Struct devices::virtio::video::EosBufferManager

struct EosBufferManager {
    stream_id: u32,
    eos_buffer: Option<u32>,
    client_awaits_eos: bool,
    responses: Vec<VideoEvtResponseType>,
}

Manages the zero-length, EOS-marked buffer signaling the end of a stream.

Both the decoder and encoder need to signal end-of-stream events using a zero-sized buffer marked with the VIRTIO_VIDEO_BUFFER_FLAG_EOS flag. This struct allows to keep a buffer aside for that purpose.

TODO(b/149725148): Remove this when libvda supports buffer flags.

Fields

stream_id: u32

eos_buffer: Option<u32>

client_awaits_eos: bool

responses: Vec<VideoEvtResponseType>

Implementations

impl EosBufferManager

fn new(stream_id: u32) -> Self

Create a new EOS manager for stream stream_id.

fn try_reserve_eos_buffer(&mut self, buffer_id: u32) -> bool

Attempt to reserve buffer buffer_id for use as EOS buffer.

This method should be called by the output buffer queueing code of the device. It returns true if the buffer has been kept aside for EOS, false otherwise (which means another buffer is already kept aside for EOS). If true is returned, the client must not use the buffer for any other purpose.

fn try_complete_eos( &mut self, responses: Vec<VideoEvtResponseType>, ) -> Option<Vec<VideoEvtResponseType>>

Attempt to complete an EOS event using the previously reserved buffer, if available.

responses is a vector of responses to be sent to the driver along with the EOS buffer. If an EOS buffer has been made available using the try_reserve_eos_buffer method, then this method returns the responses vector with the EOS buffer dequeue appended in first position.

If no EOS buffer is available, then the contents of responses is put aside, and will be returned the next time this method is called with an EOS buffer available. When this happens, client_awaits_eos will be set to true, and the client can check this member and call this method again right after queuing the next buffer to obtain the EOS response as soon as is possible.

fn reset(&mut self)

Reset the state of the manager, for use during e.g. stream resets.