.. _envoy_v3_api_file_envoy/extensions/network/socket_interface/v3/default_socket_interface.proto:

Default socket interface configuration (proto)
==============================================






.. _envoy_v3_api_msg_extensions.network.socket_interface.v3.DefaultSocketInterface:

extensions.network.socket_interface.v3.DefaultSocketInterface
-------------------------------------------------------------


:repo:`[extensions.network.socket_interface.v3.DefaultSocketInterface proto] <api/envoy/extensions/network/socket_interface/v3/default_socket_interface.proto#L18>`

Configuration for default socket interface that relies on OS dependent syscall to create
sockets.



.. code-block:: json
  :force:

  {
    "io_uring_options": {...}
  }

.. _envoy_v3_api_field_extensions.network.socket_interface.v3.DefaultSocketInterface.io_uring_options:


io_uring_options
  (:ref:`extensions.network.socket_interface.v3.IoUringOptions <envoy_v3_api_msg_extensions.network.socket_interface.v3.IoUringOptions>`) io_uring options. io_uring is only valid in Linux with at least kernel version 5.11. Otherwise,
  Envoy will fall back to use the default socket API. If not set then io_uring will not be
  enabled.



.. _envoy_v3_api_msg_extensions.network.socket_interface.v3.IoUringOptions:

extensions.network.socket_interface.v3.IoUringOptions
-----------------------------------------------------


:repo:`[extensions.network.socket_interface.v3.IoUringOptions proto] <api/envoy/extensions/network/socket_interface/v3/default_socket_interface.proto#L25>`




.. code-block:: json
  :force:

  {
    "io_uring_size": {...},
    "enable_submission_queue_polling": ...,
    "read_buffer_size": {...},
    "write_timeout_ms": {...}
  }

.. _envoy_v3_api_field_extensions.network.socket_interface.v3.IoUringOptions.io_uring_size:


io_uring_size
  (`UInt32Value <https://developers.google.com/protocol-buffers/docs/reference/google.protobuf#uint32value>`_) The size for io_uring submission queues (SQ). io_uring is built with a fixed size in each
  thread during configuration, and each io_uring operation creates a submission queue
  entry (SQE). The default is 1000.


.. _envoy_v3_api_field_extensions.network.socket_interface.v3.IoUringOptions.enable_submission_queue_polling:


enable_submission_queue_polling
  (`bool <https://developers.google.com/protocol-buffers/docs/proto#scalar>`_) Enable io_uring submission queue polling (SQPOLL). io_uring SQPOLL mode polls all SQEs in the
  SQ in the kernel thread. io_uring SQPOLL mode may reduce latency and increase CPU usage as a
  cost. The default is false.


.. _envoy_v3_api_field_extensions.network.socket_interface.v3.IoUringOptions.read_buffer_size:


read_buffer_size
  (`UInt32Value <https://developers.google.com/protocol-buffers/docs/reference/google.protobuf#uint32value>`_) The size of an io_uring socket's read buffer. Each io_uring read operation will allocate a
  buffer of the given size. If the given buffer is too small, the socket will have read multiple
  times for all the data. The default is 8192.


.. _envoy_v3_api_field_extensions.network.socket_interface.v3.IoUringOptions.write_timeout_ms:


write_timeout_ms
  (`UInt32Value <https://developers.google.com/protocol-buffers/docs/reference/google.protobuf#uint32value>`_) The write timeout of an io_uring socket on closing in ms. io_uring writes and closes
  asynchronously. If the remote stops reading, the io_uring write operation may never complete.
  The operation is canceled and the socket is closed after the timeout. The default is 1000.