BTstack is a modular dual-mode Bluetooth stack, supporting both Bluetooth Basic Rate/Enhanced Date Rate (BR/EDR) as well as Bluetooth Low Energy (LE). The BR/EDR technology, also known as Classic Bluetooth, provides a robust wireless connection between devices designed for high data rates. In contrast, the LE technology has a lower throughput but also lower energy consumption, faster connection setup, and the ability to connect to more devices in parallel.

Whether Classic or LE, a Bluetooth device implements one or more Bluetooth profiles. A Bluetooth profile specifies how one or more Bluetooth protocols are used to achieve its goals. For example, every Bluetooth device must implement the Generic Access Profile (GAP), which defines how devices find each other and how they establish a connection. This profile mainly make use of the Host Controller Interface (HCI) protocol, the lowest protocol in the stack hierarchy which implements a command interface to the Bluetooth chipset.

In addition to GAP, a popular Classic Bluetooth example would be a peripheral devices that can be connected via the Serial Port Profile (SPP). SPP basically specifies that a compatible device should provide a Service Discovery Protocol (SDP) record containing an RFCOMM channel number, which will be used for the actual communication.

Similarly, for every LE device, the Generic Attribute Profile (GATT) profile must be implemented in addition to GAP. GATT is built on top of the Attribute Protocol (ATT), and defines how one device can interact with GATT Services on a remote device.

So far, the most popular use of BTstack is in peripheral devices that can be connected via SPP (Android 2.0 or higher) and GATT (Android 4.3 or higher, and iOS 5 or higher). If higher data rates are required between a peripheral and iOS device, the iAP1 and iAP2 protocols of the Made for iPhone program can be used instead of GATT. Please contact us directly for information on BTstack and MFi.

Figure below depicts Bluetooth protocols and profiles that are currently implemented by BTstack. In the following, we first explain how the various Bluetooth protocols are used in BTstack. In the next chapter, we go over the profiles.

Architecture of a BTstack-based application.

HCI - Host Controller Interface

The HCI protocol provides a command interface to the Bluetooth chipset. In BTstack, the HCI implementation also keeps track of all active connections and handles the fragmentation and re-assembly of higher layer (L2CAP) packets.

Please note, that an application rarely has to send HCI commands on its own. Instead, BTstack provides convenience functions in GAP and higher level protocols that use HCI automatically. E.g. to set the name, you call gap_set_local_name() before powering up. The main use of HCI commands in application is during the startup phase to configure special features that are not available via the GAP API yet. How to send a custom HCI command is explained in the following section.

Defining custom HCI command templates

Each HCI command is assigned a 2-byte OpCode used to uniquely identify different types of commands. The OpCode parameter is divided into two fields, called the OpCode Group Field (OGF) and OpCode Command Field (OCF), see Bluetooth Specification - Core Version 4.0, Volume 2, Part E, Chapter 5.4.

Listing below shows the OGFs provided by BTstack in file src/hci.h:

#define OGF_LINK_CONTROL  0x01
#define OGF_LINK_POLICY  0x02
#define OGF_CONTROLLER_BASEBAND  0x03
#define OGF_INFORMATIONAL_PARAMETERS 0x04
#define OGF_LE_CONTROLLER   0x08
#define OGF_BTSTACK  0x3d
#define OGF_VENDOR  0x3f

For all existing Bluetooth commands and their OCFs see Bluetooth Specification - Core Version 4.0, Volume 2, Part E, Chapter 7.

In a HCI command packet, the OpCode is followed by parameter total length, and the actual parameters. The OpCode of a command can be calculated using the OPCODE macro. BTstack provides the hci_cmd_t struct as a compact format to define HCI command packets, see Listing below, and src/hci_cmd.h file in the source code.

// Calculate combined ogf/ocf value.
#define OPCODE(ogf, ocf) (ocf | ogf << 10)

// Compact HCI Command packet description.
typedef struct {
    uint16_t    opcode;
    const char *format;
} hci_cmd_t;

Listing below illustrates the hci_write_local_name HCI command template from library:

// Sets local Bluetooth name
const hci_cmd_t hci_write_local_name = {
    OPCODE(OGF_CONTROLLER_BASEBAND, 0x13), "N"
    // Local name (UTF-8, Null Terminated, max 248 octets)
};

It uses OGF_CONTROLLER_BASEBAND as OGF, 0x13 as OCF, and has one parameter with format “N” indicating a null terminated UTF-8 string. Table below lists the format specifiers supported by BTstack. Check for other predefined HCI commands and info on their parameters.


Format Specifier Description 1,2,3,4 one to four byte value A 31 bytes advertising data B Bluetooth Baseband Address D 8 byte data block E Extended Inquiry Information 240 octets H HCI connection handle N Name up to 248 chars, UTF8 string, null terminated P 16 byte Pairing code, e.g. PIN code or link key S Service Record (Data Element Sequence)


Table: Supported Format Specifiers of HCI Command Parameter.

Sending HCI command based on a template

You can use the hci_send_cmd function to send HCI command based on a template and a list of parameters. However, it is necessary to check that the outgoing packet buffer is empty and that the Bluetooth module is ready to receive the next command - most modern Bluetooth modules only allow to send a single HCI command. This can be done by calling hci_can_send_command_packet_now() function, which returns true, if it is ok to send.

Listing below illustrates how to manually set the device name with the HCI Write Local Name command.

if (hci_can_send_command_packet_now()){
    hci_send_cmd(&hci_write_local_name, "BTstack Demo");
}

Please note, that an application rarely has to send HCI commands on its own. Instead, BTstack provides convenience functions in GAP and higher level protocols that use HCI automatically.

The L2CAP protocol supports higher level protocol multiplexing and packet fragmentation. It provides the base for the RFCOMM and BNEP protocols. For all profiles that are officially supported by BTstack, L2CAP does not need to be used directly. For testing or the development of custom protocols, it’s helpful to be able to access and provide L2CAP services however.

Access an L2CAP service on a remote device

L2CAP is based around the concept of channels. A channel is a logical connection on top of a baseband connection. Each channel is bound to a single protocol in a many-to-one fashion. Multiple channels can be bound to the same protocol, but a channel cannot be bound to multiple protocols. Multiple channels can share the same baseband connection.

To communicate with an L2CAP service on a remote device, the application on a local Bluetooth device initiates the L2CAP layer using the l2cap_init function, and then creates an outgoing L2CAP channel to the PSM of a remote device using the l2cap_create_channel function. The l2cap_create_channel function will initiate a new baseband connection if it does not already exist. The packet handler that is given as an input parameter of the L2CAP create channel function will be assigned to the new outgoing L2CAP channel. This handler receives the L2CAP_EVENT_CHANNEL_OPENED and L2CAP_EVENT_CHANNEL_CLOSED events and L2CAP data packets, as shown in Listing below.

btstack_packet_handler_t l2cap_packet_handler;

void l2cap_packet_handler(uint8_t packet_type, uint16_t channel, uint8_t *packet, uint16_t size){
    bd_addr_t event_addr;
    switch (packet_type){
        case HCI_EVENT_PACKET:
            switch (hci_event_packet_get_type(packet)){
                case L2CAP_EVENT_CHANNEL_OPENED:
                    l2cap_event_channel_opened_get_address(packet, &event_addr);
                    psm       = l2cap_event_channel_opened_get_psm(packet); 
                    local_cid = l2cap_event_channel_opened_get_local_cid(packet); 
                    handle    = l2cap_event_channel_opened_get_handle(packet);
                    if (l2cap_event_channel_opened_get_status(packet)) {
                        printf("Connection failed\n\r");
                    } else 
                        printf("Connected\n\r");
                    }
                    break;
                case L2CAP_EVENT_CHANNEL_CLOSED:
                    break;
                    ...
            }
        case L2CAP_DATA_PACKET:
            // handle L2CAP data packet
            break;
        ...
    }
}

void create_outgoing_l2cap_channel(bd_addr_t address, uint16_t psm, uint16_t mtu){
     l2cap_create_channel(NULL, l2cap_packet_handler, remote_bd_addr, psm, mtu);
}

void btstack_setup(){
    ...
    l2cap_init();
}

Provide an L2CAP service

To provide an L2CAP service, the application on a local Bluetooth device must init the L2CAP layer and register the service with l2cap_register_service. From there on, it can wait for incoming L2CAP connections. The application can accept or deny an incoming connection by calling the l2cap_accept_connection and l2cap_deny_connection functions respectively.

If a connection is accepted and the incoming L2CAP channel gets successfully opened, the L2CAP service can send and receive L2CAP data packets to the connected device with l2cap_send.

Listing below provides L2CAP service example code.

void packet_handler (uint8_t packet_type, uint16_t channel, uint8_t *packet, uint16_t size){
    bd_addr_t event_addr;
    switch (packet_type){
        case HCI_EVENT_PACKET:
            switch (hci_event_packet_get_type(packet)){
                case L2CAP_EVENT_INCOMING_CONNECTION:
                    local_cid = l2cap_event_incoming_connection_get_local_cid(packet); 
                    l2cap_accept_connection(local_cid);
                    break;
                case L2CAP_EVENT_CHANNEL_OPENED:
                    l2cap_event_channel_opened_get_address(packet, &event_addr);
                    psm       = l2cap_event_channel_opened_get_psm(packet); 
                    local_cid = l2cap_event_channel_opened_get_local_cid(packet); 
                    handle    = l2cap_event_channel_opened_get_handle(packet);
                    if (l2cap_event_channel_opened_get_status(packet)) {
                        printf("Connection failed\n\r");
                    } else 
                        printf("Connected\n\r");
                    }
                    break;
                case L2CAP_EVENT_CHANNEL_CLOSED:
                    break;
                    ...
            }
        case L2CAP_DATA_PACKET:
            // handle L2CAP data packet
            break;
        ...
    }
}

void btstack_setup(){
    ...
    l2cap_init();
    l2cap_register_service(NULL, packet_handler, 0x11,100);
}

Sending L2CAP Data

Sending of L2CAP data packets may fail due to a full internal BTstack outgoing packet buffer, or if the ACL buffers in the Bluetooth module become full, i.e., if the application is sending faster than the packets can be transferred over the air.

Instead of directly calling l2cap_send, it is recommended to call l2cap_request_can_send_now_event(cahnnel_id) which will trigger an L2CAP_EVENT_CAN_SEND_NOW as soon as possible. It might happen that the event is received via packet handler before the l2cap_request_can_send_now_event function returns. The L2CAP_EVENT_CAN_SEND_NOW indicates a channel ID on which sending is possible.

Please note that the guarantee that a packet can be sent is only valid when the event is received. After returning from the packet handler, BTstack might need to send itself.

LE Data Channels

The full title for LE Data Channels is actually LE Connection-Oriented Channels with LE Credit-Based Flow-Control Mode. In this mode, data is sent as Service Data Units (SDUs) that can be larger than an individual HCI LE ACL packet.

LE Data Channels are similar to Classic L2CAP Channels but also provide a credit-based flow control similar to RFCOMM Channels. Unless the LE Data Packet Extension of Bluetooth Core 4.2 specification is used, the maximum packet size for LE ACL packets is 27 bytes. In order to send larger packets, each packet will be split into multiple ACL LE packets and recombined on the receiving side.

Since multiple SDUs can be transmitted at the same time and the individual ACL LE packets can be sent interleaved, BTstack requires a dedicated receive buffer per channel that has to be passed when creating the channel or accepting it. Similarly, when sending SDUs, the data provided to the l2cap_cbm_send_data must stay valid until the L2CAP_EVENT_LE_PACKET_SENT is received.

When creating an outgoing connection of accepting an incoming, the initial_credits allows to provide a fixed number of credits to the remote side. Further credits can be provided anytime with l2cap_cbm_provide_credits. If L2CAP_LE_AUTOMATIC_CREDITS is used, BTstack automatically provides credits as needed - effectively trading in the flow-control functionality for convenience.

The remainder of the API is similar to the one of L2CAP:

RFCOMM - Radio Frequency Communication Protocol

The Radio frequency communication (RFCOMM) protocol provides emulation of serial ports over the L2CAP protocol and reassembly. It is the base for the Serial Port Profile and other profiles used for telecommunication like Head-Set Profile, Hands-Free Profile, Object Exchange (OBEX) etc.

No RFCOMM packet boundaries

As RFCOMM emulates a serial port, it does not preserve packet boundaries.

On most operating systems, RFCOMM/SPP will be modeled as a pipe that allows to write a block of bytes. The OS and the Bluetooth Stack are free to buffer and chunk this data in any way it seems fit. In your BTstack application, you will therefore receive this data in the same order, but there are no guarantees as how it might be fragmented into multiple chunks.

If you need to preserve the concept of sending a packet with a specific size over RFCOMM, the simplest way is to prefix the data with a 2 or 4 byte length field and then reconstruct the packet on the receiving side.

Please note, that due to BTstack's 'no buffers' policy, BTstack will send outgoing RFCOMM data immediately and implicitly preserve the packet boundaries, i.e., it will send the data as a single RFCOMM packet in a single L2CAP packet, which will arrive in one piece. While this will hold between two BTstack instances, it's not a good idea to rely on implementation details and rather prefix the data as described.

RFCOMM flow control

RFCOMM has a mandatory credit-based flow-control. This means that two devices that established RFCOMM connection, use credits to keep track of how many more RFCOMM data packets can be sent to each. If a device has no (outgoing) credits left, it cannot send another RFCOMM packet, the transmission must be paused. During the connection establishment, initial credits are provided. BTstack tracks the number of credits in both directions. If no outgoing credits are available, the RFCOMM send function will return an error, and you can try later. For incoming data, BTstack provides channels and services with and without automatic credit management via different functions to create/register them respectively. If the management of credits is automatic, the new credits are provided when needed relying on ACL flow control - this is only useful if there is not much data transmitted and/or only one physical connection is used. If the management of credits is manual, credits are provided by the application such that it can manage its receive buffers explicitly.

Access an RFCOMM service on a remote device

To communicate with an RFCOMM service on a remote device, the application on a local Bluetooth device initiates the RFCOMM layer using the rfcomm_init function, and then creates an outgoing RFCOMM channel to a given server channel on a remote device using the rfcomm_create_channel function. The rfcomm_create_channel function will initiate a new L2CAP connection for the RFCOMM multiplexer, if it does not already exist. The channel will automatically provide enough credits to the remote side. To provide credits manually, you have to create the RFCOMM connection by calling rfcomm_create_channel_with_initial_credits - see Section on manual credit assignement.

The packet handler that is given as an input parameter of the RFCOMM create channel function will be assigned to the new outgoing channel. This handler receives the RFCOMM_EVENT_CHANNEL_OPENED and RFCOMM_EVENT_CHANNEL_CLOSED events, and RFCOMM data packets, as shown in Listing below.

void rfcomm_packet_handler(uint8_t packet_type, uint16_t channel, uint8_t *packet, uint16_t size){
    switch (packet_type){
        case HCI_EVENT_PACKET:
            switch (hci_event_packet_get_type(packet)){
                case RFCOMM_EVENT_CHANNEL_OPENED:
                    if (rfcomm_event_open_channel_complete_get_status(packet)) {
                        printf("Connection failed\n\r");
                    } else {
                        printf("Connected\n\r");
                    }
                    break;
                case RFCOMM_EVENT_CHANNEL_CLOSED:
                    break;
                ...
            }
            break;
        case RFCOMM_DATA_PACKET:
            // handle RFCOMM data packets
            return;
    }
}

void create_rfcomm_channel(uint8_t packet_type, uint8_t *packet, uint16_t size){
    rfcomm_create_channel(rfcomm_packet_handler, addr, rfcomm_channel);
}

void btstack_setup(){
    ...
    l2cap_init();
    rfcomm_init();
}

The RFCOMM channel will stay open until either side closes it with rfcomm_disconnect. BTstack will close the multiplexer after 60 seconds without an active RFCOMM channel.

Provide an RFCOMM service

To provide an RFCOMM service, the application on a local Bluetooth device must first init the L2CAP and RFCOMM layers and then register the service with rfcomm_register_service. From there on, it can wait for incoming RFCOMM connections. The application can accept or deny an incoming connection by calling the rfcomm_accept_connection and rfcomm_deny_connection functions respectively. If a connection is accepted and the incoming RFCOMM channel gets successfully opened, the RFCOMM service can send RFCOMM data packets to the connected device with rfcomm_send and receive data packets by the packet handler provided by the rfcomm_register_service call.

Listing below provides the RFCOMM service example code.

void packet_handler(uint8_t packet_type, uint16_t channel, uint8_t *packet, uint16_t size){
    switch (packet_type){
        case HCI_EVENT_PACKET:
            switch (hci_event_packet_get_type(packet)){
                case RFCOMM_EVENT_INCOMING_CONNECTION:
                    rfcomm_channel_id = rfcomm_event_incoming_connection_get_rfcomm_cid(packet);
                    rfcomm_accept_connection(rfcomm_channel_id);
                    break;
                case RFCOMM_EVENT_CHANNEL_OPENED:
                    if (rfcomm_event_open_channel_complete_get_status(packet)){
                        printf("RFCOMM channel open failed.");
                        break;
                    } 
                    rfcomm_channel_id = rfcomm_event_open_channel_complete_get_rfcomm_cid(packet);
                    mtu = rfcomm_event_open_channel_complete_get_max_frame_size(packet);
                    printf("RFCOMM channel open succeeded, max frame size %u.", mtu);
                    break;
                case RFCOMM_EVENT_CHANNEL_CLOSED:
                    printf("Channel closed.");
                    break;
                ...
            }
            break;
        case RFCOMM_DATA_PACKET:
            // handle RFCOMM data packets
            return;
        ...
    }
    ...
}

void btstack_setup(){
    ...
    l2cap_init();
    rfcomm_init();
    rfcomm_register_service(packet_handler, rfcomm_channel_nr, mtu); 
}

Slowing down RFCOMM data reception

RFCOMM’s credit-based flow-control can be used to adapt, i.e., slow down the RFCOMM data to your processing speed. For incoming data, BTstack provides channels and services with and without automatic credit management. If the management of credits is automatic, new credits are provided when needed relying on ACL flow control. This is only useful if there is not much data transmitted and/or only one physical connection is used. See Listing below.

void btstack_setup(void){
    ...
    // init RFCOMM
    rfcomm_init();
    rfcomm_register_service(packet_handler, rfcomm_channel_nr, 100); 
}

If the management of credits is manual, credits are provided by the application such that it can manage its receive buffers explicitly, see Listing below.

Manual credit management is recommended when received RFCOMM data cannot be processed immediately. In the SPP flow control example, delayed processing of received data is simulated with the help of a periodic timer. To provide new credits, you call the rfcomm_grant_credits function with the RFCOMM channel ID and the number of credits as shown in Listing below.

void btstack_setup(void){
    ...
    // init RFCOMM
    rfcomm_init();
    // reserved channel, mtu=100, 1 credit
    rfcomm_register_service_with_initial_credits(packet_handler, rfcomm_channel_nr, 100, 1);  
}

void processing(){
    // process incoming data packet
    ... 
    // provide new credit
    rfcomm_grant_credits(rfcomm_channel_id, 1);
}

Please note that providing single credits effectively reduces the credit-based (sliding window) flow control to a stop-and-wait flow-control that limits the data throughput substantially. On the plus side, it allows for a minimal memory footprint. If possible, multiple RFCOMM buffers should be used to avoid pauses while the sender has to wait for a new credit.

Sending RFCOMM data

Outgoing packets, both commands and data, are not queued in BTstack. This section explains the consequences of this design decision for sending data and why it is not as bad as it sounds.

Independent from the number of output buffers, packet generation has to be adapted to the remote receiver and/or maximal link speed. Therefore, a packet can only be generated when it can get sent. With this assumption, the single output buffer design does not impose additional restrictions. In the following, we show how this is used for adapting the RFCOMM send rate.

When there is a need to send a packet, call rcomm_request_can_send_now and wait for the reception of the RFCOMM_EVENT_CAN_SEND_NOW event to send the packet, as shown in Listing below.

Please note that the guarantee that a packet can be sent is only valid when the event is received. After returning from the packet handler, BTstack might need to send itself.

void prepare_data(uint16_t rfcomm_channel_id){
    ...
    // prepare data in data_buffer
    rfcomm_request_can_send_now_event(rfcomm_channel_id);
}

void send_data(uint16_t rfcomm_channel_id){
    rfcomm_send(rfcomm_channel_id,  data_buffer, data_len);
    // packet is handed over to BTstack, we can prepare the next one
    prepare_data(rfcomm_channel_id);
}

void packet_handler(uint8_t packet_type, uint16_t channel, uint8_t *packet, uint16_t size){
    switch (packet_type){
        case HCI_EVENT_PACKET:
            switch (hci_event_packet_get_type(packet)){
                ...
                case RFCOMM_EVENT_CAN_SEND_NOW:
                    rfcomm_channel_id = rfcomm_event_can_send_now_get_rfcomm_cid(packet);
                    send_data(rfcomm_channel_id);
                    break;
                ...
            }
            ...
        }
    }
}

Optimized sending of RFCOMM data

When sending RFCOMM data via rfcomm_send, BTstack needs to copy the data from the user provided buffer into the outgoing buffer. This requires both an additional buffer for the user data as well requires a copy operation.

To avoid this, it is possible to directly write the user data into the outgoing buffer.

When get the RFCOMM_CAN_SEND_NOW event, you call rfcomm_reserve_packet_buffer to lock the buffer for your send operation. Then, you can ask how many bytes you can send with rfcomm_get_max_frame_size and get a pointer to BTstack's buffer with rfcomm_get_outgoing_buffer. Now, you can fill that buffer and finally send the data with rfcomm_send_prepared.

SDP - Service Discovery Protocol

The SDP protocol allows to announce services and discover services provided by a remote Bluetooth device.

Create and announce SDP records

BTstack contains a complete SDP server and allows to register SDP records. An SDP record is a list of SDP Attribute {ID, Value} pairs that are stored in a Data Element Sequence (DES). The Attribute ID is a 16-bit number, the value can be of other simple types like integers or strings or can itself contain other DES.

To create an SDP record for an SPP service, you can call spp_create_sdp_record from with a pointer to a buffer to store the record, the server channel number, and a record name.

For other types of records, you can use the other functions in, using the data element de_ functions. Listing [sdpCreate] shows how an SDP record containing two SDP attributes can be created. First, a DES is created and then the Service Record Handle and Service Class ID List attributes are added to it. The Service Record Handle attribute is added by calling the de_add_number function twice: the first time to add 0x0000 as attribute ID, and the second time to add the actual record handle (here 0x1000) as attribute value. The Service Class ID List attribute has ID 0x0001, and it requires a list of UUIDs as attribute value. To create the list, de_push_sequence is called, which “opens” a sub-DES. The returned pointer is used to add elements to this sub-DES. After adding all UUIDs, the sub-DES is “closed” with de_pop_sequence.

To register an SDP record, you call sdp_register_service with a pointer to it. The SDP record can be stored in FLASH since BTstack only stores the pointer. Please note that the buffer needs to persist (e.g. global storage, dynamically allocated from the heap or in FLASH) and cannot be used to create another SDP record.

Query remote SDP service

BTstack provides an SDP client to query SDP services of a remote device. The SDP Client API is shown in here. The sdp_client_query function initiates an L2CAP connection to the remote SDP server. Upon connect, a Service Search Attribute request with a Service Search Pattern and a Attribute ID List is sent. The result of the Service Search Attribute query contains a list of Service Records, and each of them contains the requested attributes. These records are handled by the SDP parser. The parser delivers SDP_PARSER_ATTRIBUTE_VALUE and SDP_PARSER_COMPLETE events via a registered callback. The SDP_PARSER_ATTRIBUTE_VALUE event delivers the attribute value byte by byte.

On top of this, you can implement specific SDP queries. For example, BTstack provides a query for RFCOMM service name and channel number. This information is needed, e.g., if you want to connect to a remote SPP service. The query delivers all matching RFCOMM services, including its name and the channel number, as well as a query complete event via a registered callback, as shown in Listing below.

bd_addr_t remote = {0x04,0x0C,0xCE,0xE4,0x85,0xD3};

void packet_handler (void * connection, uint8_t packet_type, uint16_t channel, uint8_t *packet, uint16_t size){
    if (packet_type != HCI_EVENT_PACKET) return;

    uint8_t event = packet[0];
    switch (event) {
        case BTSTACK_EVENT_STATE:
            // bt stack activated, get started 
            if (btstack_event_state_get_state(packet) == HCI_STATE_WORKING){
                  sdp_client_query_rfcomm_channel_and_name_for_uuid(remote, 0x0003);
            }
            break;
        default:
            break;
    }
}

static void btstack_setup(){
   ...
    // init L2CAP
    l2cap_init();
    l2cap_register_packet_handler(packet_handler);
}

void handle_query_rfcomm_event(sdp_query_event_t * event, void * context){
    sdp_client_query_rfcomm_service_event_t * ve;

    switch (event->type){
        case SDP_EVENT_QUERY_RFCOMM_SERVICE:
            ve = (sdp_client_query_rfcomm_service_event_t*) event;
            printf("Service name: '%s', RFCOMM port %u\n", ve->service_name, ve->channel_nr);
            break;
        case SDP_EVENT_QUERY_COMPLETE:
            report_found_services();
            printf("Client query response done with status %d. \n", ce->status);
            break;
    }
}

int main(void){
    hw_setup();
    btstack_setup();

    // register callback to receive matching RFCOMM Services and 
    // query complete event 
    sdp_client_query_rfcomm_register_callback(handle_query_rfcomm_event, NULL);

    // turn on!
    hci_power_control(HCI_POWER_ON);
    // go!
    btstack_run_loop_execute(); 
    return 0;
}

BNEP - Bluetooth Network Encapsulation Protocol

The BNEP protocol is used to transport control and data packets over standard network protocols such as TCP, IPv4 or IPv6. It is built on top of L2CAP, and it specifies a minimum L2CAP MTU of 1691 bytes.

Receive BNEP events

To receive BNEP events, please register a packet handler with bnep_register_packet_handler.

Access a BNEP service on a remote device

To connect to a remote BNEP service, you need to know its UUID. The set of available UUIDs can be queried by a SDP query for the PAN profile. Please see section on PAN profile for details. With the remote UUID, you can create a connection using the bnep_connect function. You’ll receive a BNEP_EVENT_CHANNEL_OPENED on success or failure.

After the connection was opened successfully, you can send and receive Ethernet packets. Before sending an Ethernet frame with bnep_send, bnep_can_send_packet_now needs to return true. Ethernet frames are received via the registered packet handler with packet type BNEP_DATA_PACKET.

BTstack BNEP implementation supports both network protocol filter and multicast filters with bnep_set_net_type_filter and bnep_set_multicast_filter respectively.

Finally, to close a BNEP connection, you can call bnep_disconnect.

Provide BNEP service

To provide a BNEP service, call bnep_register_service with the provided service UUID and a max frame size.

A BNEP_EVENT_INCOMING_CONNECTION event will mark that an incoming connection is established. At this point you can start sending and receiving Ethernet packets as described in the previous section.

Sending Ethernet packets

Similar to L2CAP and RFOMM, directly sending an Ethernet packet via BNEP might fail, if the outgoing packet buffer or the ACL buffers in the Bluetooth module are full.

When there's a need to send an Ethernet packet, call bnep_request_can_send_now and send the packet when the BNEP_EVENT_CAN_SEND_NOW event gets received.

ATT - Attribute Protocol

The ATT protocol is used by an ATT client to read and write attribute values stored on an ATT server. In addition, the ATT server can notify the client about attribute value changes. An attribute has a handle, a type, and a set of properties.

The Generic Attribute (GATT) profile is built upon ATT and provides higher level organization of the ATT attributes into GATT Services and GATT Characteristics. In BTstack, the complete ATT client functionality is included within the GATT Client. See GATT client for more.

On the server side, one ore more GATT profiles are converted ahead of time into the corresponding ATT attribute database and provided by the att_server implementation. The constant data are automatically served by the ATT server upon client request. To receive the dynamic data, such is characteristic value, the application needs to register read and/or write callback. In addition, notifications and indications can be sent. Please see Section on GATT server for more.

SMP - Security Manager Protocol

The SMP protocol allows to setup authenticated and encrypted LE connection. After initialization and configuration, SMP handles security related functions on its own but emits events when feedback from the main app or the user is required. The two main tasks of the SMP protocol are: bonding and identity resolving.

LE Legacy Pairing and LE Secure Connections

The original pairing algorithm introduced in Bluetooth Core V4.0 does not provide security in case of an attacker present during the initial pairing. To fix this, the Bluetooth Core V4.2 specification introduced the new LE Secure Connections method, while referring to the original method as LE Legacy Pairing.

BTstack supports both pairing methods. To enable the more secure LE Secure Connections method, ENABLE_LE_SECURE_CONNECTIONS needs to be defined in btstack_config.h.

LE Secure Connections are based on Elliptic Curve Diffie-Hellman (ECDH) algorithm for the key exchange. On start, a new public/private key pair is generated. During pairing, the Long Term Key (LTK) is generated based on the local keypair and the remote public key. To facilitate the creation of such a keypairs and the calculation of the LTK, the Bluetooth Core V4.2 specification introduced appropriate commands for the Bluetooth controller.

As an alternative for controllers that don't provide these primitives, BTstack provides the relevant cryptographic functions in software via the BSD-2-Clause licensed micro-ecc library. When using using LE Secure Connections, the Peripheral must store LTK in non-volatile memory.

To only allow LE Secure Connections, you can call sm_set_secure_connections_only_mode(true).

Initialization

To activate the security manager, call sm_init().

If you’re creating a product, you should also call sm_set_ir() and sm_set_er() with a fixed random 16 byte number to create the IR and ER key seeds. If possible use a unique random number per device instead of deriving it from the product serial number or something similar. The encryption key generated by the BLE peripheral will be ultimately derived from the ER key seed. See Bluetooth Specification - Bluetooth Core V4.0, Vol 3, Part G, 5.2.2 for more details on deriving the different keys. The IR key is used to identify a device if private, resolvable Bluetooth addresses are used.

Configuration

To receive events from the Security Manager, a callback is necessary. How to register this packet handler depends on your application configuration.

When att_server is used to provide a GATT/ATT service, att_server registers itself as the Security Manager packet handler. Security Manager events are then received by the application via the att_server packet handler.

If att_server is not used, you can directly register your packet handler with the security manager by calling sm_register_packet_handler.

The default SMP configuration in BTstack is to be as open as possible:

You can configure these items by calling following functions respectively:

Identity Resolving

Identity resolving is the process of matching a private, resolvable Bluetooth address to a previously paired device using its Identity Resolving (IR) key. After an LE connection gets established, BTstack automatically tries to resolve the address of this device. During this lookup, BTstack will emit the following events:

and later:

User interaction during Pairing

BTstack will inform the app about pairing via theses events:

Depending on the authentication requirements, IO capabilities, available OOB data, and the enabled STK generation methods, BTstack will request feedback from the app in the form of an event:

To accept Just Works/Numeric Comparison, or provide a Passkey, sm_just_works_confirm or sm_passkey_input can be called respectively. Othwerise, sm_bonding_decline aborts the pairing.

After the bonding process, SM_EVENT_PAIRING_COMPLETE, is emitted. Any active dialog can be closed on this.

Connection with Bonded Devices

During pairing, two devices exchange bonding information, notably a Long-Term Key (LTK) and their respective Identity Resolving Key (IRK). On a subsequent connection, the Securit Manager will use this information to establish an encrypted connection.

To inform about this, the following events are emitted:

The SM_EVENT_REENCRYPTION_COMPLETE with ERROR_CODE_PIN_OR_KEY_MISSING can be caused:

In Peripheral role, pairing will start even in case of an re-encryption error. It might be helpful to inform the user about the lost bonding or reject it right away due to security considerations.

Keypress Notifications

As part of Bluetooth Core V4.2 specification, a device with a keyboard but no display can send keypress notifications to provide better user feedback. In BTstack, the sm_keypress_notification() function is used for sending notifications. Notifications are received by BTstack via the SM_EVENT_KEYPRESS_NOTIFICATION event.

Cross-transport Key Derivation (CTKD) for LE Secure Connections

In a dual-mode configuration, BTstack generates an BR/EDR Link Key from the LE LTK via the Link Key Conversion functions h6 , (or h7 if supported) when ENABLE_CROSS_TRANSPORT_KEY_DERIVATION is defined. The derived key then stored in local LE Device DB.

The main use case for this is connections with smartphones. E.g. iOS provides APIs for LE scanning and connection, but none for BR/EDR. This allows an application to connect and pair with a device and also later setup a BR/EDR connection without the need for the smartphone user to use the system Settings menu.

To derive an LE LTK from a BR/EDR link key, the Bluetooth controller needs to support Secure Connections via NIST P-256 elliptic curves. BTstack does not support LE Secure Connections via LE Transport currently.

Out-of-Band Data with LE Legacy Pairing

LE Legacy Pairing can be made secure by providing a way for both devices to acquire a pre-shared secret 16 byte key by some fancy method. In most cases, this is not an option, especially since popular OS like iOS don’t provide a way to specify it. In some applications, where both sides of a Bluetooth link are developed together, this could provide a viable option.

To provide OOB data, you can register an OOB data callback with sm_register_oob_data_callback.