ESP-NETIF

The purpose of ESP-NETIF library is twofold:

It provides an abstraction layer for the application on top of the TCP/IP stack. This will allow applications to choose between IP stacks in the future.
The APIs it provides are thread safe, even if the underlying TCP/IP stack APIs are not.

ESP-IDF currently implements ESP-NETIF for the lwIP TCP/IP stack only. However, the adapter itself is TCP/IP implementation agnostic and different implementations are possible.

Some ESP-NETIF API functions are intended to be called by application code, for example to get/set interface IP addresses, configure DHCP. Other functions are intended for internal ESP-IDF use by the network driver layer.

In many cases, applications do not need to call ESP-NETIF APIs directly as they are called from the default network event handlers.

ESP-NETIF architecture

                       |          (A) USER CODE                 |
                       |                                        |
      .................| init          settings      events     |
      .                +----------------------------------------+
      .                   .                |           *
      .                   .                |           *
  --------+            +===========================+   *     +-----------------------+
          |            | new/config     get/set    |   *     |                       |
          |            |                           |...*.....| init                  |
          |            |---------------------------|   *     |                       |
    init  |            |                           |****     |                       |
    start |************|  event handler            |*********|  DHCP                 |
    stop  |            |                           |         |                       |
          |            |---------------------------|         |                       |
          |            |                           |         |    NETIF              |
    +-----|            |                           |         +-----------------+     |
    | glue|---<----|---|  esp_netif_transmit       |--<------| netif_output    |     |
    |     |        |   |                           |         |                 |     |
    |     |--->----|---|  esp_netif_receive        |-->------| netif_input     |     |
    |     |        |   |                           |         + ----------------+     |
    |     |...<....|...|  esp_netif_free_rx_buffer |...<.....| packet buffer         |
    +-----|     |  |   |                           |         |                       |
          |     |  |   |                           |         |         (D)           |
    (B)   |     |  |   |          (C)              |         +-----------------------+
  --------+     |  |   +===========================+
communication   |  |                                               NETWORK STACK
DRIVER          |  |           ESP-NETIF
                |  |                                         +------------------+
                |  |   +---------------------------+.........| open/close       |
                |  |   |                           |         |                  |
                |  -<--|  l2tap_write              |-----<---|  write           |
                |      |                           |         |                  |
                ---->--|  esp_vfs_l2tap_eth_filter |----->---|  read            |
                       |                           |         |                  |
                       |            (E)            |         +------------------+
                       +---------------------------+
                                                                   USER CODE
                             ESP-NETIF L2 TAP

Data and event flow in the diagram

........ Initialization line from user code to ESP-NETIF and communication driver
--<--->-- Data packets going from communication media to TCP/IP stack and back
******** Events aggregated in ESP-NETIF propagates to driver, user code and network stack
| User settings and runtime configuration

ESP-NETIF interaction

A) User code, boiler plate

Overall application interaction with a specific IO driver for communication media and configured TCP/IP network stack is abstracted using ESP-NETIF APIs and outlined as below:

Initialization code

Initializes IO driver

Creates a new instance of ESP-NETIF and configure with

ESP-NETIF specific options (flags, behaviour, name)

Network stack options (netif init and input functions, not publicly available)

IO driver specific options (transmit, free rx buffer functions, IO driver handle)

Attaches the IO driver handle to the ESP-NETIF instance created in the above steps

Configures event handlers

use default handlers for common interfaces defined in IO drivers; or define a specific handlers for customised behaviour/new interfaces

register handlers for app related events (such as IP lost/acquired)

Interaction with network interfaces using ESP-NETIF API

Getting and setting TCP/IP related parameters (DHCP, IP, etc)

Receiving IP events (connect/disconnect)

Controlling application lifecycle (set interface up/down)

B) Communication driver, IO driver, media driver

Communication driver plays these two important roles in relation with ESP-NETIF:

Event handlers: Define behaviour patterns of interaction with ESP-NETIF (for example: ethernet link-up -> turn netif on)
Glue IO layer: Adapts the input/output functions to use ESP-NETIF transmit, receive and free receive buffer

Installs driver_transmit to appropriate ESP-NETIF object, so that outgoing packets from network stack are passed to the IO driver

Calls esp_netif_receive() to pass incoming data to network stack

C) ESP-NETIF

ESP-NETIF is an intermediary between an IO driver and a network stack, connecting packet data path between these two. As that it provides a set of interfaces for attaching a driver to ESP-NETIF object (runtime) and configuring a network stack (compile time). In addition to that a set of API is provided to control network interface lifecycle and its TCP/IP properties. As an overview, the ESP-NETIF public interface could be divided into these 6 groups:

Initialization APIs (to create and configure ESP-NETIF instance)
Input/Output API (for passing data between IO driver and network stack)
Event or Action API

Used for network interface lifecycle management

ESP-NETIF provides building blocks for designing event handlers

Setters and Getters for basic network interface properties
Network stack abstraction: enabling user interaction with TCP/IP stack

Set interface up or down

DHCP server and client API

DNS API

Driver conversion utilities

D) Network stack

Network stack has no public interaction with application code with regard to public interfaces and shall be fully abstracted by ESP-NETIF API.

E) ESP-NETIF L2 TAP Interface

The ESP-NETIF L2 TAP interface is ESP-IDF mechanism utilized to access Data Link Layer (L2 per OSI/ISO) for frame reception and transmission from user application. Its typical usage in embedded world might be implementation of non-IP related protocols such as PTP, Wake on LAN and others. Note that only Ethernet (IEEE 802.3) is currently supported.

From user perspective, the ESP-NETIF L2 TAP interface is accessed using file descriptors of VFS which provides a file-like interfacing (using functions like open(), read(), write(), etc). Refer to 虚拟文件系统组件 to learn more.

There is only one ESP-NETIF L2 TAP interface device (path name) available. However multiple file descriptors with different configuration can be opened at a time since the ESP-NETIF L2 TAP interface can be understood as generic entry point to Layer 2 infrastructure. Important is then specific configuration of particular file descriptor. It can be configured to give an access to specific Network Interface identified by if_key (e.g. ETH_DEF) and to filter only specific frames based on their type (e.g. Ethernet type in case of IEEE 802.3). Filtering only specific frames is crucial since the ESP-NETIF L2 TAP needs to exist along with IP stack and so the IP related traffic (IP, ARP, etc.) should not be passed directly to the user application. Even though such option is still configurable, it is not recommended in standard use cases. Filtering is also advantageous from a perspective the user’s application gets access only to frame types it is interested in and the remaining traffic is either passed to other L2 TAP file descriptors or to IP stack.

ESP-NETIF L2 TAP Interface Usage Manual

Initialization

To be able to use the ESP-NETIF L2 TAP interface, it needs to be enabled in Kconfig by CONFIG_ESP_NETIF_L2_TAP first and then registered by esp_vfs_l2tap_intf_register() prior usage of any VFS function.

open()

Once the ESP-NETIF L2 TAP is registered, it can be opened at path name “/dev/net/tap”. The same path name can be opened multiple times up to CONFIG_ESP_NETIF_L2_TAP_MAX_FDS and multiple file descriptors with with different configuration may access the Data Link Layer frames.

The ESP-NETIF L2 TAP can be opened with O_NONBLOCK file status flag to the read() does not block. Note that the write() may block in current implementation when accessing a Network interface since it is a shared resource among multiple ESP-NETIF L2 TAP file descriptors and IP stack, and there is currently no queuing mechanism deployed. The file status flag can be retrieved and modified using fcntl().

On success, open() returns the new file descriptor (a nonnegative integer). On error, -1 is returned and errno is set to indicate the error.

ioctl()

The newly opened ESP-NETIF L2 TAP file descriptor needs to be configured prior its usage since it is not bounded to any specific Network Interface and no frame type filter is configured. The following configuration options are available to do so:

L2TAP_S_INTF_DEVICE - bounds the file descriptor to specific Network Interface which is identified by its if_key. ESP-NETIF Network Interface if_key is passed to ioctl() as the third parameter. Note that default Network Interfaces if_key’s used in ESP-IDF can be found in esp_netif/include/esp_netif_defaults.h.

L2TAP_S_DEVICE_DRV_HNDL - is other way how to bound the file descriptor to specific Network Interface. In this case the Network interface is identified directly by IO Driver handle (e.g. esp_eth_handle_t in case of Ethernet). The IO Driver handle is passed to ioctl() as the third parameter.

L2TAP_S_RCV_FILTER - sets the filter to frames with this type to be passed to the file descriptor. In case of Ethernet frames, the frames are to be filtered based on Length/Ethernet type field. In case the filter value is set less than or equal to 0x05DC, the Ethernet type field is considered to represent IEEE802.3 Length Field and all frames with values in interval <0, 0x05DC> at that field are to be passed to the file descriptor. The IEEE802.2 logical link control (LLC) resolution is then expected to be performed by user’s application. In case the filter value is set greater than 0x05DC, the Ethernet type field is considered to represent protocol identification and only frames which are equal to the set value are to be passed to the file descriptor.

All above set configuration options have getter counterpart option to read the current settings.

警告

The file descriptor needs to be firstly bounded to specific Network Interface by L2TAP_S_INTF_DEVICE or L2TAP_S_DEVICE_DRV_HNDL to be L2TAP_S_RCV_FILTER option available.

备注

VLAN tagged frames are currently not recognized. If user needs to process VLAN tagged frames, they need set filter to be equal to VLAN tag (i.e. 0x8100 or 0x88A8) and process the VLAN tagged frames in user application.

备注

L2TAP_S_DEVICE_DRV_HNDL is particularly useful when user’s application does not require usage of IP stack and so ESP-NETIF is not required to be initialized too. As a result, Network Interface cannot be identified by its if_key and hence it needs to be identified directly by its IO Driver handle.

On success, ioctl() returns 0. On error, -1 is returned, and errno is set to indicate the error.
EBADF - not a valid file descriptor.
EACCES - option change is denied in this state (e.g. file descriptor has not be bounded to Network interface yet).
EINVAL - invalid configuration argument. Ethernet type filter is already used by other file descriptor on that same Network interface.
ENODEV - no such Network Interface which is tried to be assigned to the file descriptor exists.
ENOSYS - unsupported operation, passed configuration option does not exists.

fcntl()

fcntl() is used to manipulate with properties of opened ESP-NETIF L2 TAP file descriptor.

The following commands manipulate the status flags associated with file descriptor:

F_GETFD - the function returns the file descriptor flags, the third argument is ignored.

F_SETFD - sets the file descriptor flags to the value specified by the third argument. Zero is returned.

On error, -1 is returned, and errno is set to indicate the error.
EBADF - not a valid file descriptor.
ENOSYS - unsupported command.

read()

Opened and configured ESP-NETIF L2 TAP file descriptor can be accessed by read() to get inbound frames. The read operation can be either blocking or non-blocking based on actual state of O_NONBLOCK file status flag. When the file status flag is set blocking, the read operation waits until a frame is received and context is switched to other task. When the file status flag is set non-blocking, the read operation returns immediately. In such case, either a frame is returned if it was already queued or the function indicates the queue is empty. The number of queued frames associated with one file descriptor is limited by CONFIG_ESP_NETIF_L2_TAP_RX_QUEUE_SIZE Kconfig option. Once the number of queued frames reach configured threshold, the newly arriving frames are dropped until the queue has enough room to accept incoming traffic (Tail Drop queue management).

On success, read() returns the number of bytes read. Zero is returned when size of the destination buffer is 0. On error, -1 is returned, and errno is set to indicate the error.
EBADF - not a valid file descriptor.
EAGAIN - the file descriptor has been marked non-blocking (O_NONBLOCK), and the read would block.

write()

A raw Data Link Layer frame can be sent to Network Interface via opened and configured ESP-NETIF L2 TAP file descriptor. User’s application is responsible to construct the whole frame except for fields which are added automatically by the physical interface device. The following fields need to be constructed by the user’s application in case of Ethernet link: source/destination MAC addresses, Ethernet type, actual protocol header and user data. See below for more information about Ethernet frame structure.

+-------------------+-------------------+-------------+-------------------------------     --+
|  Destination MAC  |     Source MAC    | Type/Length | Payload (protocol header/data) ...   |
+-------------------+-------------------+-------------+-------------------------------     --+
        6B                   6B                2B                 0-1486B

In other words, there is no additional frame processing performed by the ESP-NETIF L2 TAP interface. It only checks the Ethernet type of the frame is the same as the filter configured in the file descriptor. If the Ethernet type is different, an error is returned and the frame is not sent. Note that the write() may block in current implementation when accessing a Network interface since it is a shared resource among multiple ESP-NETIF L2 TAP file descriptors and IP stack, and there is currently no queuing mechanism deployed.

On success, write() returns the number of bytes written. Zero is returned when size of the input buffer is 0. On error, -1 is returned, and errno is set to indicate the error.
EBADF - not a valid file descriptor.
EBADMSG - Ethernet type of the frame is different then file descriptor configured filter.
EIO - Network interface not available or busy.

close()

Opened ESP-NETIF L2 TAP file descriptor can be closed by the close() to free its allocated resources. The ESP-NETIF L2 TAP implementation of close() may block. On the other hand, it is thread safe and can be called from different task than the file descriptor is actually used. If such situation occurs and one task is blocked in I/O operation and another task tries to close the file descriptor, the first task is unblocked. The first’s task read operation then ends with error.

On success, close() returns zero. On error, -1 is returned, and errno is set to indicate the error.

EBADF - not a valid file descriptor.

select()

Select is used in a standard way, just CONFIG_VFS_SUPPORT_SELECT needs to be enabled to be the select() function available.

ESP-NETIF programmer’s manual

Please refer to the example section for basic initialization of default interfaces:

WiFi Station: wifi/getting_started/station/main/station_example_main.c
Ethernet: ethernet/basic/main/ethernet_example_main.c
L2 TAP: protocols/l2tap/main/l2tap_main.c

WiFi Access Point: wifi/getting_started/softAP/main/softap_example_main.c

For more specific cases please consult this guide: ESP-NETIF Custom I/O Driver.

WiFi default initialization

The initialization code as well as registering event handlers for default interfaces, such as softAP and station, are provided in separate APIs to facilitate simple startup code for most applications:

esp_netif_create_default_wifi_sta()

esp_netif_create_default_wifi_ap()

Please note that these functions return the esp_netif handle, i.e. a pointer to a network interface object allocated and configured with default settings, which as a consequence, means that:

The created object has to be destroyed if a network de-initialization is provided by an application using esp_netif_destroy_default_wifi().
These default interfaces must not be created multiple times, unless the created handle is deleted using esp_netif_destroy().

When using Wifi in AP+STA mode, both these interfaces has to be created.

API Reference

Header File

components/esp_netif/include/esp_netif.h

Functions

esp_err_t esp_netif_init(void)

Initialize the underlying TCP/IP stack.

备注

This function should be called exactly once from application code, when the application starts up.

返回

ESP_OK on success
ESP_FAIL if initializing failed

esp_err_t esp_netif_deinit(void)

Deinitialize the esp-netif component (and the underlying TCP/IP stack)

     Note: Deinitialization is not supported yet

返回

ESP_ERR_INVALID_STATE if esp_netif not initialized
ESP_ERR_NOT_SUPPORTED otherwise

esp_netif_t *esp_netif_new(const esp_netif_config_t *esp_netif_config)

Creates an instance of new esp-netif object based on provided config.

参数

esp_netif_config – pointer esp-netif configuration

返回

pointer to esp-netif object on success
NULL otherwise

void esp_netif_destroy(esp_netif_t *esp_netif)

Destroys the esp_netif object.

参数: esp_netif – [in] pointer to the object to be deleted

esp_err_t esp_netif_set_driver_config(esp_netif_t *esp_netif, const esp_netif_driver_ifconfig_t *driver_config)

Configures driver related options of esp_netif object.

参数

esp_netif – [inout] pointer to the object to be configured
driver_config – [in] pointer esp-netif io driver related configuration

返回

ESP_OK on success
ESP_ERR_ESP_NETIF_INVALID_PARAMS if invalid parameters provided

esp_err_t esp_netif_attach(esp_netif_t *esp_netif, esp_netif_iodriver_handle driver_handle)

Attaches esp_netif instance to the io driver handle.

Calling this function enables connecting specific esp_netif object with already initialized io driver to update esp_netif object with driver specific configuration (i.e. calls post_attach callback, which typically sets io driver callbacks to esp_netif instance and starts the driver)

参数

esp_netif – [inout] pointer to esp_netif object to be attached
driver_handle – [in] pointer to the driver handle

返回

ESP_OK on success
ESP_ERR_ESP_NETIF_DRIVER_ATTACH_FAILED if driver’s pot_attach callback failed

esp_err_t esp_netif_receive(esp_netif_t *esp_netif, void *buffer, size_t len, void *eb)

Passes the raw packets from communication media to the appropriate TCP/IP stack.

This function is called from the configured (peripheral) driver layer. The data are then forwarded as frames to the TCP/IP stack.

参数

esp_netif – [in] Handle to esp-netif instance
buffer – [in] Received data
len – [in] Length of the data frame
eb – [in] Pointer to internal buffer (used in Wi-Fi driver)

返回

ESP_OK

void esp_netif_action_start(void *esp_netif, esp_event_base_t base, int32_t event_id, void *data)

Default building block for network interface action upon IO driver start event Creates network interface, if AUTOUP enabled turns the interface on, if DHCPS enabled starts dhcp server.

备注