_job_ids()\n+ return rq.job.Job.fetch_many(job_ids, queue.connection)\n" } ]

[ { "change_type": "MODIFY", "old_path": "common/gce.py", "new_path": "common/gce.py", "diff": "@@ -54,3 +54,55 @@ def get_preempted_instances(project, zone):\nif (instance['scheduling']['preemptible'] and\ninstance['status'] == 'TERMINATED'):\nyield instance['name']\n+\n+\n+def get_instance_group_managers():\n+ \"\"\"Returns the instance group managers resource.\"\"\"\n+ return thread_local.service.instanceGroupManagers()\n+\n+\n+def get_instance_group_size(instance_group: str, project: str,\n+ zone: str) -> int:\n+ \"\"\"Returns the number of instances running in |instance_group|.\"\"\"\n+ managers = get_instance_group_managers()\n+ request = managers.get(instanceGroupManager=instance_group,\n+ project=project,\n+ zone=zone)\n+ return request.execute()['targetSize']\n+\n+\n+def resize_instance_group(size, instance_group, project, zone):\n+ \"\"\"Changes the number of instances running in |instance_group| to |size|.\"\"\"\n+ assert size >= 1\n+ managers = get_instance_group_managers()\n+ request = managers.resize(instanceGroupManager=instance_group,\n+ size=size,\n+ project=project,\n+ zone=zone)\n+ return request.execute()\n+\n+\n+def delete_instance_group(instance_group, project, zone):\n+ \"\"\"Deletes |instance_group|.\"\"\"\n+ managers = get_instance_group_managers()\n+ request = managers.delete(instanceGroupManager=instance_group,\n+ zone=zone,\n+ project=project)\n+ return request.execute()\n+\n+\n+def create_instance_group(name: str, instance_template_url: str,\n+ base_instance_name: str, project: str, zone: str):\n+ \"\"\"Creates an instance group named |name| from the template specified by\n+ |instance_template_url|.\"\"\"\n+ managers = get_instance_group_managers()\n+ target_size = 1\n+\n+ body = {\n+ 'baseInstanceName': base_instance_name,\n+ 'targetSize': target_size,\n+ 'name': name,\n+ 'instanceTemplate': instance_template_url\n+ }\n+ request = managers.insert(body=body, project=project, zone=zone)\n+ return request.execute()\n" }, { "change_type": "ADD", "old_path": null, "new_path": "common/test_gce.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Tests for gce.py.\"\"\"\n+from unittest import mock\n+\n+from common import gce\n+\n+PROJECT = 'my-cloud-project'\n+ZONE = 'my-compute-zone'\n+INSTANCE_GROUP = 'my-instance-group'\n+INSTANCE_TEMPLATE_URL = 'resource/my-instance-group'\n+EXPERIMENT = 'my-experiment'\n+\n+\[email protected]('common.gce.get_instance_group_managers')\n+def test_delete_instance_group(mocked_get_instance_group_managers):\n+ \"\"\"Tests that delete_instance_group uses the GCE API correctly.\"\"\"\n+ mock_managers = mock.Mock()\n+ mocked_get_instance_group_managers.return_value = mock_managers\n+ gce.delete_instance_group(INSTANCE_GROUP, PROJECT, ZONE)\n+ assert mock_managers.delete.call_args_list == [\n+ mock.call(instanceGroupManager=INSTANCE_GROUP,\n+ project=PROJECT,\n+ zone=ZONE)\n+ ]\n+\n+\[email protected]('common.gce.get_instance_group_managers')\n+def test_resize_instance_group(mocked_get_instance_group_managers):\n+ \"\"\"Tests that resize_instance_group uses the GCE API correctly.\"\"\"\n+ size = 10\n+ mock_managers = mock.Mock()\n+ mocked_get_instance_group_managers.return_value = mock_managers\n+ gce.resize_instance_group(size, INSTANCE_GROUP, PROJECT, ZONE)\n+ assert mock_managers.resize.call_args_list == [\n+ mock.call(instanceGroupManager=INSTANCE_GROUP,\n+ size=size,\n+ project=PROJECT,\n+ zone=ZONE)\n+ ]\n+\n+\[email protected]('common.gce.get_instance_group_managers')\n+def test_create_instance_group(mocked_get_instance_group_managers):\n+ \"\"\"Tests that create_instance_group uses the GCE API correctly.\"\"\"\n+ mock_managers = mock.Mock()\n+ mocked_get_instance_group_managers.return_value = mock_managers\n+ base_instance_name = 'm-' + EXPERIMENT\n+ gce.create_instance_group(INSTANCE_GROUP, INSTANCE_TEMPLATE_URL,\n+ base_instance_name, PROJECT, ZONE)\n+ body = {\n+ 'baseInstanceName': 'm-' + EXPERIMENT,\n+ 'targetSize': 1,\n+ 'name': INSTANCE_GROUP,\n+ 'instanceTemplate': INSTANCE_TEMPLATE_URL,\n+ }\n+ assert mock_managers.insert.call_args_list == [\n+ mock.call(body=body, project=PROJECT, zone=ZONE)\n+ ]\n+\n+\[email protected]('common.gce.get_instance_group_managers')\n+def test_get_instance_group_size(mocked_get_instance_group_managers):\n+ \"\"\"Tests that get_instance_group_size uses the GCE API correctly and returns\n+ the right value.\"\"\"\n+ mock_managers = mock.Mock()\n+ mocked_get_instance_group_managers.return_value = mock_managers\n+ mock_req = mock.Mock()\n+ mock_managers.get.return_value = mock_req\n+ size = 1\n+ mock_req.execute.return_value = {'targetSize': size}\n+ result = gce.get_instance_group_size(INSTANCE_GROUP, PROJECT, ZONE)\n+ assert mock_managers.get.call_args_list == [\n+ mock.call(instanceGroupManager=INSTANCE_GROUP,\n+ project=PROJECT,\n+ zone=ZONE)\n+ ]\n+ assert result == size\n" } ]

[ { "change_type": "MODIFY", "old_path": "docker/generate_makefile.py", "new_path": "docker/generate_makefile.py", "diff": "@@ -91,7 +91,7 @@ def _get_makefile_run_template(image):\ndef get_rules_for_image(name, image):\n\"\"\"Returns makefile section for |image|.\"\"\"\n- if not ('base-' in name or 'dispatcher-' in name):\n+ if not ('base-' in name or 'dispatcher-' in name or name == 'worker'):\nsection = '.'\nelse:\nsection = ''\n" }, { "change_type": "MODIFY", "old_path": "docker/image_types.yaml", "new_path": "docker/image_types.yaml", "diff": "tag: 'dispatcher-image'\ntype: 'dispatcher'\n+'worker':\n+ depends_on:\n+ - 'base-image'\n+ dockerfile: 'docker/worker/Dockerfile'\n+ context: '.'\n+ tag: 'worker'\n+ type: 'worker'\n+\n# TODO: It would be better to call this benchmark builder. But that would be\n# confusing because this doesn't involve benchmark-builder/Dockerfile. Rename\n# that and then rename this.\n" }, { "change_type": "ADD", "old_path": null, "new_path": "docker/worker/Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+FROM gcr.io/fuzzbench/base-image\n+\n+ENV WORK /work\n+ENV SRC $WORK/src\n+RUN mkdir -p $SRC\n+\n+ADD . $SRC/\n+\n+ENTRYPOINT /bin/bash $SRC/docker/worker/startup-worker.sh\n\\ No newline at end of file\n" }, { "change_type": "ADD", "old_path": null, "new_path": "docker/worker/startup-worker.sh", "diff": "+#! /bin/bash\n+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+rq worker $EXPERIMENT --url redis://$REDIS_HOST:6379\n" }, { "change_type": "MODIFY", "old_path": "experiment/build/gcb_build.py", "new_path": "experiment/build/gcb_build.py", "diff": "@@ -46,7 +46,10 @@ def _get_buildable_images(fuzzer=None, benchmark=None):\ndef build_base_images():\n\"\"\"Build base images on GCB.\"\"\"\n- image_templates = {'base-image': _get_buildable_images()['base-image']}\n+ buildable_images = _get_buildable_images()\n+ image_templates = {\n+ image: buildable_images[image] for image in ['base-image', 'worker']\n+ }\nconfig = generate_cloudbuild.create_cloudbuild_spec(image_templates,\nbuild_base_images=True)\n_build(config, 'base-images')\n" }, { "change_type": "MODIFY", "old_path": "experiment/build/local_build.py", "new_path": "experiment/build/local_build.py", "diff": "@@ -35,7 +35,7 @@ def make(targets):\ndef build_base_images() -> Tuple[int, str]:\n\"\"\"Build base images locally.\"\"\"\n- return make(['base-image'])\n+ return make(['base-image', 'worker'])\ndef get_shared_coverage_binaries_dir():\n" }, { "change_type": "MODIFY", "old_path": "experiment/build/test_docker_images.py", "new_path": "experiment/build/test_docker_images.py", "diff": "@@ -23,7 +23,7 @@ def test_images_to_build_list():\nbenchmarks = ['libxml', 'libpng']\nall_images = docker_images.get_images_to_build(fuzzers, benchmarks)\nassert set(all_images.keys()) == set([\n- 'base-image', 'dispatcher-image', 'libxml-project-builder',\n+ 'base-image', 'worker', 'dispatcher-image', 'libxml-project-builder',\n'libpng-project-builder', 'afl-libxml-builder-intermediate',\n'afl-libxml-intermediate-runner', 'afl-libxml-builder',\n'coverage-libxml-builder', 'afl-libpng-builder',\n" } ]

[ { "change_type": "MODIFY", "old_path": "common/gcloud.py", "new_path": "common/gcloud.py", "diff": "\"\"\"Google cloud related code.\"\"\"\nimport enum\n+import posixpath\nimport subprocess\nfrom typing import List\n@@ -112,3 +113,33 @@ def run_local_instance(startup_script: str = None) -> bool:\ncommand = ['/bin/bash', startup_script]\nsubprocess.Popen(command, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)\nreturn new_process.ProcessResult(0, '', False)\n+\n+\n+def create_instance_template(template_name, docker_image, env, project, zone):\n+ \"\"\"Returns a ProcessResult from running the command to create an instance\n+ template.\"\"\"\n+ # Creating an instance template cannot be done using the GCE API because\n+ # there is no public API for handling some docker related functionality that\n+ # we need.\n+ command = [\n+ 'gcloud', 'compute', '--project', project, 'instance-templates',\n+ 'create-with-container', template_name, '--no-address',\n+ '--image-family=cos-stable', '--image-project=cos-cloud',\n+ '--region=%s' % zone, '--scopes=cloud-platform',\n+ '--machine-type=n1-standard-1', '--boot-disk-size=50GB',\n+ '--preemptible', '--container-image', docker_image\n+ ]\n+ for item in env.items():\n+ command.extend(['--container-env', '%s=%s' % item])\n+ new_process.execute(command)\n+ return posixpath.join('https://www.googleapis.com/compute/v1/projects/',\n+ project, 'global', 'instanceTemplates', template_name)\n+\n+\n+def delete_instance_template(template_name: str):\n+ \"\"\"Returns a ProcessResult from running the command to delete the\n+ measure_worker template for this |experiment|.\"\"\"\n+ command = [\n+ 'gcloud', 'compute', 'instance-templates', 'delete', template_name\n+ ]\n+ return new_process.execute(command)\n" }, { "change_type": "MODIFY", "old_path": "common/test_gcloud.py", "new_path": "common/test_gcloud.py", "diff": "@@ -147,3 +147,42 @@ def test_delete_instances_fail(mocked_execute):\nresult = gcloud.delete_instances(instances, zone)\nassert not result\nmocked_execute.assert_called_with(expected_command, expect_zero=False)\n+\n+\[email protected]('common.new_process.execute')\n+def test_create_instance_template(mocked_execute):\n+ \"\"\"Tests that create_instance_template uses the correct gcloud command and\n+ returns the correct instance template URL.\"\"\"\n+ template_name = 'my-template'\n+ docker_image = 'docker_image'\n+ env = {'ENV_VAR': 'value'}\n+ project = 'fuzzbench'\n+ result = gcloud.create_instance_template(template_name, docker_image, env,\n+ project, ZONE)\n+ expected_command = [\n+ 'gcloud', 'compute', '--project', project, 'instance-templates',\n+ 'create-with-container', template_name, '--no-address',\n+ '--image-family=cos-stable', '--image-project=cos-cloud',\n+ '--region=zone-a', '--scopes=cloud-platform',\n+ '--machine-type=n1-standard-1', '--boot-disk-size=50GB',\n+ '--preemptible', '--container-image', docker_image, '--container-env',\n+ 'ENV_VAR=value'\n+ ]\n+ mocked_execute.assert_called_with(expected_command)\n+ expected_result = (\n+ 'https://www.googleapis.com/compute/v1/projects/{project}'\n+ '/global/instanceTemplates/{name}').format(project=project,\n+ name=template_name)\n+ assert result == expected_result\n+\n+\[email protected]('common.new_process.execute')\n+def test_delete_instance_template(mocked_execute):\n+ \"\"\"Tests that delete_instance_template uses the correct gcloud command to\n+ delete an instance template.\"\"\"\n+ template_name = 'my-template'\n+ gcloud.delete_instance_template(template_name)\n+ expected_command = [\n+ 'gcloud', 'compute', 'instance-templates', 'delete', template_name\n+ ]\n+ mocked_execute.assert_called_with(expected_command)\n" } ]

[ { "change_type": "MODIFY", "old_path": ".gitignore", "new_path": ".gitignore", "diff": "@@ -36,3 +36,5 @@ docker/generated.mk\n# Emacs backup files.\n*~\n\\#*\\#\n+\n+.vscode\n\\ No newline at end of file\n" }, { "change_type": "DELETE", "old_path": ".vscode/settings.json", "new_path": null, "diff": "-{\n- \"python.linting.pylintEnabled\": true\n-}\n\\ No newline at end of file\n" } ]

[ { "change_type": "ADD", "old_path": null, "new_path": "experiment/schedule_measure_workers.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Module for starting instances to run measure workers.\"\"\"\n+import collections\n+import os\n+import posixpath\n+import sys\n+import time\n+\n+from common import experiment_utils\n+from common import gce\n+from common import gcloud\n+from common import logs\n+from common import queue_utils\n+from common import yaml_utils\n+\n+logger = logs.Logger('schedule_measure_workers') # pylint: disable=invalid-name\n+\n+# This is the default quota on GCE.\n+# TODO(metzman): Use the GCE API to determine this quota.\n+MAX_INSTANCES_PER_GROUP = 1000\n+\n+\n+def get_instance_group_name(experiment: str):\n+ \"\"\"Returns the name of the instance group of measure workers for\n+ |experiment|.\"\"\"\n+ # \"worker-\" needs to come first because name cannot start with number.\n+ return 'worker-' + experiment\n+\n+\n+def get_measure_worker_instance_template_name(experiment: str):\n+ \"\"\"Returns an instance template name for measurer workers running in\n+ |experiment|.\"\"\"\n+ return 'worker-' + experiment\n+\n+\n+def initialize(experiment_config: dict):\n+ \"\"\"Initialize everything that will be needed to schedule measurers.\"\"\"\n+ logger.info('Initializing worker scheduling.')\n+ gce.initialize()\n+ experiment = experiment_config['experiment']\n+ project = experiment_config['project']\n+ instance_template_name = get_measure_worker_instance_template_name(\n+ experiment)\n+ docker_image = posixpath.join(experiment_config['docker_registry'],\n+ 'measure-worker:{}'.format(experiment))\n+\n+ redis_host = experiment_config['redis_host']\n+ experiment_filestore = experiment_config['experiment_filestore']\n+ local_experiment = experiment_utils.is_local_experiment()\n+ cloud_compute_zone = experiment_config.get('cloud_compute_zone')\n+ env = {\n+ 'REDIS_HOST': redis_host,\n+ 'EXPERIMENT_FILESTORE': experiment_filestore,\n+ 'EXPERIMENT': experiment,\n+ 'LOCAL_EXPERIMENT': local_experiment,\n+ 'CLOUD_COMPUTE_ZONE': cloud_compute_zone,\n+ }\n+\n+ zone = experiment_config['cloud_compute_zone']\n+ instance_template_url = gcloud.create_instance_template(\n+ instance_template_name, docker_image, env, project, zone)\n+\n+ instance_group_name = get_instance_group_name(experiment)\n+\n+ # GCE will create instances for this group in the format\n+ # \"m-$experiment-$UNIQUE_ID\". Use 'm' is short for \"measurer\".\n+ base_instance_name = 'm-' + experiment\n+\n+ gce.create_instance_group(instance_group_name, instance_template_url,\n+ base_instance_name, project, zone)\n+ queue = queue_utils.initialize_queue(redis_host)\n+ return queue\n+\n+\n+def teardown(experiment_config: dict):\n+ \"\"\"Teardown all resources used for running measurer workers.\"\"\"\n+ instance_group_name = get_instance_group_name(\n+ experiment_config['experiment'])\n+ project = experiment_config['cloud_project']\n+ zone = experiment_config['cloud_compute_zone']\n+ gce.delete_instance_group(instance_group_name, project, zone)\n+ gcloud.delete_instance_template(experiment_config['experiment'])\n+\n+\n+def schedule(experiment_config: dict, queue):\n+ \"\"\"Schedule measurer workers. This cannot be called before\n+ initialize_measurers.\"\"\"\n+ logger.info('Scheduling measurer workers.')\n+\n+ # TODO(metzman): This method doesn't seem to correctly take into account\n+ # jobs that are running (the API provided by rq doesn't work intuitively).\n+ # That is OK for now since scheduling only happens while nothing is being\n+ # measured but this should be fixed.\n+ jobs = queue_utils.get_all_jobs(queue)\n+ counts = collections.defaultdict(int)\n+ for job in jobs:\n+ counts[job.get_status(refresh=False)] += 1\n+\n+ num_instances_needed = counts['queued'] + counts['started']\n+ num_instances_needed = min(num_instances_needed, MAX_INSTANCES_PER_GROUP)\n+\n+ logger.info('Scheduling %d workers.', num_instances_needed)\n+ instance_group_name = get_instance_group_name(\n+ experiment_config['experiment'])\n+ project = experiment_config['cloud_project']\n+ zone = experiment_config['cloud_compute_zone']\n+ num_instances = gce.get_instance_group_size(instance_group_name, project,\n+ zone)\n+\n+ # TODO(metzman): Use autoscaling as it probably can deal with quotas more\n+ # easily.\n+ if not num_instances_needed:\n+ # Can't go below 1 instance per group.\n+ logs.info('num_instances_needed = 0, resizing to 1.')\n+ num_instances_needed = 1\n+\n+ if num_instances_needed != num_instances:\n+ # TODO(metzman): Add some limits so always have some measurers but not\n+ # too many.\n+ gce.resize_instance_group(num_instances_needed, instance_group_name,\n+ project, zone)\n+\n+\n+def main():\n+ \"\"\"Run schedule_measure_workers as a standalone script by calling schedule\n+ in a loop. Useful for debugging.\"\"\"\n+ logs.initialize(\n+ default_extras={\n+ 'experiment': os.environ['EXPERIMENT'],\n+ 'component': 'dispatcher',\n+ 'subcomponent': 'scheduler'\n+ })\n+ gce.initialize()\n+ config_path = sys.argv[1]\n+ config = yaml_utils.read(config_path)\n+ queue = initialize(config)\n+ while True:\n+ schedule(config, queue)\n+ time.sleep(30)\n+\n+\n+if __name__ == '__main__':\n+ main()\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/build_and_test_run_fuzzer_benchmarks.py", "new_path": ".github/workflows/build_and_test_run_fuzzer_benchmarks.py", "diff": "@@ -25,26 +25,19 @@ ALWAYS_BUILD_FUZZER = 'afl'\nNUM_RETRIES = 2\nRETRY_DELAY = 60\n-# TODO(tanq16): Get list of Benchmarks automatically.\n-\n# Don't build php benchmark since it fills up disk in GH actions.\nOSS_FUZZ_BENCHMARKS = {\n'bloaty_fuzz_target',\n'curl_curl_fuzzer_http',\n'harfbuzz_hb-subset-fuzzer',\n'jsoncpp_jsoncpp_fuzzer',\n- 'libhevc_hevc_dec_fuzzer',\n'libpcap_fuzz_both',\n'libxslt_xpath',\n- 'matio_matio_fuzzer',\n'mbedtls_fuzz_dtlsclient',\n- 'ndpi_fuzz_ndpi_reader',\n'openh264_decoder_fuzzer',\n'openssl_x509',\n'sqlite3_ossfuzz',\n- 'stb_stbi_read_fuzzer',\n'systemd_fuzz-link-parser',\n- 'wabt_wasm2wat_fuzzer',\n'zlib_zlib_uncompress_fuzzer',\n}\n@@ -62,6 +55,14 @@ STANDARD_BENCHMARKS = {\n'woff2-2016-05-06',\n}\n+BUG_BENCHMARKS = {\n+ 'libhevc_hevc_dec_fuzzer',\n+ 'matio_matio_fuzzer',\n+ 'ndpi_fuzz_ndpi_reader',\n+ 'stb_stbi_read_fuzzer',\n+ 'wabt_wasm2wat_fuzzer',\n+}\n+\ndef get_make_target(fuzzer, benchmark):\n\"\"\"Return test target for a fuzzer and benchmark.\"\"\"\n@@ -121,6 +122,8 @@ def do_build(build_type, fuzzer, always_build):\nbenchmarks = OSS_FUZZ_BENCHMARKS\nelif build_type == 'standard':\nbenchmarks = STANDARD_BENCHMARKS\n+ elif build_type == 'bug':\n+ benchmarks = BUG_BENCHMARKS\nelse:\nraise Exception('Invalid build_type: %s' % build_type)\n" }, { "change_type": "MODIFY", "old_path": ".github/workflows/fuzzers.yml", "new_path": ".github/workflows/fuzzers.yml", "diff": "@@ -5,7 +5,10 @@ on:\n- 'docker/**' # Base image changes.\n- 'fuzzers/**' # Changes to fuzzers themselves.\n- 'benchmarks/**' # Changes to benchmarks.\n- - 'src_analysis/**' # Changes that affect what gets built.\n+ # Changes that affect what gets built.\n+ - 'src_analysis/**'\n+ - '.github/worfkflows/fuzzers.yml'\n+ - '.github/worfkflows/build_and_test_run_fuzzer_benchmarks.py'\njobs:\nbuild:\n@@ -56,6 +59,7 @@ jobs:\nbenchmark_type:\n- oss-fuzz\n- standard\n+ - bug\nsteps:\n- uses: actions/checkout@v2\n" } ]

[ { "change_type": "MODIFY", "old_path": "experiment/schedule_measure_workers.py", "new_path": "experiment/schedule_measure_workers.py", "diff": "@@ -45,6 +45,12 @@ def get_measure_worker_instance_template_name(experiment: str):\nreturn 'worker-' + experiment\n+def get_base_worker_instance_name(experiment):\n+ \"\"\"GCE will create instances for this group in the format\n+ \"w-|experiment|-$UNIQUE_ID\". 'w' is short for \"worker\".\"\"\"\n+ return 'w-' + experiment\n+\n+\ndef initialize(experiment_config: dict):\n\"\"\"Initialize everything that will be needed to schedule measurers.\"\"\"\nlogger.info('Initializing worker scheduling.')\n@@ -74,9 +80,7 @@ def initialize(experiment_config: dict):\ninstance_group_name = get_instance_group_name(experiment)\n- # GCE will create instances for this group in the format\n- # \"m-$experiment-$UNIQUE_ID\". Use 'm' is short for \"measurer\".\n- base_instance_name = 'm-' + experiment\n+ base_instance_name = get_base_worker_instance_name(experiment)\ngce.create_instance_group(instance_group_name, instance_template_url,\nbase_instance_name, project, zone)\n" } ]

[ { "change_type": "MODIFY", "old_path": "benchmarks/harfbuzz_hb-subset-fuzzer/Dockerfile", "new_path": "benchmarks/harfbuzz_hb-subset-fuzzer/Dockerfile", "diff": "#\n################################################################################\n-FROM gcr.io/oss-fuzz-base/base-builder\n+FROM gcr.io/oss-fuzz-base/base-builder@sha256:7c6edc92d725ec4117ed7f380fb2fa51c2382a037c405dd3fe1ee2870deeb1a1\nRUN apt-get update && apt-get install -y python3-pip ragel pkg-config && \\\npip3 install meson==0.53.0 ninja\nRUN git clone --depth 1 https://github.com/harfbuzz/harfbuzz.git\n" }, { "change_type": "MODIFY", "old_path": "benchmarks/libhevc_hevc_dec_fuzzer/Dockerfile", "new_path": "benchmarks/libhevc_hevc_dec_fuzzer/Dockerfile", "diff": "#\n################################################################################\n-FROM gcr.io/oss-fuzz-base/base-builder\n+FROM gcr.io/oss-fuzz-base/base-builder@sha256:270e52ac2c38de7d7d678ab488252c81652a161b378fc9fc9f3df6c79e475afa\nMAINTAINER [email protected]\nRUN apt-get update && apt-get install -y wget cmake\nRUN git clone https://android.googlesource.com/platform/external/libhevc\n" }, { "change_type": "MODIFY", "old_path": "benchmarks/ndpi_fuzz_ndpi_reader/Dockerfile", "new_path": "benchmarks/ndpi_fuzz_ndpi_reader/Dockerfile", "diff": "#\n################################################################################\n-FROM gcr.io/oss-fuzz-base/base-builder\n+FROM gcr.io/oss-fuzz-base/base-builder@sha256:83cf1d8893892d6bc471008e50c5fd078a3b8efe80c421b56334e0a28a60fa53\nMAINTAINER [email protected]\nRUN apt-get update && apt-get install -y make autoconf automake autogen pkg-config libtool flex bison\nRUN git clone --depth 1 https://github.com/ntop/nDPI.git ndpi\n" }, { "change_type": "MODIFY", "old_path": "benchmarks/openh264_decoder_fuzzer/Dockerfile", "new_path": "benchmarks/openh264_decoder_fuzzer/Dockerfile", "diff": "#\n################################################################################\n-FROM gcr.io/oss-fuzz-base/base-builder\n+FROM gcr.io/oss-fuzz-base/base-builder@sha256:d321113ffaca3cd8dcec0ceff2c594c01b6a860971d02058da7d42c0c82e3d23\nMAINTAINER [email protected]\nRUN dpkg --add-architecture i386 && \\\napt-get update && \\\n" }, { "change_type": "MODIFY", "old_path": "benchmarks/stb_stbi_read_fuzzer/Dockerfile", "new_path": "benchmarks/stb_stbi_read_fuzzer/Dockerfile", "diff": "#\n################################################################################\n-FROM gcr.io/oss-fuzz-base/base-builder\n+FROM gcr.io/oss-fuzz-base/base-builder@sha256:12b7b470479f04fd1b6a124291486f555169a5496a355ee75cf333c117d4bb92\nRUN apt-get update && \\\napt-get install -y wget tar\n" }, { "change_type": "MODIFY", "old_path": "benchmarks/wabt_wasm2wat_fuzzer/Dockerfile", "new_path": "benchmarks/wabt_wasm2wat_fuzzer/Dockerfile", "diff": "#\n################################################################################\n-FROM gcr.io/oss-fuzz-base/base-builder\n+FROM gcr.io/oss-fuzz-base/base-builder@sha256:c0466abc5bd8515aaceb3fa06725c982522c1127d3559da8615ad22fc43c78a7\nMAINTAINER [email protected]\nRUN apt-get update && apt-get install -y cmake libtool make python\nRUN git clone --recursive https://github.com/WebAssembly/wabt\n" } ]

[ { "change_type": "MODIFY", "old_path": "experiment/dispatcher.py", "new_path": "experiment/dispatcher.py", "diff": "@@ -31,7 +31,7 @@ from common import yaml_utils\nfrom database import models\nfrom database import utils as db_utils\nfrom experiment.build import builder\n-from experiment.measurer import measurer\n+from experiment.measurer import measure_manager\nfrom experiment import reporter\nfrom experiment import scheduler\nfrom experiment import stop_experiment\n@@ -156,7 +156,7 @@ def dispatcher_main():\nscheduler_loop_thread.start()\nmeasurer_main_process = multiprocessing.Process(\n- target=measurer.measure_main, args=(experiment.config,))\n+ target=measure_manager.measure_main, args=(experiment.config,))\nmeasurer_main_process.start()\n" }, { "change_type": "RENAME", "old_path": "experiment/measurer/measurer.py", "new_path": "experiment/measurer/measure_manager.py", "diff": "-#!/usr/bin/env python3\n# Copyright 2020 Google LLC\n#\n# Licensed under the Apache License, Version 2.0 (the \"License\");\n" }, { "change_type": "RENAME", "old_path": "experiment/measurer/test_measurer.py", "new_path": "experiment/measurer/test_measure_manager.py", "diff": "# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n# See the License for the specific language governing permissions and\n# limitations under the License.\n-\"\"\"Tests for measurer.py.\"\"\"\n+\"\"\"Tests for measure_manager.py.\"\"\"\nimport os\nimport shutil\n@@ -25,7 +25,7 @@ from common import new_process\nfrom database import models\nfrom database import utils as db_utils\nfrom experiment.build import build_utils\n-from experiment.measurer import measurer\n+from experiment.measurer import measure_manager\nfrom test_libs import utils as test_utils\nTEST_DATA_PATH = os.path.join(os.path.dirname(__file__), 'test_data')\n@@ -41,7 +41,7 @@ MAX_TOTAL_TIME = 100\nGIT_HASH = 'FAKE-GIT-HASH'\nCYCLE = 1\n-SNAPSHOT_LOGGER = measurer.logger\n+SNAPSHOT_LOGGER = measure_manager.logger\n# pylint: disable=unused-argument,invalid-name,redefined-outer-name,protected-access\n@@ -58,8 +58,8 @@ def db_experiment(experiment_config, db):\ndef test_get_current_coverage(fs, experiment):\n\"\"\"Tests that get_current_coverage reads the correct data from json file.\"\"\"\n- snapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\n- SNAPSHOT_LOGGER)\n+ snapshot_measurer = measure_manager.SnapshotMeasurer(\n+ FUZZER, BENCHMARK, TRIAL_NUM, SNAPSHOT_LOGGER)\njson_cov_summary_file = get_test_data_path('cov_summary.json')\nfs.add_real_file(json_cov_summary_file, read_only=False)\nsnapshot_measurer.cov_summary_file = json_cov_summary_file\n@@ -70,8 +70,8 @@ def test_get_current_coverage(fs, experiment):\ndef test_get_current_coverage_error(fs, experiment):\n\"\"\"Tests that get_current_coverage returns None from a\ndefective json file.\"\"\"\n- snapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\n- SNAPSHOT_LOGGER)\n+ snapshot_measurer = measure_manager.SnapshotMeasurer(\n+ FUZZER, BENCHMARK, TRIAL_NUM, SNAPSHOT_LOGGER)\njson_cov_summary_file = get_test_data_path('cov_summary_defective.json')\nfs.add_real_file(json_cov_summary_file, read_only=False)\nsnapshot_measurer.cov_summary_file = json_cov_summary_file\n@@ -81,8 +81,8 @@ def test_get_current_coverage_error(fs, experiment):\ndef test_get_current_coverage_no_file(fs, experiment):\n\"\"\"Tests that get_current_coverage returns None with no json file.\"\"\"\n- snapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\n- SNAPSHOT_LOGGER)\n+ snapshot_measurer = measure_manager.SnapshotMeasurer(\n+ FUZZER, BENCHMARK, TRIAL_NUM, SNAPSHOT_LOGGER)\njson_cov_summary_file = get_test_data_path('cov_summary_not_exist.json')\nsnapshot_measurer.cov_summary_file = json_cov_summary_file\ncovered_regions = snapshot_measurer.get_current_coverage()\n@@ -93,8 +93,8 @@ def test_get_current_coverage_no_file(fs, experiment):\ndef test_generate_profdata_create(mocked_execute, experiment, fs):\n\"\"\"Tests that generate_profdata can run the correct command.\"\"\"\nmocked_execute.return_value = new_process.ProcessResult(0, '', False)\n- snapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\n- SNAPSHOT_LOGGER)\n+ snapshot_measurer = measure_manager.SnapshotMeasurer(\n+ FUZZER, BENCHMARK, TRIAL_NUM, SNAPSHOT_LOGGER)\nsnapshot_measurer.profdata_file = '/work/reports/data.profdata'\nsnapshot_measurer.profraw_file_pattern = '/work/reports/data-%m.profraw'\nprofraw_file = '/work/reports/data-123.profraw'\n@@ -115,8 +115,8 @@ def test_generate_profdata_create(mocked_execute, experiment, fs):\ndef test_generate_profdata_merge(mocked_execute, experiment, fs):\n\"\"\"Tests that generate_profdata can run correctly with existing profraw.\"\"\"\nmocked_execute.return_value = new_process.ProcessResult(0, '', False)\n- snapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\n- SNAPSHOT_LOGGER)\n+ snapshot_measurer = measure_manager.SnapshotMeasurer(\n+ FUZZER, BENCHMARK, TRIAL_NUM, SNAPSHOT_LOGGER)\nsnapshot_measurer.profdata_file = '/work/reports/data.profdata'\nsnapshot_measurer.profraw_file_pattern = '/work/reports/data-%m.profraw'\nprofraw_file = '/work/reports/data-123.profraw'\n@@ -143,8 +143,8 @@ def test_generate_summary(mocked_get_coverage_binary, mocked_execute,\ncoverage_binary_path = '/work/coverage-binaries/benchmark-a/fuzz-target'\nmocked_get_coverage_binary.return_value = coverage_binary_path\n- snapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\n- SNAPSHOT_LOGGER)\n+ snapshot_measurer = measure_manager.SnapshotMeasurer(\n+ FUZZER, BENCHMARK, TRIAL_NUM, SNAPSHOT_LOGGER)\nsnapshot_measurer.cov_summary_file = \"/reports/cov_summary.txt\"\nsnapshot_measurer.profdata_file = \"/reports/data.profdata\"\nfs.create_dir('/reports')\n@@ -164,17 +164,18 @@ def test_generate_summary(mocked_get_coverage_binary, mocked_execute,\[email protected]('common.logs.error')\[email protected]('experiment.measurer.measurer.initialize_logs')\[email protected]('experiment.measurer.measure_manager.initialize_logs')\[email protected]('multiprocessing.Queue')\[email protected]('experiment.measurer.measurer.measure_snapshot_coverage')\[email protected]('experiment.measurer.measure_manager.measure_snapshot_coverage')\ndef test_measure_trial_coverage(mocked_measure_snapshot_coverage, mocked_queue,\n_, __):\n\"\"\"Tests that measure_trial_coverage works as expected.\"\"\"\nmin_cycle = 1\nmax_cycle = 10\n- measure_request = measurer.SnapshotMeasureRequest(FUZZER, BENCHMARK,\n- TRIAL_NUM, min_cycle)\n- measurer.measure_trial_coverage(measure_request, max_cycle, mocked_queue())\n+ measure_request = measure_manager.SnapshotMeasureRequest(\n+ FUZZER, BENCHMARK, TRIAL_NUM, min_cycle)\n+ measure_manager.measure_trial_coverage(measure_request, max_cycle,\n+ mocked_queue())\nexpected_calls = [\nmock.call(FUZZER, BENCHMARK, TRIAL_NUM, cycle)\nfor cycle in range(min_cycle, max_cycle + 1)\n@@ -187,9 +188,9 @@ def test_measure_trial_coverage(mocked_measure_snapshot_coverage, mocked_queue,\ndef test_measure_all_trials_not_ready(mocked_rsync, mocked_ls, experiment):\n\"\"\"Test running measure_all_trials before it is ready works as intended.\"\"\"\nmocked_ls.return_value = new_process.ProcessResult(1, '', False)\n- assert measurer.measure_all_trials(experiment_utils.get_experiment_name(),\n- MAX_TOTAL_TIME, test_utils.MockPool(),\n- queue.Queue())\n+ assert measure_manager.measure_all_trials(\n+ experiment_utils.get_experiment_name(), MAX_TOTAL_TIME,\n+ test_utils.MockPool(), queue.Queue())\nassert not mocked_rsync.called\n@@ -204,7 +205,7 @@ def test_measure_all_trials_no_more(mocked_directories_have_same_files,\nmocked_directories_have_same_files.return_value = True\nmocked_execute.return_value = new_process.ProcessResult(0, '', False)\nmock_pool = test_utils.MockPool()\n- assert not measurer.measure_all_trials(\n+ assert not measure_manager.measure_all_trials(\nexperiment_utils.get_experiment_name(), MAX_TOTAL_TIME, mock_pool,\nqueue.Queue())\n@@ -212,8 +213,8 @@ def test_measure_all_trials_no_more(mocked_directories_have_same_files,\ndef test_is_cycle_unchanged_doesnt_exist(experiment):\n\"\"\"Test that is_cycle_unchanged can properly determine if a cycle is\nunchanged or not when it needs to copy the file for the first time.\"\"\"\n- snapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\n- SNAPSHOT_LOGGER)\n+ snapshot_measurer = measure_manager.SnapshotMeasurer(\n+ FUZZER, BENCHMARK, TRIAL_NUM, SNAPSHOT_LOGGER)\nthis_cycle = 1\nwith test_utils.mock_popen_ctx_mgr(returncode=1):\nassert not snapshot_measurer.is_cycle_unchanged(this_cycle)\n@@ -224,8 +225,8 @@ def test_is_cycle_unchanged_doesnt_exist(experiment):\ndef test_is_cycle_unchanged_first_copy(mocked_read, mocked_cp, experiment):\n\"\"\"Test that is_cycle_unchanged can properly determine if a cycle is\nunchanged or not when it needs to copy the file for the first time.\"\"\"\n- snapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\n- SNAPSHOT_LOGGER)\n+ snapshot_measurer = measure_manager.SnapshotMeasurer(\n+ FUZZER, BENCHMARK, TRIAL_NUM, SNAPSHOT_LOGGER)\nthis_cycle = 100\nunchanged_cycles_file_contents = (\n'\\n'.join([str(num) for num in range(10)] + [str(this_cycle)]))\n@@ -239,8 +240,8 @@ def test_is_cycle_unchanged_first_copy(mocked_read, mocked_cp, experiment):\ndef test_is_cycle_unchanged_update(fs, experiment):\n\"\"\"Test that is_cycle_unchanged can properly determine that a\ncycle has changed when it has the file but needs to update it.\"\"\"\n- snapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\n- SNAPSHOT_LOGGER)\n+ snapshot_measurer = measure_manager.SnapshotMeasurer(\n+ FUZZER, BENCHMARK, TRIAL_NUM, SNAPSHOT_LOGGER)\nthis_cycle = 100\ninitial_unchanged_cycles_file_contents = (\n@@ -263,8 +264,8 @@ def test_is_cycle_unchanged_update(fs, experiment):\ndef test_is_cycle_unchanged_skip_cp(mocked_cp, fs, experiment):\n\"\"\"Check that is_cycle_unchanged doesn't call filestore_utils.cp\nunnecessarily.\"\"\"\n- snapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\n- SNAPSHOT_LOGGER)\n+ snapshot_measurer = measure_manager.SnapshotMeasurer(\n+ FUZZER, BENCHMARK, TRIAL_NUM, SNAPSHOT_LOGGER)\nthis_cycle = 100\ninitial_unchanged_cycles_file_contents = (\n'\\n'.join([str(num) for num in range(10)] + [str(this_cycle + 1)]))\n@@ -279,8 +280,8 @@ def test_is_cycle_unchanged_no_file(mocked_cp, fs, experiment):\n\"\"\"Test that is_cycle_unchanged returns False when there is no\nunchanged-cycles file.\"\"\"\n# Make sure we log if there is no unchanged-cycles file.\n- snapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\n- SNAPSHOT_LOGGER)\n+ snapshot_measurer = measure_manager.SnapshotMeasurer(\n+ FUZZER, BENCHMARK, TRIAL_NUM, SNAPSHOT_LOGGER)\nmocked_cp.return_value = new_process.ProcessResult(1, '', False)\nassert not snapshot_measurer.is_cycle_unchanged(0)\n@@ -296,8 +297,8 @@ def test_run_cov_new_units(_, mocked_execute, fs, environ):\n'EXPERIMENT': 'experiment',\n}\nmocked_execute.return_value = new_process.ProcessResult(0, '', False)\n- snapshot_measurer = measurer.SnapshotMeasurer(FUZZER, BENCHMARK, TRIAL_NUM,\n- SNAPSHOT_LOGGER)\n+ snapshot_measurer = measure_manager.SnapshotMeasurer(\n+ FUZZER, BENCHMARK, TRIAL_NUM, SNAPSHOT_LOGGER)\nsnapshot_measurer.initialize_measurement_dirs()\nshared_units = ['shared1', 'shared2']\nfs.create_file(snapshot_measurer.measured_files_path,\n@@ -351,8 +352,8 @@ class TestIntegrationMeasurement:\n# portable binary.\[email protected](not os.getenv('FUZZBENCH_TEST_INTEGRATION'),\nreason='Not running integration tests.')\n- @mock.patch(\n- 'experiment.measurer.measurer.SnapshotMeasurer.is_cycle_unchanged')\n+ @mock.patch('experiment.measurer.measure_manager.SnapshotMeasurer'\n+ '.is_cycle_unchanged')\ndef test_measure_snapshot_coverage( # pylint: disable=too-many-locals\nself, mocked_is_cycle_unchanged, db, experiment, tmp_path):\n\"\"\"Integration test for measure_snapshot_coverage.\"\"\"\n@@ -384,9 +385,8 @@ class TestIntegrationMeasurement:\nexperiment=os.environ['EXPERIMENT'])\ndb_utils.add_all([trial])\n- snapshot_measurer = measurer.SnapshotMeasurer(trial.fuzzer,\n- trial.benchmark, trial.id,\n- SNAPSHOT_LOGGER)\n+ snapshot_measurer = measure_manager.SnapshotMeasurer(\n+ trial.fuzzer, trial.benchmark, trial.id, SNAPSHOT_LOGGER)\n# Set up the snapshot archive.\ncycle = 1\n@@ -400,7 +400,7 @@ class TestIntegrationMeasurement:\nmocked_cp.return_value = new_process.ProcessResult(0, '', False)\n# TODO(metzman): Create a system for using actual buckets in\n# integration tests.\n- snapshot = measurer.measure_snapshot_coverage(\n+ snapshot = measure_manager.measure_snapshot_coverage(\nsnapshot_measurer.fuzzer, snapshot_measurer.benchmark,\nsnapshot_measurer.trial_num, cycle)\nassert snapshot\n@@ -413,7 +413,7 @@ class TestIntegrationMeasurement:\ndef test_extract_corpus(archive_name, tmp_path):\n\"\"\"\"Tests that extract_corpus unpacks a corpus as we expect.\"\"\"\narchive_path = get_test_data_path(archive_name)\n- measurer.extract_corpus(archive_path, set(), tmp_path)\n+ measure_manager.extract_corpus(archive_path, set(), tmp_path)\nexpected_corpus_files = {\n'5ea57dfc9631f35beecb5016c4f1366eb6faa810',\n'2f1507c3229c5a1f8b619a542a8e03ccdbb3c29c',\n@@ -423,8 +423,8 @@ def test_extract_corpus(archive_name, tmp_path):\[email protected]('time.sleep', return_value=None)\[email protected]('experiment.measurer.measurer.set_up_coverage_binaries')\[email protected]('experiment.measurer.measurer.measure_all_trials',\[email protected]('experiment.measurer.measure_manager.set_up_coverage_binaries')\[email protected]('experiment.measurer.measure_manager.measure_all_trials',\nreturn_value=False)\[email protected]('multiprocessing.Manager')\[email protected]('multiprocessing.pool')\n@@ -433,17 +433,17 @@ def test_measure_loop_end(_, __, ___, ____, _____, ______, experiment_config,\ndb_experiment):\n\"\"\"Tests that measure_loop stops when there is nothing left to measure. In\nthis test, there is nothing left to measure on the first call.\"\"\"\n- measurer.measure_loop(experiment_config, 100)\n+ measure_manager.measure_loop(experiment_config, 100)\n# If everything went well, we should get to this point without any\n# exceptions.\[email protected]('time.sleep', return_value=None)\[email protected]('experiment.measurer.measurer.set_up_coverage_binaries')\[email protected]('experiment.measurer.measure_manager.set_up_coverage_binaries')\[email protected]('multiprocessing.Manager')\[email protected]('multiprocessing.pool')\[email protected]('experiment.scheduler.all_trials_ended', return_value=True)\[email protected]('experiment.measurer.measurer.measure_all_trials')\[email protected]('experiment.measurer.measure_manager.measure_all_trials')\ndef test_measure_loop_loop_until_end(mocked_measure_all_trials, _, __, ___,\n____, _____, experiment_config,\ndb_experiment):\n@@ -464,7 +464,7 @@ def test_measure_loop_loop_until_end(mocked_measure_all_trials, _, __, ___,\nreturn True\nmocked_measure_all_trials.side_effect = mock_measure_all_trials\n- measurer.measure_loop(experiment_config, 100)\n+ measure_manager.measure_loop(experiment_config, 100)\nassert call_count == loop_iterations\n@@ -475,7 +475,7 @@ def test_path_exists_in_experiment_filestore(mocked_execute, environ):\nos.environ['WORK'] = work_dir\nos.environ['EXPERIMENT_FILESTORE'] = 'gs://cloud-bucket'\nos.environ['EXPERIMENT'] = 'example-experiment'\n- measurer.exists_in_experiment_filestore(work_dir)\n+ measure_manager.exists_in_experiment_filestore(work_dir)\nmocked_execute.assert_called_with(\n['gsutil', 'ls', 'gs://cloud-bucket/example-experiment'],\nexpect_zero=False)\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/fuzzers.yml", "new_path": ".github/workflows/fuzzers.yml", "diff": "@@ -55,6 +55,7 @@ jobs:\n- aflplusplus_fast_branches_v2\n- aflplusplus_cmplog\n- aflplusplus_dict2file\n+ - aflplusplus_268c_qemu\n- aflplusplus_fast_v2_add\n- aflplusplus_fast_v2_depth\n- aflplusplus_fast_v2_cutoff\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/aflplusplus_268c_qemu/builder.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+ARG parent_image\n+FROM $parent_image\n+\n+# Install wget to download afl_driver.cpp. Install libstdc++ to use llvm_mode.\n+RUN apt-get update && \\\n+ apt-get install -y wget libstdc++-5-dev libtool-bin automake flex bison \\\n+ libglib2.0-dev libpixman-1-dev python3-setuptools unzip\n+\n+# Why do some build images have ninja, other not? Weird.\n+RUN cd / && wget https://github.com/ninja-build/ninja/releases/download/v1.10.1/ninja-linux.zip && \\\n+ unzip ninja-linux.zip && chmod 755 ninja && mv ninja /usr/local/bin\n+\n+# Build afl++ without Python support as we don't need it.\n+# Set AFL_NO_X86 to skip flaky tests.\n+RUN git clone https://github.com/AFLplusplus/AFLplusplus.git /afl && \\\n+ cd /afl && git checkout ee206da3897fd2d9f72206c3c5ea0e3fab109001 && \\\n+ unset CFLAGS && unset CXXFLAGS && \\\n+ AFL_NO_X86=1 CC=clang PYTHON_INCLUDE=/ make && \\\n+ cd qemu_mode && ./build_qemu_support.sh && cd .. && \\\n+ make -C examples/aflpp_driver && \\\n+ cp examples/aflpp_driver/libAFLQemuDriver.a /libAFLDriver.a && \\\n+ cp examples/aflpp_driver/aflpp_qemu_driver_hook.so /\n+\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/aflplusplus_268c_qemu/description.md", "diff": "+# aflplusplus_qemu\n+\n+AFL++ fuzzer instance for binary-only fuzzing with qemu_mode.\n+The following config active for all benchmarks:\n+ - qemu_mode with:\n+ - laf-intel (integers and floats)\n+ - entrypoint set to LLVMFuzzerTestOneInput\n+ - persisten mode set to LLVMFuzzerTestOneInput\n+ - in-memory shared memory test cases\n+ - \"coe\" power schedule\n+\n+Repository: [https://github.com/AFLplusplus/AFLplusplus/](https://github.com/AFLplusplus/AFLplusplus/)\n+\n+[builder.Dockerfile](builder.Dockerfile)\n+[fuzzer.py](fuzzer.py)\n+[runner.Dockerfile](runner.Dockerfile)\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/aflplusplus_268c_qemu/fuzzer.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Integration code for AFLplusplus fuzzer.\"\"\"\n+\n+import os\n+import subprocess\n+\n+from fuzzers.aflplusplus import fuzzer as aflplusplus_fuzzer\n+\n+\n+def build():\n+ \"\"\"Build benchmark.\"\"\"\n+ aflplusplus_fuzzer.build('qemu')\n+\n+\n+def fuzz(input_corpus, output_corpus, target_binary):\n+ \"\"\"Run fuzzer.\"\"\"\n+ # Get LLVMFuzzerTestOneInput address.\n+ nm_proc = subprocess.run([\n+ 'sh', '-c',\n+ 'nm \\'' + target_binary + '\\' | grep \\'T afl_qemu_driver_stdin_input\\''\n+ ],\n+ stdout=subprocess.PIPE,\n+ check=True)\n+ target_func = \"0x\" + nm_proc.stdout.split()[0].decode(\"utf-8\")\n+ print('[fuzz] afl_qemu_driver_stdin_input() address =', target_func)\n+\n+ # Fuzzer option for qemu_mode.\n+ flags = ['-Q']\n+\n+ os.environ['AFL_COMPCOV_LEVEL'] = '3' # Complete compcov including floats\n+ os.environ['AFL_QEMU_PERSISTENT_ADDR'] = target_func\n+ os.environ['AFL_ENTRYPOINT'] = target_func\n+ os.environ['AFL_QEMU_PERSISTENT_CNT'] = \"100000\"\n+ os.environ['AFL_QEMU_DRIVER_NO_HOOK'] = \"1\"\n+ aflplusplus_fuzzer.fuzz(input_corpus,\n+ output_corpus,\n+ target_binary,\n+ flags=flags)\n" }, { "change_type": "ADD", "old_path": null, "new_path": "fuzzers/aflplusplus_268c_qemu/runner.Dockerfile", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\n+FROM gcr.io/fuzzbench/base-image\n+\n+# This makes interactive docker runs painless:\n+ENV LD_LIBRARY_PATH=\"$LD_LIBRARY_PATH:/out\"\n+ENV AFL_MAP_SIZE=1048576\n+ENV PATH=\"$PATH:/out\"\n+ENV AFL_SKIP_CPUFREQ=1\n+ENV AFL_I_DONT_CARE_ABOUT_MISSING_CRASHES=1\n" }, { "change_type": "MODIFY", "old_path": "service/experiment-requests.yaml", "new_path": "service/experiment-requests.yaml", "diff": "#\n# You can run \"make presubmit\" to do basic validation on this file.\n# Please add new experiment requests towards the top of this file.\n+- experiment: 2020-12-04\n+ description: \"AFL++ 2.68c qemu mode\"\n+ fuzzers:\n+ - aflplusplus_268c_qemu\n+ - aflplusplus_qemu\n- experiment: 2020-12-03\ndescription: \"AFL++ some fast variants\"\n" } ]

[ { "change_type": "MODIFY", "old_path": ".github/workflows/fuzzers.yml", "new_path": ".github/workflows/fuzzers.yml", "diff": "@@ -17,7 +17,6 @@ jobs:\nfail-fast: false\nmatrix:\nfuzzer:\n- - coverage\n- afl\n- aflcc\n- aflfast\n" }, { "change_type": "MODIFY", "old_path": "presubmit.py", "new_path": "presubmit.py", "diff": "@@ -125,7 +125,7 @@ class FuzzerAndBenchmarkValidator:\n# We know this is invalid and have already complained about it.\nreturn False\n- if not is_fuzzer_tested_in_ci(fuzzer):\n+ if fuzzer != 'coverage' and not is_fuzzer_tested_in_ci(fuzzer):\nself.invalid_fuzzers.add(fuzzer)\nreturn False\n" } ]

[ { "change_type": "MODIFY", "old_path": "experiment/build/gcb_build.py", "new_path": "experiment/build/gcb_build.py", "diff": "# limitations under the License.\n\"\"\"Module for building things on Google Cloud Build for use in trials.\"\"\"\n+import subprocess\nimport tempfile\nfrom typing import Dict\n@@ -100,8 +101,12 @@ def _build(\nresult = new_process.execute(command,\nwrite_to_stdout=False,\nkill_children=True,\n- timeout=timeout_seconds)\n+ timeout=timeout_seconds,\n+ expect_zero=False)\n+ # TODO(metzman): Refactor code so that local_build stores logs as well.\nbuild_utils.store_build_logs(config_name, result)\n+ if result.retcode != 0:\n+ raise subprocess.CalledProcessError(result.retcode, command)\nreturn result\n" }, { "change_type": "ADD", "old_path": null, "new_path": "experiment/build/test_gcb_build.py", "diff": "+# Copyright 2020 Google LLC\n+#\n+# Licensed under the Apache License, Version 2.0 (the \"License\");\n+# you may not use this file except in compliance with the License.\n+# You may obtain a copy of the License at\n+#\n+# http://www.apache.org/licenses/LICENSE-2.0\n+#\n+# Unless required by applicable law or agreed to in writing, software\n+# distributed under the License is distributed on an \"AS IS\" BASIS,\n+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n+# See the License for the specific language governing permissions and\n+# limitations under the License.\n+\"\"\"Tests for builder.py.\"\"\"\n+\n+import subprocess\n+from unittest import mock\n+\n+import pytest\n+\n+from common import new_process\n+from experiment.build import gcb_build\n+\n+# pylint: disable=protected-access\n+\n+FAIL_RESULT = new_process.ProcessResult(1, '', False)\n+\n+\[email protected]('common.new_process.execute', return_value=FAIL_RESULT)\[email protected]('experiment.build.build_utils.store_build_logs')\n+def test_build_error(mocked_store_build_logs, _):\n+ \"\"\"Tests that on error, _build raises subprocess.CalledProcessError and\n+ calls store_build_logs.\"\"\"\n+ config_name = 'config'\n+ with pytest.raises(subprocess.CalledProcessError):\n+ gcb_build._build({}, config_name)\n+ mocked_store_build_logs.assert_called_with(config_name, FAIL_RESULT)\n+\n+\n+SUCCESS_RESULT = new_process.ProcessResult(0, '', False)\n+\n+\[email protected]('common.new_process.execute', return_value=SUCCESS_RESULT)\[email protected]('experiment.build.build_utils.store_build_logs')\n+def test_build_success_store_logs(mocked_store_build_logs, _):\n+ \"\"\"Tests that on success _buiild stores build logs.\"\"\"\n+ config_name = 'config'\n+ gcb_build._build({}, config_name)\n+ mocked_store_build_logs.assert_called_with(config_name, SUCCESS_RESULT)\n" } ]