channel - Bidirectional interprocess communication
A channel is a bidirectional transport of messages consisting of some amount of byte data and some number of handles.
Channels maintain an ordered queue of messages to be delivered in either direction. A message consists of some amount of data and some number of handles. A call to mx_channel_write() enqueues one message, and a call to mx_channel_read() dequeues one message (if any are queued). A thread can block until messages are pending via mx_object_wait_one() or other waiting mechanisms.
Alternatively, a call to mx_channel_call() enqueues a message in one direction of the channel, waits for a corresponding response, and dequeues the response message. In call mode, corresponding responses are identified via the first 4 bytes of the message, called the transaction ID. Coming up with distinct transaction IDs is up to the users of mx_channel_call().
The process of sending a message via a channel has two steps. The first is to atomically write the data into the channel and move ownership of all handles in the message into this channel. This operation cannot partially succeed: at the end of the call, all handles are either still in the calling process‘s handle table or are all in the channel. The second operation is similar: after a channel read, all the handles in the next message to read are either atomically moved into the process’s handle table, all remain in the channel, or are discarded (only when the MX_CHANNEL_READ_MAY_DISCARD option is given).
Unlike many other kernel object types, channels are not duplicatable. Thus there is only ever one handle associated to a handle endpoint.
Because of these properties (that channel messages move their handle contents atomically, and that channels are not duplicatable), the kernel is able to avoid complicated garbage collection, lifetime management, or cycle detection simply by enforcing the simple rule that a channel handle may not be written into itself.