pipewire/doc/dox/internals/protocol.dox

/** \page page_native_protocol Native Protocol

PipeWire has a pluggable client/server IPC protocol.

The reference implementation uses unix sockets and is implemented in
\ref page_module_protocol_native.

We document the messages here.

\tableofcontents

# Message header {#native-protocol-message-header}

Each message on the unix socket contains a 16 bytes header and a
variable length payload size:

```
    0                   1                   2                   3
    0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                     Id                                        |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |    opcode     |                   size                        |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                         seq                                   |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                         n_fds                                 |
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                     payload POD                               |
   .                                                               .
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
   |                     optional footer POD                       |
   .                                                               .
   +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
```


These are four uint32 words to be read in native endianness.

- Id: the message id this is the destination resource/proxy id
- opcode: the opcode on the resource/proxy interface
- size: the size of the payload and optional footer of the message
- seq: an increasing sequence number for each message
- n_fds: number of file descriptors in this message.

The sender should send along with each message the fds that belong to
the message. If there are more than the maximum number of fds in the
message than can fit in one message, the message is split into multiple
parts.

A receiver needs to first read the header to know the size of the
complete message. After collecting the complete message, it can start
processing it.

The payload is a single POD see \ref page_spa_pod for details.

After the payload, there is an optional footer POD object.

# Making a connection {#native-protocol-making-connection}

First a connection is made to a unix domain socket. By default, the socket is
named as "pipewire-0" and searched in the following directories:

 - getenv("PIPEWIRE_RUNTIME_DIR")
 - getenv("XDG_RUNTIME_DIR")
 - getenv("USERPROFILE")


The client opens the socket and allocates a Core proxy with id 0 and a Client
proxy with id 1. The proxy is nothing more than a way to identify the remote
objects with an id. Usually they also contain the type and version of the
remote object and some helpers (interfaces) to dispatch messages.

The server will allocate a new Core resource with id 0. Resources are the
counterpart of client proxies and also typically contain extra information
and helpers (interfaces) to dispatch messages.

The client sends the Core::Hello message on the socket to start the
communication.

The server binds the client resource with id 1 to the client. This means
that the client object on the server will now have a resource with a
client side proxy associated with it.

When the client sends a message for a certain proxy, the server side
resource with the same id for the client is found and the message is
dispatched to the object associated with the resource.

Similarly when a server sends a message on a resource, the client will
find the proxy with the same id and dispatch the message to it's
associated object.

The client then sends client properties to the server.

```
   client                                    server
     |---------------------------------------->|
     |            open socket                  |
     |                                         |
     |---------------------------------------->|
     |            Core::Hello(version)         |
     |                                         |
     |---------------------------------------->|
     |       Client::UpdateProperties          |
     .                                         .
```

This completes the setup of the client. The newly connected client will
appear in the registry at this point.

# Core proxy/resource {#native-protocol-core}

The core is always the object with Id 0.

## Core Methods (Id 0)

### Core::Hello (Opcode 1)

The first message sent by a client is the Hello message and contains the
version number of the client.

```
   Struct(
      Int: version
   )
```

The version is 3.

### Core::Sync (Opcode 2)

The Sync message will result in a Done event from the server. When the Done
event is received, the client can be sure that all operations before the Sync
method have been completed.

```
   Struct(
      Int: id
      Int: seq
   )
```

- id: the id will be returned in the Done event.
- seq: is usually generated automatically and will be returned in the Done event.

### Core::Pong (Opcode 3)

Is sent from the client to the server when the server emits the Ping event.
The id and seq should be copied from the Ping event.

```
   Struct(
      Int: id
      Int: seq
   )
```

### Core::Error (Opcode 4)

An error occurred in an object on the client.

```
   Struct(
      Int: id
      Int: seq
      Int: res
      String: message
   )
```

- id: The id of the proxy that is in error.
- seq: a seq number from the failing request (if any)
- res: a negative errno style error code
- message: an error message

### Core::GetRegistry (Opcode 5)

A client requests to bind to the registry object and list the available objects
on the server.

Like with all bindings, first the client allocates a new proxy id and puts this
as the new_id field. Methods and Events can then be sent and received on the 
new_id (in the message Id field).

```
   Struct(
      Int: version
      Int: new_id
   )
```
- version: the version of the registry interface used on the client
- new_id: the id of the new proxy with the registry interface

After this method, the server will start sending Registry::Global events
to the proxy with new_id.

```
   client                                    server
     |                                         |
     | new proxy(new_id)                       |
     |---------------------------------------->|  new resource(new_id)
     |      Core::GetRegistry(version)         |
     |                                         |
     |<----------------------------------------| new_id
     |      Registry::Global()                 |
     |<----------------------------------------|
     |      Registry::Global()                 |
     .                                         .
```

There is no explicit last Global event to signal that the last object
has been received. The usual way of knowing this is to send a Core::Sync
method right after the Core::GetRegistry method and to wait for the
Core::Done event.

```
   client                                    server
     |                                         |
     | new proxy(new_id)                       |
     |---------------------------------------->|  new resource(new_id)
     |      Core::GetRegistry(version)         |
     |---------------------------------------->|
     |  seq = Core::Sync                       |
     |                                         |
     |<----------------------------------------| new_id
     |      Registry::Global()                 |
     |<----------------------------------------|
     |      Registry::Global()                 |
     .                                         .
     |<----------------------------------------|
     |         Core::Done(seq)                 |
     .                                         .
```

### Core::CreateObject (Opcode 6)

Create a new object from a factory of a certain type.

The client allocates a new_id for the proxy. The server will allocate a new
resource with the same new_id and from then on, Methods and Events will be 
exchanged between the new object of the given type.

```
   Struct(
      String: factory_name
      String: type
      Int: version
      Struct(
         Int: n_items
	 (String: key
	  String: value)*
      ): props
      Int: new_id
   )
```

- factory_name: the name of a server factory object to use
- type: the type of the object to create, this is also the type of the
      interface of the new_id proxy.
- props: extra properties to create the object
- new_id: the proxy id of the new object

### Core::Destroy (Opcode 7)

Destroy an object.

```
   Struct(
      Int: id
   )
```

- id: the proxy id of the object to destroy.

## Core Events

Core events are emitted by the server.

### Core::Info (Opcode 0)

Emitted by the server upon connection with the more information about the
server.

```
   Struct(
      Int: id
      Int: cookie
      String: user_name
      String: host_name
      String: version
      String: name
      Long: change_mask
      Struct( 
         Int: n_items
	 (String: key
	  String: value)*
      ): props
   )
```

- id: the id of the server (0)
- cookie: a unique cookie for this server
- user_name: the name of the user running the server
- host_name: the name of the host running the server
- version: a version string of the server
- name: the name of the server
- change_mask: a set of bits with changes to the info
- props: optional key/value properties, valid when change_mask has (1<<0)

### Core::Done (Opcode 1)

Emitted as a result of a client Sync method.

```
   Struct(
      Int: id
      Int: seq
   )
```

id and seq are passed from the client Sync request.

### Core::Ping (Opcode 2)

Emitted by the server when it wants to check if a client is alive or ensure
that it has processed the previous events.

```
   Struct(
      Int: id
      Int: seq
   )
```

- id: the object id to ping
- seq: usually automatically generated. The client should pass this in the Pong
       method reply.

### Core::Error (Opcode 3)

The error event is sent out when a fatal (non-recoverable)
error has occurred. The id argument is the proxy object where
the error occurred, most often in response to a request to that
object. The message is a brief description of the error,
for (debugging) convenience.

```
   Struct(
      Int: id
      Int: seq
      Int: res
      String: message
   )
```

- id: The id of the resource that is in error.
- seq: a seq number from the failing request (if any)
- res: a negative errno style error code
- message: an error message


### Core::RemoveId (Opcode 4)

This event is used internally by the object ID management
logic. When a client deletes an object, the server will send
this event to acknowledge that it has seen the delete request.
When the client receives this event, it will know that it can
safely reuse the object ID.

```
   Struct(
     Int: id
   )
```

- id: a proxy id that was removed

### Core::BoundId (Opcode 5)

This event is emitted when a local object ID is bound to a
global ID. It is emitted before the global becomes visible in the
registry. This event is deprecated, the BoundProps event should
be used because it also contains extra properties.

```
   Struct(
     Int: id
     Int: global_id
   )
```

- id: a proxy id
- global_id: the global_id as it will appear in the registry.

### Core::AddMem (Opcode 6)

Memory is given to a client as fd of a certain memory type.
Further references to this fd will be made with the per memory
unique identifier id.

```
   Struct(
     Int: id
     Id: type
     Fd: fd
     Int: flags
   )
```

- id: a server allocated id for this memory
- type: the memory type, see enum spa_data_type
- fd: the index of the fd sent with this message
- flags: extra flags

### Core::RemoveMem (Opcode 7)

Remove memory for a client with the given id

```
   Struct(
     Int: id
   )
```

- id: the id of the memory to remove. This is the Id from AddMem.

### Core::BoundProps (Opcode 8)

This event is emitted when a local object ID is bound to a
global ID. It is emitted before the global becomes visible in the
registry.

```
   Struct(
      Int: id
      Int: global_id
      Struct( 
         Int: n_items
	 (String: key
	  String: value)*
      ): props
   )
```

- id: a proxy id
- global_id: the global_id as it will appear in the registry.
- props: the properties of the global

# Registry proxy/resource {#native-protocol-registry}

The registry is obtained with the GetRegistry method on the Core object.
The Id depends on the new_id that was provided.

## Registry Methods

### Registry::Bind (Opcode 1)

Bind to the global object with id and use the client proxy
with new_id as the proxy. After this call, methods can be
send to the remote global object and events can be received.

```
   Struct(
      Int: id
      String: type
      Int: version
      Int: new_id
   )
```

- id: the global_id to bind to
- type: the type of the global id
- version: the client version of the interface for type
- new_id: the client proxy id for the global object

### Registry::Destroy (Opcode 2)

Try to destroy the global object with id. This might fail when the
client does not have permission.

```
   Struct(
      Int: id
   )
```

- id: the global id to destroy.

## Registry Events

### Registry::Global (Opcode 0)

Notify a client about a new global object.

```
   Struct(
      Int: id
      Int: permissions
      String: type
      Int: version
      Struct( 
         Int: n_items
	 (String: key
	  String: value)*
      ): props
   )
```

- id: the global id
- permissions: permission bits
- type: the type of object
- version: the server version of the object
- props: extra global properties

### Registry::GlobalRemove (Opcode 1)

A global with id was removed

```
   Struct(
      Int: id
   )
```

- id: the global id that was removed.

# PipeWire:Interface:Client {#native-protocol-client}

The client object represents a client connect to the PipeWire server.
Permissions of the client can be managed.

The currently connected client always has the Client object with
proxy id 1.

## Client Methods

### Client::Error (Opcode 1)

Is used to send an error to a client.

```
   Struct(
      Int: id
      Int: res
      String: error
   )
```

- id: a client proxy id to send the error to
- res: a negative errno style error code
- error: an error message

### Client::UpdateProperties (Opcode 2)

Is used to update the properties of a client.

```
   Struct(
      Struct( 
         Int: n_items
	 (String: key
	  String: value)*
      ): props
   )
```
- props: properties to update on the client.

### Client::GetPermissions (Opcode 3)

Get the currently configured permissions on the client.

```
   Struct(
      Int: index
      Int: num
   )
```
- index: the start index of the permissions to get
- num: the number of permissions to get

This method will result in at most num Permission Events.

### Client::UpdatePermissions (Opcode 4)

Update the permissions of the global objects using the
provided array with permissions


```
   Struct(
      Int: n_permissions
      ( Int: id
        Int: permission
      )*
   )
```

- n_permissions: number of permissions
- id: the global id
- permissions: the permissions for the global id

## Client Events

### Client::Info (Opcode 0)

Get client information updates. This is emitted when binding to a client or
when the client info is updated later.

```
   Struct(
      Int: id
      Long: change_mask
      Struct( 
         Int: n_items
	 (String: key
	  String: value)*
      ): props
   )
```
- id: the global id of the client
- change_mask: the changes emitted by this event
- props: properties of this object, valid when change_mask has (1<<0)

### Client::Permissions (Opcode 1)

Emitted as the reply of the GetPermissions method.

```
   Struct(
      Int: index
      Struct( 
         Int: n_permissions
	 (Int: id
	  Int: permission)*
      )
   )
```
- index: index of the first permission
- n_permissions: the number of permission entries
- id: the global id of the object
- permissions: the permission for the given id

# PipeWire:Interface:Device {#native-protocol-device}

A device is an object that manages other devices or nodes.

The usual flow is to bind to the Device object. This will result in an
Info event. From the Info event one can find the available params to
enumerate.

## Device methods

### Device::SubscribeParams (Opcode 1)

Automatically emit Param events for the given ids when they are changed.

```
   Struct(
      Array[Id]: ids
   )
```
- ids: and array of param Id to subscribe to

### Device::EnumParams (Opcode 2)

Enumerate the values of a param. This will result in Param events.

```
   Struct(
      Int: seq
      Id: id
      Int: index
      Int: num
      Pod: filter
   )
```
- seq: an automatically generated sequence number, will be copied into the reply
- id: the param id to enumerate.
- index: the first param index to retrieve
- num: the number of params to retrieve
- filter: an optional filter object for the param.

### Device::SetParam (Opcode 3)

Set a parameter on the Device.

```
   Struct(
      Id: id
      Int: flags
      Pod: param
   )
```
- id: the param id to set.
- flags: extra flags
- param: the param object to set

## Device events

### Device::Info (Opcode 0)

The info event is emitted when binding or when the device information changed.

```
   Struct(
      Int: id
      Long: change_mask
      Struct( 
         Int: n_items
	 ( String: key
	   String: value )*
      ): props
      Struct( 
         Int: n_params
	 ( Int: id
	   Int: flags )* 
      ): param_info
   )
```

- id: the id of the global
- change_mask: a bitmask of changed fields
- props: extra properties, valid when change_mask is (1<<0)
- param_info: info about the parameters, valid when change_mask is (1<<1)
            For each parameter, the id and current flags are given.
	- param_info.id : see enum spa_param_type
	- param_info.flags: struct spa_param_info.flags

### Device::Param (Opcode 1)

Emitted as a result of EnumParams or SubscribeParams.

```
   Struct(
      Int: seq
      Id: id
      Int: index
      Int: next
      Pod: param
   )
```
- seq: the sequence number send by the client EnumParams or server generated
     in the SubscribeParams case.
- id: the param id that is reported, see enum spa_param_type 
- index: the index of the parameter
- next: the index of the next parameter
- param: the parameter. The object type depends on the id


# PipeWire:Interface:Factory {#native-protocol-factory}

A factory is an object that allows one to create new objects.

## Factory methods

A factory has no methods

## Factory events

### Factory::Info (Opcode 0)

Info is emitted when binding to the factory global or when the information changed.

```
   Struct(
      Int: id
      String: name
      String: type
      Int: version
      Long: change_mask
      Struct( 
         Int: n_items
	 ( String: key
	   String: value )*
      ): props
   )
```
- id: the global id of the factory
- name: the name of the factory. This can be used as the name for Core::CreateObject
- type: the object type produced by this factory
- version: the version of the object interface
- change_mask: bitfield of changed values.
- props: optional properties of the factory, valid when change_mask is (1<<0)


# PipeWire:Interface:Link {#native-protocol-link}

A link is a connection between 2 ports.

## Link methods

A link has no methods

## Link events

### Link::Info (Opcode 0)

Info is emitted when binding to the link global or when the information changed.

```
   Struct(
      Int: id
      Int: output_node_id
      Int: output_port_id
      Int: input_node_id
      Int: input_port_id
      Long: change_mask
      Int: state
      String: error
      Pod: format
      Struct( 
         Int: n_items
	 ( String: key
	   String: value )*
      ): props
   )
```

- id: the global id of the link
- output_node_id: the global id of the output node
- output_port_id: the global id of the output port
- input_node_id: the global id of the input node
- input_port_id: the global id of the input port
- change_mask: bitfield of changed values.
- state: the state of the link, valid when change_mask has (1<<0)
       - see enum pw_link_state for values
- error: an optional error string
- format: an optional format for the link, valid when change_mask has (1<<1)
- props: optional properties of the link, valid when change_mask is (1<<2)


# PipeWire:Interface:Module {#native-protocol-module}

A Module provides dynamically loaded functionality

## Module methods

A module has no methods

## Module events

### Module::Info (Opcode 0)

Info is emitted when binding to the module global or when the information changed.

```
   Struct(
      Int: id
      String: name
      String: filename
      String: args
      Long: change_mask
      Struct( 
         Int: n_items
	 ( String: key
	   String: value )*
      ): props
   )
```

- id: the global id of the module
- name: the name of the module
- filename: the filename of the module
- args: arguments passed when loading the module
- change_mask: bitfield of changed values.
- props: optional properties of the module, valid when change_mask has (1<<0)


# PipeWire:Interface:Node {#native-protocol-node}

A Node is a processing element in the graph

## Node methods

### Node::SubscribeParams (Opcode 1)

Automatically emit Param events for the given ids when they are changed.

```
   Struct(
      Array[Id]: ids
   )
```
- ids: and array of param Id to subscribe to

### Node::EnumParams (Opcode 2)

Enumerate the values of a param. This will result in Param events.

```
   Struct(
      Int: seq
      Id: id
      Int: index
      Int: num
      Pod: filter
   )
```
- seq: an automatically generated sequence number, will be copied into the reply
- id: the param id to enumerate.
- index: the first param index to retrieve
- num: the number of params to retrieve
- filter: an optional filter object for the param.

### Node::SetParam (Opcode 3)

Set a parameter on the Node.

```
   Struct(
      Id: id
      Int: flags
      Pod: param
   )
```
- id: the param id to set.
- flags: extra flags
- param: the param object to set

### Node::SendCommand (Opcode 4)

Send a Command to the node.

```
   Struct(
      Pod: command
   )
```
- command: the command to send. See enum spa_node_command 

## Node events

### Node::Info (Opcode 0)

The info event is emitted when binding or when the node information changed.

```
   Struct(
      Int: id
      Int: max_input_ports
      Int: max_output_ports
      Long: change_mask
      Int: n_input_ports
      Int: n_output_ports
      Id: state
      String: error
      Struct( 
         Int: n_items
	 ( String: key
	   String: value )*
      ): props
      Struct( 
         Int: n_params
	 ( Int: id
	   Int: flags )* 
      ): param_info
   )
```

- id: the id of the node global
- max_input_port: the maximum input ports for the node
- max_output_port: the maximum output ports for the node
- change_mask: a bitmask of changed fields
- n_input_port: the number of input ports, when change_mask has (1<<0)
- n_output_port: the number of output ports, when change_mask has (1<<1)
- state: the current node state, when change_mask has (1<<2)
       - See enum pw_node_state  for values
- error: an error message.
- props: extra properties, valid when change_mask is (1<<3)
- param_info: info about the parameters, valid when change_mask is (1<<4)
            For each parameter, the id and current flags are given.
	- param_info.id : see enum spa_param_type
	- param_info.flags: struct spa_param_info.flags

### Node::Param (Opcode 1)

Emitted as a result of EnumParams or SubscribeParams.

```
   Struct(
      Int: seq
      Id: id
      Int: index
      Int: next
      Pod: param
   )
```
- seq: the sequence number send by the client EnumParams or server generated
     in the SubscribeParams case.
- id: the param id that is reported, see enum spa_param_type 
- index: the index of the parameter
- next: the index of the next parameter
- param: the parameter. The object type depends on the id


# PipeWire:Interface:Port {#native-protocol-port}

A port is part of a node and allows links with other ports.

## Port methods

### Port::SubscribeParams (Opcode 1)

Automatically emit Param events for the given ids when they are changed.

```
   Struct(
      Array[Id]: ids
   )
```
- ids: and array of param Id to subscribe to

### Port::EnumParams (Opcode 2)

Enumerate the values of a param. This will result in Param events.

```
   Struct(
      Int: seq
      Id: id
      Int: index
      Int: num
      Pod: filter
   )
```
- seq: an automatically generated sequence number, will be copied into the reply
- id: the param id to enumerate.
- index: the first param index to retrieve
- num: the number of params to retrieve
- filter: an optional filter object for the param.

## Port events

### Port::Info (Opcode 0)

The info event is emitted when binding or when the port information changed.

```
   Struct(
      Int: id
      Int: direction
      Long: change_mask
      Struct( 
         Int: n_items
	 ( String: key
	   String: value )*
      ): props
      Struct( 
         Int: n_params
	 ( Int: id
	   Int: flags )* 
      ): param_info
   )
```

- id: the id of the port global
- direction: the direction of the port, see enum pw_direction
- change_mask: a bitmask of changed fields
- props: extra properties, valid when change_mask is (1<<0)
- param_info: info about the parameters, valid when change_mask is (1<<1)
            For each parameter, the id and current flags are given.
	- param_info.id : see enum spa_param_type
	- param_info.flags: struct spa_param_info.flags

### Port::Param (Opcode 1)

Emitted as a result of EnumParams or SubscribeParams.

```
   Struct(
      Int: seq
      Id: id
      Int: index
      Int: next
      Pod: param
   )
```
- seq: the sequence number send by the client EnumParams or server generated
     in the SubscribeParams case.
- id: the param id that is reported, see enum spa_param_type 
- index: the index of the parameter
- next: the index of the next parameter


# PipeWire:Interface:ClientNode {#native-protocol-clientnode}

The ClientNode object is created from the `client-node` factory that is provided
by the `libpipewire-module-client-node` module.

It creates a server Node that can be controlled from a client. Processing will happen
in the client. It is used by pw-stream and pw-filter to implement the PipeWire media
processing nodes.

To create a client-node, one must first connect to the server and then make a
ClientNode proxy and do Core::CreateObject with the client-node factory.
This will create a server side ClientNode resource that can be controlled with the
proxy.

After the proxy is set up, the conversation between client and server goes as
follows:

```
   client                                    server
     |                                         |
     |---------------------------------------->|
     |        ClientNode::Update               | send initial node configuration
     |                                         |
     |<----------------------------------------|
     |       Core::AddMem                      | memory for node activation
     |<----------------------------------------|
     |       ClientNode::SetActivation         | the node activation
     |<----------------------------------------|
     |       ClientNode::Transport             | the node transport
     |<----------------------------------------|
     |       ClientNode::SetIO                 | clock IO
     |                                         |
     |---------------------------------------->|
     |       ClientNode::SetActive(true)       | set the node active
     |                                         |
     |<----------------------------------------|
     |       ClientNode::SetParam              | optional volume restore/settings
     |<----------------------------------------|
     |       ClientNode::SetParam              | optional PortConfig if needed
     |---------------------------------------->|
     |       ClientNode::Update                | Upload changed params
     |                                         |
     |---------------------------------------->|
     |       ClientNode::PortUpdate            | config for each port
     .                                         .
     |<----------------------------------------|
     |       ClientNode::PortSetMixInfo        | mixer inputs for each linked port
     |<----------------------------------------|
     |       ClientNode::PortSetParam          | Latency of the ports
     |<----------------------------------------|
     |       ClientNode::SetActivation         | activation of port peers
     |---------------------------------------->|
     |       ClientNode::PortUpdate            | Ack port updates
     .                                         .
     |<----------------------------------------|
     |       ClientNode::PortSetParam          | formats of the ports
     |---------------------------------------->|
     |       ClientNode::PortUpdate            | Ack port format update
     .                                         .
     |<----------------------------------------|
     |       Core::AddMem                      | memory for the port buffers
     |<----------------------------------------|
     |       ClientNode::PortUseBuffers        | buffers for a port
     .                                         .
     |<----------------------------------------|
     |       Core::AddMem                      | memory for driver activation
     |<----------------------------------------|
     |       ClientNode::SetActivation         | the driver activation
     |<----------------------------------------|
     |       ClientNode::SetIO                 | driver Position IO
     |<----------------------------------------|
     |       Core::AddMem                      | memory for port IO
     |<----------------------------------------|
     |       ClientNode::PortSetIO             | buffers/async-buffers port IO
     .                                         .
     |<----------------------------------------|
     |       ClientNode::Command               | Start command
     .                                         .
```

## ClientNode methods

### ClientNode::GetNode (Opcode 1)

Get the node object associated with the client-node. This binds to the server side
Node object.

```
   Struct(
      Int: version
      Int: new_id
   )
```
- version: the Node version to bind as
- new_id: the proxy id

### ClientNode::Update (Opcode 2)

Update the params and info of the node.

```
   Struct(
      Int: change_mask
      Int: n_params
      ( Pod: param )*
      Struct(
         Int: max_input_ports
         Int: max_output_ports
	 Long: change_mask
	 Long: flags
	 Int: n_items
	 ( String: key
	   String: value )*
	 Int: n_params
	 ( Id: id
	   Int: flags )*
      ): info
   )
```
- change_mask: a bitfield of changed items
- n_params: number of update params, valid when change_mask has (1<<0)
- param: an updated param
- info: an updated `struct spa_node_info`, valid when change_mask has (1<<1)
  - max_input_ports: maximum input ports
  - max_output_ports: maximum output ports
  - change_mask: bitmask of changed items
  - flags: flags, see `struct spa_node_info`, when change_mask has (1<<0)
  - n_items: updated properties, valid when info.change_mask has (1<<1)
  - n_params: updated `struct spa_param_info`, valid when info.change_mask has (1<<2)


### ClientNode::PortUpdate (Opcode 3)

Create, Update or destroy a node port.

When the port is not known on the server, the port is created.
When info is None, the port is destroyed.
Otherwise, the port information is updated.

```
   Struct(
      Int: direction
      Int: port_id
      Int: change_mask
      Int: n_params
      ( Pod: param )*
      Struct(
	 Long: change_mask
	 Long: flags
	 Int: rate_num
	 Int: rate_denom
	 Int: n_items
	 ( String: key
	   String: value )*
	 Int: n_params
	 ( Id: id
	   Int: flags )*
      ): info
   )
```
- direction: the port direction
- port_id: the port id
- change_mask: a bitfield of changed items
- n_params: number of updated params, valid when change_mask has (1<<0)
- param: n_params updated params
- info: an updated `struct spa_port_info`, valid when change_mask has (1<<1)
  - change_mask: bitmask of changed items
  - flags: flags, see `struct spa_port_info`, when change_mask has (1<<0)
  - rate_num: updated rate numerator
  - rate_denom: updated rate denominator, when info.change_mask has (1<<1)
  - n_items: updated properties, valid when info.change_mask has (1<<2)
  - n_params: updated `struct spa_param_info`, valid when info.change_mask has (1<<3)

### ClientNode::SetActive (Opcode 4)

Set the node active or inactive.

```
   Struct(
      Bool: active
   )
```
- active: the new state of the node

### ClientNode::Event (Opcode 5)

Emit an event on the node.

```
   Struct(
      Pod: event
   )
```
- event: the event to emit. See `enum spa_node_event`.

### ClientNode::PortBuffers (Opcode 6)

This method is used by the client when it has allocated buffers for a port.
It is usually called right after the UseBuffers event to let the server know
about the the newly allocated buffer memory.

```
   Struct(
      Int: direction
      Int: port_id
      Int: mix_id
      Int: n_buffers
      (  Int: n_datas
         (  Id: type
	    Fd: memfd
	    Int: flags
	    Int: mapoffset
	    Int: maxsize
	 )*
      )*
   )
```
- direction: the direction of the port
- port_id: the port id
- mix_id: the mix id of the port
- n_buffer: the number of buffers
- n_data: for each buffer the number of data planes
- type: the plane memory type
- memfd: the plane memfd
- mapoffset: the start offset of where the buffer memory starts
- maxsize: the maximum size of the memory.

## ClientNode events

### ClientNode::Transport (Opcode 0)

The server will allocate the activation record and eventfd for the node and
transfer this to the client with the Transport event.

```
   Struct(
      Fd: readfd
      Fd: writefd
      Int: memfd
      Int: offset
      Int: size
   )
```
- readfd: the eventfd to start processing
- writefd: the eventfd to signal when the driver completes and profiling is
           enabled.
- memfd: the index of the memfd of the activation record
- offset: the offset in memfd of the start of the activation record
- size: the size of the activation record

The activation record is currently an internal data structure that is
not yet ABI stable.

The writefd is meant to wake up the server after the driver completes so
that the profiler can collect the information. The profiler is active
when the pw_node_activation::flags fields has PW_NODE_ACTIVATION_FLAG_PROFILER
set. When the profiler is disabled (or when the node is not driving), this
eventfd should not be signaled.

### ClientNode::SetParam (Opcode 1)

Set a parameter on the Node.

```
   Struct(
      Id: id
      Int: flags
      Pod: param
   )
```
- id: the param id to set.
- flags: extra flags
- param: the param object to set


### ClientNode::SetIO (Opcode 2)

Set an IO area on the node.

```
   Struct(
      Id: id
      Int: memid
      Int: offset
      Int: size
   )
```
- id: the io area id to set.
- the memid to use, this is signaled with Core::AddMem
- offset: the start offset in the memory area
- the size of the io area

### ClientNode::Event (Opcode 3)

Emit an event on the node.

```
   Struct(
      Pod: event
   )
```
- event: the event to emit. See `enum spa_node_event`.

### ClientNode::Command (Opcode 4)

Send a command on the node.

```
   Struct(
      Pod: command
   )
```
- command: the command to send. See `enum spa_node_command`.

### ClientNode::AddPort (Opcode 5)

Add a new port to the node

```
   Struct(
      Int: direction
      Int: port_id
      Struct( 
         Int: n_items
	 ( String: key
	   String: value )*
      ): props
   )
```
- direction: the direction of the new port
- port_id: the port id of the new port
- props: optional extra properties for the port

### ClientNode::RemovePort (Opcode 6)

Remove a port from the node

```
   Struct(
      Int: direction
      Int: port_id
   )
```
- direction: the direction of the port to remove
- port_id: the port id of the port to remove


### ClientNode::PortSetParam (Opcode 7)

Set a parameter on the Port of the node.

```
   Struct(
      Int: direction
      Int: port_id
      Id: id
      Int: flags
      Pod: param
   )
```
- direction: the direction of the port
- port_id: the port id of the port
- id: the param id to set.
- flags: extra flags
- param: the param object to set

### ClientNode::UseBuffers (Opcode 8)

Use a set of buffers on the mixer port

```
   Struct(
      Int: direction
      Int: port_id
      Int: mix_id
      Int: flags
      Int: n_buffers
      (  Int: memid
         Int: offset
	 Int: size
	 Int: n_metas
	 (  Id: type
	    Int: size
	 )*: meta
	 Int: n_datas
	 (  Id: type
	    Int: data
	    Int: flags
	    Int: mapoffset
	    Int: maxsize
	 )*: data
      )*: buffer
   )
```
- direction: the direction of the port
- port_id: the port id of the port
- mix_id: the mixer id of the port
- flags: extra flags
- id: the param id to set.
- flags: extra flags
   - SPA_NODE_BUFFERS_FLAG_ALLOC (1<<0) to let the client allocate
     buffers, in which case the PortBuffers method needs to be called
     with the allocated buffer memory.
- n_buffers: the number of buffers
- memid: the memory id of the buffer metadata and or data
- offset: the offset in memid of the buffer
- size: the size of the buffer metadata or data
- n_metas: number of metadata. The buffer memory first contains this
           number of metadata parts of the given type and size
  - type: metadata type
  - size: metadata size, round up with 8 to get to the next metadata
- n_data: the number of datablocks (planes) per buffer
  - type: the data type, this can be:
  		- SPA_DATA_MemId to reference a memfd from Core:AddMem
		- SPA_DATA_MemPtr to reference this buffer memid
  - data: contains the memid or offset in the memid
  - flags: extra flags for the data
  - mapoffset: the offset in memfd
  - maxsize: the maxsize of the memory in memfd

### ClientNode::PortSetIO (Opcode 9)

Set an IO area on a mixer port.

```
   Struct(
      Int: direction
      Int: port_id
      Int: mix_id
      Id: id
      Int: memid
      Int: offset
      Int: size
   )
```
- direction: the direction of the port
- port_id: the port id of the port
- mix_id: the mix id of the port
- id: the IO area to set. See enum spa_io_type
- memid: the memid of the io area, added with Core::AddMem
- offset: the offset in the memid
- size: the size of the IO area


### ClientNode::SetActivation (Opcode 10)

Notify the client of the activation record of a peer node. This activation record
should be triggered when this node finishes processing.

```
   Struct(
      Int: node_id
      Fd: signalfd
      Int: memid
      Int: offset
      Int: size
   )
```
- node_id: the node_id of the peer node
- signalfd: the eventfd of the peer node
- memid: the memid of the activation record of the peer from Core:AddMem
- offset: the offset in memid
- size: the size of the activation record

### ClientNode::PortSetMixInfo (Opcode 11)

Notify the node of the peer of a mixer port. This can be used to track the peer
ports of a node.

```
   Struct(
      Int: direction
      Int: port_id
      Int: mix_id
      Int: peer_id
      Struct( 
         Int: n_items
	 ( String: key
	   String: value )*
      ): props
   )
```
- direction: the direction of the port
- port_id: the port id of the port
- mix_id: the mix id of the port
- peer_id: the id of the peer port
- props: optional properties

# PipeWire:Interface:Metadata {#native-protocol-metadata}

Metadata is a shared database of settings and properties.

## Metadata methods

### Metadata::SetProperty (Opcode 1)

Set a property in the metadata store.

```
   Struct(
      Int: subject
      String: key
      String: type
      String: value
   )
```
- subject: the id of the object, this needs to be a valid global_id
- key: a key
- type: an optional type
- value: a value


### Metadata::Clear (Opcode 2)

Clear the metadata store.

```
   Struct(
      None
   )
```

## Metadata events

### Metadata::Property (Opcode 0)

A metadata key changed. This is also emitted when binding to the metadata.

```
   Struct(
      Int: subject
      String: key
      String: type
      String: value
   )
```
- subject: the id of the object, this is a valid global_id
- key: a key
- type: an optional type
- value: a value

# PipeWire:Interface:Profiler {#native-protocol-profiler}

The profiler object allows one to receive profiler information of the pipewire
graph.

## Profiler methods

The profiler has no methods

## Profiler events

### Profiler::Profile (Opcode 0)

```
   Struct(
      Pod: object
   )
```
- object: a SPA_TYPE_OBJECT_Profiler object. See enum spa_profiler 


# Footer {#native-protocol-footer}

The message footer contains additional messages, not directed to the
destination object defined by the `Id` field.

The footer consists of a single POD, immediately following the payload
POD.  The footer POD consists of a sequence of footer opcode Ids and
footer payload Structs containing their arguments:

```
   Struct(
      Id: opcode1,
      Struct { ... },
      Id: opcode2,
      Struct { ... },
      ...
   )
```

The footer opcodes are separate for server-to-client (`core`) and
client-to-server (`client`) directions.

The message footer is processed before other parts of the message,
including the object Id lookup.

## Core Generation (Footer Opcode 0)

```
   Struct(
      Long: registry_generation,
   )
```

Indicates to the client what is the current registry generation
number of the \ref pw_context on the server side.

The server shall include this footer in the next message it sends that
follows the increment of the registry generation number.

\see \ref native-protocol-registry-generation

## Client Generation (Footer Opcode 0)

```
   Struct(
      Long: client_generation,
   )
```

Indicates to the server what is the last registry generation number
the client has processed.

The client shall include this footer in the next message it sends,
after it has processed an incoming message whose footer includes a
registry generation update.

\see \ref native-protocol-registry-generation

# Registry generation  {#native-protocol-registry-generation}

The registry generation is a 64-bit integer in the PipeWire server
\ref pw_context that increments by one when the server allocates a new
global \ref PW_KEY_OBJECT_ID for an object.

The server keeps track of each global *id* as a tuple ( *id*, *object
generation* ) where *object generation* is the registry generation
value when the *id* was allocated.

When an object is destroyed, its *id* value may be reallocated to a
different object. Because the protocol is asynchronous, the
object *id* alone is not sufficient to uniquely identify objects.

The server looks up objects based on a tuple ( *id*, *generation* ) as
follows:

1. Look up the \ref pw_global based on the *id*.

2. The lookup fails if there is no global for the *id*, or
   if *generation* < *object generation*.

The protocol message generally contains only the object *id*.  The
registry generation part is passed around as follows:

1. The server sends the current *registry generation* to clients in the
   protocol footer, if it has changed.

2. The clients keep track of the latest registry generation of the
   messages they have processed. This is the *client generation*.

3. Each client sends their *client generation* in the protocol footer
   of the next message to the server, if its value has changed.

4. The server keeps track for each client the *client generation* they
   have sent back.

5. In each protocol message received from client, the server considers
   each object *id* as tuple ( *id*, *client generation* ).

This allows the server to know if the object *id* the client refers to was
already destroyed, but the client has not yet processed the message
indicating that the *id* is gone. The server indicates failed lookups
of this type with error code ESTALE to the client.

If a client has not sent any *client generation* updates to the
server, then the server will not do any registry generation checks in
object lookups.  This is for backward compatibility only.

The registry generation is an internal detail of the server
implementation and the native protocol, and is not visible to clients
in the PipeWire API.  To identify objects uniquely, clients can use
\ref PW_KEY_OBJECT_SERIAL, which are unique for objects and not
reused, unlike \ref PW_KEY_OBJECT_ID

\see \ref PW_KEY_OBJECT_SERIAL

*/