diff --git a/doc/adding-fuzzer.rst b/doc/adding-fuzzer.rst index 976e33c..6664883 100644 --- a/doc/adding-fuzzer.rst +++ b/doc/adding-fuzzer.rst @@ -4,15 +4,15 @@ Adding a Fuzzer Prerequisites ------------- -In order to be integrated in PASTIS a fuzzer should support: +To be integrated in PASTIS a fuzzer should support: * **receiving new inputs** while running * providing input/crashes it generates *(preferably as soon as they are produced)* * providing some telemetry that could be sent to the broker *(current coverage, execution per/sec etc..)* -If the fuzzer does not support such features it first it needs to be modified in order to be compliant. +If the fuzzer does not support such features, it first it needs to be modified in order to be compliant. As a matter of example, Honggfuzz does not support such feature and has been modified -to periodically reading a specific folder for new input files. +to periodically read a specific folder for new input files. Python Driver ------------- @@ -20,11 +20,11 @@ Python Driver Once the fuzzer is ready and satisfies PASTIS constraints, it can be wrapped in a Python driver that will act as an interface between the fuzzer and the broker. For that, all the required objects and types have been defined in ``libpastis``. It notably provides a :py:class:`ClientAgent` object -(:ref:`label_agent_api`) class that one can inherit from in order to interact with the broker. The whole purpose +(:ref:`label_agent_api`) class that one can inherit from to interact with the broker. The whole purpose of a driver is to send data generated by the fuzzer to the broker and to react to some messages it receives from the broker. -A client can send to following message to the broker: +A client can send the following message to the broker: * **Hello**: Indicating its own architecture, number of CPUs, memory, platform and the engines it supports * **Seed**: To seed an input or a crash it has generated @@ -32,11 +32,11 @@ A client can send to following message to the broker: * **Log**: Logs that seem's relevant to be sent to the broker * **Telemetry**: Periodic data about the current state of the fuzzing (exec per/sec, coverage, iterations ...) -Conversely it will receive messages, materialized by callbacks that will be called and that the driver should +Conversely, it will receive messages, materialized by callbacks, that will be called and that the driver should implement. The messages are the following: -* **Start**: Message to start a fuzzing campain. It contains the program, all the runtime configuration parameters - that should be used. When receive such callbacks the driver should: +* **Start**: Message to start a fuzzing campaign. It contains the program, all the runtime configuration parameters + that should be used. When receiving such callbacks, the driver should: * check that a campaign is not already running * perform all the initializations * start the underlying fuzzer with the program and appropriate parameters @@ -44,7 +44,7 @@ implement. The messages are the following: any pre-processing that it considers useful. * **Stop**: Should stop the campaign -The following snippet show how to instanciate a very simple client agent. +The following snippet shows how to instantiate a basic client agent. .. code-block:: python @@ -74,18 +74,18 @@ The following snippet show how to instanciate a very simple client agent. agent.start() # start reception thread agent.send_hello([(FuzzingEngine.TRITON, "v0.8")]) -.. warning:: We recommand inheriting ``ClientAgent`` rather than using it a a standalone object. +.. warning:: We recommend inheriting ``ClientAgent`` rather than using it a a standalone object. Broker Addon ------------ -The broker is designed to be independent from the fuzzing engine with which it interacts. +The broker is designed to be independent of the fuzzing engine with which it interacts. Nonetheless, the broker needs to know which program variant to send the fuzzing engine. Thus, the fuzzing engine should provide a ``FuzzingEngineDescriptor`` providing basic information about the engine and a ``accept_file(file)`` method to know whether the file -is suitable for the engine. The broker is then design with an "addon" mecanism allowing -to provide the broker a ``FuzzingEngineDescriptor`` object. The following figure summarizes +is suitable for the engine. The broker is then designed with an "add-on" mechanism allowing +to provide the broker with a ``FuzzingEngineDescriptor`` object. The following figure summarizes the handshake process between a client and the broker. @@ -94,24 +94,24 @@ the handshake process between a client and the broker. :align: center :alt: Loading of a Fuzzing Engine descriptor -As shown on the figure when a fuzzer do connect on the broker it should advertize one or multiple +As shown in the figure, when a fuzzer connects to the broker, it should advertise one or multiple fuzzing engine that it supports with ``FuzzingEngineInfo`` *(all that is encapsulated in the HelloMsg)*. That object also indicates the name -of the python module to side-load on the broker in order to obtain the ``FuzzingEngineDescriptor`` -associated with the engine. The python module should contains a class inheriting ``FuzzingEngineDescriptor`` +of the python module to side-load on the broker to obtain the ``FuzzingEngineDescriptor`` +associated with the engine. The python module should contain a class inheriting ``FuzzingEngineDescriptor`` that describe the engine. *(The broker will automatically find the subclass object within the module).* -You fuzzing driver should thus provide this module to be installed on the broker *(either as a -separate python module or directly within the main driver code)*. The object is is documented +Your fuzzing driver should thus provide this module to be installed on the broker *(either as a +separate Python module or directly within the main driver code)*. The object is documented in the API: :ref:`label_engine_desc`. Examples -------- -There is nothing like examples. Thus one can see existing drivers to see how they are +There is nothing like an example. Thus, one can see existing drivers to see how they are implemented and how they work. They can be used as a basis for other fuzzing engines. * tiny test clients: `test_client.py `_ -* Honggfuzz driver: `driver.py `_ -* TritonDSE driver: `driver.py `_ +* Honggfuzz driver: `driver.py `_ +* TritonDSE driver: `driver.py `_