20. Client message protocol

20.1. Introduction

.intro: The client message protocol provides a means by which clients can receive messages from the MPS. The motivating use case is finalization notification (see design.mps.finalize), but the mechanism is also used for feedback about collections.

.contents: This document describes the design of the external and internal interfaces and concludes with a sketch of an example design of an internal client. The example is that of implementing finalization using the MRG pool.

.readership: Any MPS developer.

20.2. Requirements

.req.synchronous: The message protocol must be synchronous with the client program: that is, the client program must be able to choose when to collect and act on messages. Justification: [Boehm_2002] shows that asynchronous finalization is impossible to implement correctly.

.req.reliable: Posting a message must be reliable: that is, it must not fail for a dynamic reason such as running out memory to store the message. Justification: messages can’t be used to implement finalization unless the messages can be delivered reliably.

.req.extensible.types: The message mechanism must be extensible with new types of message in future versions of the MPS, without breaking client programs that do not receive those types of message.

.req.resources: It follows from .req.extensible.types that messages must not use resources unless the client program has requested them (otherwise resources would leak in client programs that have not been updated to handle new types of message).

.req.extensible.fields: It must be possible to add new fields to existing types of message in future versions of the MPS, without breaking client programs that do not receive those types of message.

20.3. Design

.sol.synchronous: Messages are stored on a ring belonging to the arena. An interface is provided that allows the client program to collect messages from the ring at a time of its choosing.

.sol.reliable: The memory needed for the message is allocated at an earlier point in time, when it possible to communicate an allocation failure via a result code. In particular, space for a finalization message is allocated when the client program calls mps_finalize(), and space for trace messages is allocated in the arena (there can be at most one instance of each message per trace, and the maximum number of traces is known statically).

.sol.resources: Messages are not posted unless they belong to a type that has been enabled by the client program calling mps_message_enable(). This means that message types that are not understood by the client program are not posted and use no resources.

.sol.extensible.fields: Message fields are retrieved by calling accessor functions.

20.4. External interface

20.4.1. Functions

.if.fun: The following functions are provided:

.if.fun.poll: mps_message_poll() sees whether there are any messages pending. Returns 1 only if there is a message on the queue of arena. Returns 0 otherwise.

.if.fun.enable: mps_message_type_enable() enables the flow of messages of a certain type. The queue of messages of a arena will contain only messages whose types have been enabled. Initially all message types are disabled. Effectively this function allows the client to declare to the MPS what message types the client understands.

.if.fun.disable: mps_message_type_disable() disables the flow of messages of a certain type. The antidote to mps_message_type_enable(). Disables the specified message type. Flushes any existing messages of that type on the queue, and stops any further generation of messages of that type. This permits clients to dynamically decline interest in a message type, which may help to avoid a memory leak or bloated queue when the messages are only required temporarily.

.if.fun.get: mps_message_get() begins a message “transaction”. If there is a message of the specified type on the queue then the first such message will be removed from the queue and a handle to it will be returned to the client via the messageReturn argument; in this case the function will return TRUE. Otherwise it will return FALSE. Having obtained a handle on a message in this way, the client can use the type-specific accessors to find out about the message. When the client is done with the message the client should call mps_message_discard(); failure to do so will result in a resource leak.

.if.fun.discard: mps_message_discard() ends a message “transaction”. It indicates to the MPS that the client is done with this message and its resources may be reclaimed.

.if.fun.type.any: mps_message_queue_type() determines the type of a message in the queue. Returns TRUE only if there is a message on the queue of arena, and in this case updates the typeReturn argument to be the type of a message in the queue. Otherwise returns FALSE.

.if.fun.type: mps_message_type() determines the type of a message (that has already been got). Only legal when inside a message transaction (that is, after mps_message_get() and before mps_message_discard()). Note that the type will be the same as the type that the client passed in the call to mps_message_get().

20.4.2. Types of messages

.type: The type governs the “shape” and meaning of the message.

.type.int: A message type is an integer belonging to the MessageType enumeration.

.type.semantics: A type indicates the semantics of the message.

.type.semantics.interpret: The semantics of a message are interpreted by the client by calling various accessor methods on the message.

.type.accessor: The type of a message governs which accessor methods are legal to apply to the message.

.type.finalization: There is a finalization type, MessageTypeFINALIZATION.

.type.finalization.semantics: A finalization message indicates that an object has been discovered to be finalizable (see design.mps.poolmrg.def.final.object for a definition of finalizable).

.type.finalization.ref: The accessor function mps_message_finalization_ref() retrieves the reference to the object which is finalizable.

.type.finalization.ref.scan: Note that the reference returned must be stored in scanned memory.

20.5. Internal interface

20.5.1. Types

struct MessageStruct *Message

.message.type: Message is the type of messages.

.message.instance: Messages are instances of Message Classes.

.message.concrete: Concretely a message is represented by a MessageStruct. A MessageStruct has the usual signature field (see design.mps.sig). A MessageStruct has a type field which defines its type, a ring node, which is used to attach the message to the queue of pending messages, a class field, which identifies a MessageClass object.

.message.intent: The intention is that a MessageStruct will be embedded in some richer object which contains information relevant to that specific type of message.

.message.struct: The structure is declared as follows:

typedef struct mps_message_s {
  Sig sig;                      /* <design/sig/> */
  Arena arena;                  /* owning arena */
  MessageClass klass;           /* Message Class Structure */
  Clock postedClock;            /* mps_clock() at post time, or 0 */
  RingStruct queueRing;         /* Message queue ring */
} MessageStruct;
struct MessageClassStruct *MessageClass

.class: A message class is an encapsulation of methods. It encapsulates methods that are applicable to all types of messages (generic) and methods that are applicable to messages only of a certain type (type-specific).

.class.concrete: Concretely a message class is represented by a MessageClassStruct (a struct). Clients of the Message module are expected to allocate storage for and initialise the MessageClassStruct. It is expected that such storage will be allocated and initialised statically.

.class.one-type: A message class implements exactly one message type. The identifier for this type is stored in the type field of the MessageClassStruct. Note that the converse is not true: a single message type may be implemented by two (or more) different message classes (for example: for two pool classes that require different implementations for that message type).

.class.methods.generic: The generic methods are as follows:

  • delete – used when the message is destroyed (by the client calling mps_message_discard()). The class implementation should finish the message (by calling MessageFinish()) and storage for the message should be reclaimed (if applicable).

.class.methods.specific: The type specific methods are:

.class.methods.specific.finalization: Specific to MessageTypeFINALIZATION:

  • finalizationRef – returns a reference to the finalizable object represented by this message.

.class.methods.specific.gc: Specific to MessageTypeGC:

  • gcLiveSize – returns the number of bytes (of objects) that were condemned by the trace but survived.

  • gcCondemnedSize – returns the number of bytes condemned by the trace.

  • gcNotCondemnedSize – returns the the number of bytes (of objects) that are collectable but were not condemned by the trace.

.class.methods.specific.gcstart: Specific to MessageTypeGCSTART:

  • gcStartWhy – returns an English-language description of the reason why the trace was started.

.class.sig.double: The MessageClassStruct has a signature field at both ends. This is so that if the MessageClassStruct changes size (by adding extra methods for example) then any static initializers will generate errors from the compiler (there will be a type error causes by initialising a non-signature type field with a signature) unless the static initializers are changed as well.

.class.struct: The structure is declared as follows:

typedef struct MessageClassStruct {
  Sig sig;                      /* <design/sig/> */
  const char *name;             /* Human readable Class name */

  MessageType type;             /* Message Type */

  /* generic methods */
  MessageDeleteMethod delete;   /* terminates a message */

  /* methods specific to MessageTypeFINALIZATION */
  MessageFinalizationRefMethod finalizationRef;

  /* methods specific to MessageTypeGC */
  MessageGCLiveSizeMethod gcLiveSize;
  MessageGCCondemnedSizeMethod gcCondemnedSize;
  MessageGCNotCondemnedSizeMethod gcNotCondemnedSize;

  /* methods specific to MessageTypeGCSTART */
  MessageGCStartWhyMethod gcStartWhy;

  Sig endSig;                   /* <design/message/#class.sig.double> */
} MessageClassStruct;

.space.queue: The arena structure is augmented with a structure for managing for queue of pending messages. This is a ring in the ArenaStruct:

struct ArenaStruct
{
  ...
  RingStruct messageRing;
  ...
}

20.5.2. Functions

void MessageInit(Arena arena, Message message, MessageClass klass, MessageType type)

.fun.init: Initializes the MessageStruct pointed to by message. The caller of this function is expected to manage the store for the MessageStruct.

void MessageFinish(Message message)

.fun.finish: Finishes the MessageStruct pointed to by message. The caller of this function is expected to manage the store for the MessageStruct.

void MessagePost(Arena arena, Message message)

.fun.post: Places a message on the queue of an arena.

.fun.post.precondition: Prior to calling the function, the queueRing field of the message must be a singleton (design.mps.ring.def.singleton). After the call to the function the message will be available for MPS client to access. After the call to the function the message fields must not be manipulated except from the message’s class’s method functions (that is, you mustn’t poke about with the queueRing field in particular).

void MessageEmpty(Arena arena)

.fun.empty: Empties the message queue. This function has the same effect as discarding all the messages on the queue. After calling this function there will be no messages on the queue.

.fun.empty.internal-only: This functionality is not exposed to clients. We might want to expose this functionality to our clients in the future.

20.6. Message life cycle

.life.alloc: Space for the message structure is allocated at the earliest point in time when the MPS knows that the message might be needed.

.life.init: The message structure is initialized by calling MessageInit().

.life.post: The message is posted on the arena’s message queue by calling MessagePost().

.life.get: The client program retrieves the message by calling mps_message_get().

.life.discard: The client program indicates that it is finished with the message by calling mps_message_discard().

.life.reuse: The MPS may reuse the message structure, in which case the lifecycle continues from .life.post.

.life.delete: When the MPS no longer needs the message structure, its delete method is called.

20.7. References

Boehm_2002

Hans-J. Boehm. 2002. “Destructors, Finalizers, and Synchronization”. HP Labs technical report HPL-2002-335.