In the following, we explain how the various Bluetooth profiles are used in BTstack.

A2DP - Advanced Audio Distribution

The A2DP profile defines how to stream audio over a Bluetooth connection from one device, such as a mobile phone, to another device such as a headset. A device that acts as source of audio stream implements the A2DP Source role. Similarly, a device that receives an audio stream implements the A2DP Sink role. As such, the A2DP service allows uni-directional transfer of an audio stream, from single channel mono, up to two channel stereo. Our implementation includes mandatory support for the low-complexity SBC codec. Signaling for optional codes (FDK AAC, LDAC, APTX) is supported as well, by you need to provide your own codec library.

AVRCP - Audio/Video Remote Control Profile

The AVRCP profile defines how audio playback on a remote device (e.g. a music app on a smartphone) can be controlled as well as how to state changes such as volume, information on currently played media, battery, etc. can be received from a remote device (e.g. a speaker). Usually, each device implements two roles: - The Controller role allows to query information on currently played media, such are title, artist and album, as well as to control the playback, i.e. to play, stop, repeat, etc. - The Target role responds to commands, e.g. playback control, and queries, e.g. playback status, media info, from the Controller currently played media.

GAP - Generic Access Profile: Classic

The GAP profile defines how devices find each other and establish a secure connection for other profiles. As mentioned before, the GAP functionality is split between and . Please check both.

Become discoverable

A remote unconnected Bluetooth device must be set as “discoverable” in order to be seen by a device performing the inquiry scan. To become discoverable, an application can call gap_discoverable_control with input parameter 1. If you want to provide a helpful name for your device, the application can set its local name by calling gap_set_local_name. To save energy, you may set the device as undiscoverable again, once a connection is established. See Listing below for an example.

int main(void){
    ...
    // make discoverable
    gap_discoverable_control(1);
    btstack_run_loop_execute();
    return 0;
}
void packet_handler (uint8_t packet_type, uint8_t *packet, uint16_t size){
     ...
     switch(state){
          case W4_CHANNEL_COMPLETE:
              // if connection is successful, make device undiscoverable
              gap_discoverable_control(0);
          ...
     }
 }

Discover remote devices

To scan for remote devices, the hci_inquiry command is used. Found remote devices are reported as a part of:

Each response contains at least the Bluetooth address, the class of device, the page scan repetition mode, and the clock offset of found device. The latter events add information about the received signal strength or provide the Extended Inquiry Result (EIR). A code snippet is shown in Listing below.

void print_inquiry_results(uint8_t *packet){
    int event = packet[0];
    int numResponses = hci_event_inquiry_result_get_num_responses(packet);
    uint16_t classOfDevice, clockOffset;
    uint8_t  rssi, pageScanRepetitionMode;
    for (i=0; i<numResponses; i++){
        bt_flip_addr(addr, &packet[3+i*6]);
        pageScanRepetitionMode = packet [3 + numResponses*6 + i];
        if (event == HCI_EVENT_INQUIRY_RESULT){
            classOfDevice = little_endian_read_24(packet, 3 + numResponses*(6+1+1+1) + i*3);
            clockOffset =   little_endian_read_16(packet, 3 + numResponses*(6+1+1+1+3) + i*2) & 0x7fff;
            rssi  = 0;
        } else {
            classOfDevice = little_endian_read_24(packet, 3 + numResponses*(6+1+1)     + i*3);
            clockOffset =   little_endian_read_16(packet, 3 + numResponses*(6+1+1+3)   + i*2) & 0x7fff;
            rssi  = packet [3 + numResponses*(6+1+1+3+2) + i*1];
        }
        printf("Device found: %s with COD: 0x%06x, pageScan %u, clock offset 0x%04x, rssi 0x%02x\n", bd_addr_to_str(addr), classOfDevice, pageScanRepetitionMode, clockOffset, rssi);
    }
}

void packet_handler (uint8_t packet_type, uint8_t *packet, uint16_t size){
    ...
    switch (event) {
         case HCI_STATE_WORKING:
            hci_send_cmd(&hci_write_inquiry_mode, 0x01); // with RSSI
            break;
        case HCI_EVENT_COMMAND_COMPLETE:
            if (COMMAND_COMPLETE_EVENT(packet, hci_write_inquiry_mode) ) {
                start_scan();
            }
        case HCI_EVENT_COMMAND_STATUS:
            if (COMMAND_STATUS_EVENT(packet, hci_write_inquiry_mode) ) {
                printf("Ignoring error (0x%x) from hci_write_inquiry_mode.\n", packet[2]);
                hci_send_cmd(&hci_inquiry, HCI_INQUIRY_LAP, INQUIRY_INTERVAL, 0);
            }
            break;
        case HCI_EVENT_INQUIRY_RESULT:
        case HCI_EVENT_INQUIRY_RESULT_WITH_RSSI:
            print_inquiry_results(packet);
            break;
       ...
    }
}

By default, neither RSSI values nor EIR are reported. If the Bluetooth device implements Bluetooth Specification 2.1 or higher, the hci_write_inquiry_mode command enables reporting of this advanced features (0 for standard results, 1 for RSSI, 2 for RSSI and EIR).

A complete GAP inquiry example is provided here.

Pairing of Devices

By default, Bluetooth communication is not authenticated, and any device can talk to any other device. A Bluetooth device (for example, cellular phone) may choose to require authentication to provide a particular service (for example, a Dial-Up service). The process of establishing authentication is called pairing. Bluetooth provides two mechanism for this.

On Bluetooth devices that conform to the Bluetooth v2.0 or older specification, a PIN code (up to 16 bytes ASCII) has to be entered on both sides. This isn’t optimal for embedded systems that do not have full I/O capabilities. To support pairing with older devices using a PIN, see Listing below.

void packet_handler (uint8_t packet_type, uint8_t *packet, uint16_t size){
    ...
    switch (event) {
        case HCI_EVENT_PIN_CODE_REQUEST:
            // inform about pin code request
            printf("Pin code request - using '0000'\n\r");
            hci_event_pin_code_request_get_bd_addr(packet, bd_addr);

            // baseband address, pin length, PIN: c-string
            hci_send_cmd(&hci_pin_code_request_reply, &bd_addr, 4, "0000");
            break;
       ...
    }
}

The Bluetooth v2.1 specification introduces Secure Simple Pairing (SSP), which is a better approach as it both improves security and is better adapted to embedded systems. With SSP, the devices first exchange their IO Capabilities and then settle on one of several ways to verify that the pairing is legitimate. If the Bluetooth device supports SSP, BTstack enables it by default and even automatically accepts SSP pairing requests. Depending on the product in which BTstack is used, this may not be desired and should be replaced with code to interact with the user.

Regardless of the authentication mechanism (PIN/SSP), on success, both devices will generate a link key. The link key can be stored either in the Bluetooth module itself or in a persistent storage, see here. The next time the device connects and requests an authenticated connection, both devices can use the previously generated link key. Please note that the pairing must be repeated if the link key is lost by one device.

Dedicated Bonding

Aside from the regular bonding, Bluetooth also provides the concept of “dedicated bonding”, where a connection is established for the sole purpose of bonding the device. After the bonding process is over, the connection will be automatically terminated. BTstack supports dedicated bonding via the gap_dedicated_bonding function.

SPP - Serial Port Profile

The SPP profile defines how to set up virtual serial ports and connect two Bluetooth enabled devices. Please keep in mind that a serial port does not preserve packet boundaries if you try to send data as packets and read about RFCOMM packet boundaries.

Accessing an SPP Server on a remote device

To access a remote SPP server, you first need to query the remote device for its SPP services. Section on querying remote SDP service shows how to query for all RFCOMM channels. For SPP, you can do the same but use the SPP UUID 0x1101 for the query. After you have identified the correct RFCOMM channel, you can create an RFCOMM connection as shown here.

Providing an SPP Server

To provide an SPP Server, you need to provide an RFCOMM service with a specific RFCOMM channel number as explained in section on RFCOMM service. Then, you need to create an SDP record for it and publish it with the SDP server by calling sdp_register_service. BTstack provides the spp_create_sdp_record function in that requires an empty buffer of approximately 200 bytes, the service channel number, and a service name. Have a look at the SPP Counter example.

PAN - Personal Area Networking Profile

The PAN profile uses BNEP to provide on-demand networking capabilities between Bluetooth devices. The PAN profile defines the following roles:

PANU is a Bluetooth device that communicates as a client with GN, or NAP, or with another PANU Bluetooth device, through a point-to-point connection. Either the PANU or the other Bluetooth device may terminate the connection at anytime.

NAP is a Bluetooth device that provides the service of routing network packets between PANU by using BNEP and the IP routing mechanism. A NAP can also act as a bridge between Bluetooth networks and other network technologies by using the Ethernet packets.

The GN role enables two or more PANUs to interact with each other through a wireless network without using additional networking hardware. The devices are connected in a piconet where the GN acts as a master and communicates either point-to-point or a point-to-multipoint with a maximum of seven PANU slaves by using BNEP.

Currently, BTstack supports only PANU.

Accessing a remote PANU service

To access a remote PANU service, you first need perform an SDP query to get the L2CAP PSM for the requested PANU UUID. With these two pieces of information, you can connect BNEP to the remote PANU service with the bnep_connect function. The Section on PANU Demo example shows how this is accomplished.

Providing a PANU service

To provide a PANU service, you need to provide a BNEP service with the service UUID, e.g. the PANU UUID, and a maximal ethernet frame size, as explained in Section on BNEP service. Then, you need to create an SDP record for it and publish it with the SDP server by calling sdp_register_service. BTstack provides the pan_create_panu_sdp_record function in src/pan.c that requires an empty buffer of approximately 200 bytes, a description, and a security description.

HSP - Headset Profile

The HSP profile defines how a Bluetooth-enabled headset should communicate with another Bluetooth enabled device. It relies on SCO for audio encoded in 64 kbit/s CVSD and a subset of AT commands from GSM 07.07 for minimal controls including the ability to ring, answer a call, hang up and adjust the volume.

The HSP defines two roles:

There are following restrictions: - The CVSD is used for audio transmission.

%TODO: audio paths

HFP - Hands-Free Profile

The HFP profile defines how a Bluetooth-enabled device, e.g. a car kit or a headset, can be used to place and receive calls via a audio gateway device, typically a mobile phone. It relies on SCO for audio encoded in 64 kbit/s CVSD and a bigger subset of AT commands from GSM 07.07 then HSP for controls including the ability to ring, to place and receive calls, join a conference call, to answer, hold or reject a call, and adjust the volume.

The HFP defines two roles:

Supported Features

The supported features define the HFP capabilities of the device. The enumeration unfortunately differs between HF and AG sides.

The AG supported features are set by combining the flags that start with HFP_AGSF_xx and calling hfp_ag_init_supported_features, followed by creating SDP record for the service using the same feature set.

Similarly, the HF supported features are a combination of HFP_HFSF_xx flags and are configured by calling hfp_hf_init_supported_features, as well as creating an SDP record.

Define for AG Supported Feature Description
HFP_AGSF_THREE_WAY_CALLING Three-way calling
HFP_AGSF_EC_NR_FUNCTION Echo Canceling and/or Noise Reduction function
HFP_AGSF_VOICE_RECOGNITION_FUNCTION Voice recognition function
HFP_AGSF_IN_BAND_RING_TONE In-band ring tone capability
HFP_AGSF_ATTACH_A_NUMBER_TO_A_VOICE_TAG Attach a number to a voice tag
HFP_AGSF_ABILITY_TO_REJECT_A_CALL Ability to reject a call
HFP_AGSF_ENHANCED_CALL_STATUS Enhanced call status
HFP_AGSF_ENHANCED_CALL_CONTROL Enhanced call control
HFP_AGSF_EXTENDED_ERROR_RESULT_CODES Extended Error Result Codes
HFP_AGSF_CODEC_NEGOTIATION Codec negotiation
HFP_AGSF_HF_INDICATORS HF Indicators
HFP_AGSF_ESCO_S4 eSCO S4 (and T2) Settings Supported
HFP_AGSF_ENHANCED_VOICE_RECOGNITION_STATUS Enhanced voice recognition status
HFP_AGSF_VOICE_RECOGNITION_TEXT Voice recognition text
Define for HF Supported Feature Description
HFP_HFSF_THREE_WAY_CALLING Three-way calling
HFP_HFSF_EC_NR_FUNCTION Echo Canceling and/or Noise Reduction function
HFP_HFSF_CLI_PRESENTATION_CAPABILITY CLI presentation capability
HFP_HFSF_VOICE_RECOGNITION_FUNCTION Voice recognition function
HFP_HFSF_REMOTE_VOLUME_CONTROL Remote volume control
HFP_HFSF_ATTACH_A_NUMBER_TO_A_VOICE_TAG Attach a number to a voice tag
HFP_HFSF_ENHANCED_CALL_STATUS Enhanced call status
HFP_HFSF_ENHANCED_CALL_CONTROL Enhanced call control
HFP_HFSF_CODEC_NEGOTIATION Codec negotiation
HFP_HFSF_HF_INDICATORS HF Indicators
HFP_HFSF_ESCO_S4 eSCO S4 (and T2) Settings Supported
HFP_HFSF_ENHANCED_VOICE_RECOGNITION_STATUS Enhanced voice recognition status
HFP_HFSF_VOICE_RECOGNITION_TEXT Voice recognition text

Audio Voice Recognition Activation

Audio voice recognition (AVR) requires that HF and AG have the following features enabled:

It can be activated or deactivated on both sides by calling:

// AG
uint8_t hfp_ag_activate_voice_recognition(hci_con_handle_t acl_handle);
uint8_t hfp_ag_deactivate_voice_recognition(hci_con_handle_t acl_handle);

// HF
uint8_t hfp_hf_activate_voice_recognition(hci_con_handle_t acl_handle);
uint8_t hfp_hf_deactivate_voice_recognition(hci_con_handle_t acl_handle);

On activation change, the HFP_SUBEVENT_VOICE_RECOGNITION_(DE)ACTIVATED event will be emitted with status field set to ERROR_CODE_SUCCESS on success.

Voice recognition will stay active until either the deactivation command is called, or until the current Service Level Connection between the AG and the HF is dropped for any reason.

Use cases Expected behavior
No previous audio connection, AVR activated then deactivated Audio connection will be opened by AG upon AVR activation, and upon AVR deactivation closed
AVR activated and deactivated during existing audio connection Audio remains active upon AVR deactivation
Call to close audio connection during active AVR session The audio connection shut down will be refused
AVR activated, but audio connection failed to be established AVR will stay activated

Beyond the audio routing and voice recognition activation capabilities, the rest of the voice recognition functionality is implementation dependent - the stack only provides the signaling for this.

Enhanced Audio Voice Recognition

Similarly to AVR, Enhanced Audio voice recognition (eAVR) requires that HF and AG have the following features enabled:

In addition, to allow textual representation of audio that is parsed by eAVR (note that parsing is not part of Bluetooth specification), both devices must enable:

eAVR implements the same use cases as AVR (see previous section) and it can be activated or deactivated using the same API as for AVR, see above.

When eAVR and audio channel are established there are several additional commands that can be sent:

HFP Role eVRA API Description
HF hfp_hf_enhanced_voice_recognition_report_ready_for_audio Ready to accept audio input.
AG hfp_ag_enhanced_voice_recognition_report_ready_for_audio Voice recognition engine is ready to accept audio input.
AG hfp_ag_enhanced_voice_recognition_report_sending_audio The voice recognition engine will play a sound, e.g. starting sound.
AG hfp_ag_enhanced_voice_recognition_report_processing_input Voice recognition engine is processing the audio input.
AG hfp_ag_enhanced_voice_recognition_send_message Report textual representation from the voice recognition engine.

HID - Human-Interface Device Profile

The HID profile allows an HID Host to connect to one or more HID Devices and communicate with them. Examples of Bluetooth HID devices are keyboards, mice, joysticks, gamepads, remote controls, and also voltmeters and temperature sensors. Typical HID hosts would be a personal computer, tablets, gaming console, industrial machine, or data-recording device.

Please refer to:

GAP LE - Generic Access Profile for Low Energy

As with GAP for Classic, the GAP LE profile defines how to discover and how to connect to a Bluetooth Low Energy device. There are several GAP roles that a Bluetooth device can take, but the most important ones are the Central and the Peripheral role. Peripheral devices are those that provide information or can be controlled. Central devices are those that consume information or control the peripherals. Before the connection can be established, devices are first going through an advertising process.

Private addresses.

To better protect privacy, an LE device can choose to use a private i.e. random Bluetooth address. This address changes at a user-specified rate. To allow for later reconnection, the central and peripheral devices will exchange their Identity Resolving Keys (IRKs) during bonding. The IRK is used to verify if a new address belongs to a previously bonded device.

To toggle privacy mode using private addresses, call the gap_random_address_set_mode function. The update period can be set with gap_random_address_set_update_period.

After a connection is established, the Security Manager will try to resolve the peer Bluetooth address as explained in Section on SMP.

Advertising and Discovery

An LE device is discoverable and connectable, only if it periodically sends out Advertisements. An advertisement contains up to 31 bytes of data. To configure and enable advertisement broadcast, the following GAP functions can be used:

In addition to the Advertisement data, a device in the peripheral role can also provide Scan Response data, which has to be explicitly queried by the central device. It can be set with gap_scan_response_set_data.

Please have a look at the SPP and LE Counter example.

The scan parameters can be set with gap_set_scan_parameters. The scan can be started/stopped with gap_start_scan/gap_stop_scan.

Finally, if a suitable device is found, a connection can be initiated by calling gap_connect. In contrast to Bluetooth classic, there is no timeout for an LE connection establishment. To cancel such an attempt, gap_connect_cancel has be be called.

By default, a Bluetooth device stops sending Advertisements when it gets into the Connected state. However, it does not start broadcasting advertisements on disconnect again. To re-enable it, please send the hci_le_set_advertise_enable again .

GATT Client

The GATT profile uses ATT Attributes to represent a hierarchical structure of GATT Services and GATT Characteristics. Each Service has one or more Characteristics. Each Characteristic has meta data attached like its type or its properties. This hierarchy of Characteristics and Services are queried and modified via ATT operations.

GATT defines both a server and a client role. A device can implement one or both GATT roles.

The GATT Client is used to discover services, characteristics and their descriptors on a peer device. It allows to subscribe for notifications or indications that the characteristic on the GATT server has changed its value.

To perform GATT queries, it provides a rich interface. Before calling queries, the GATT client must be initialized with gatt_client_init once.

To allow for modular profile implementations, GATT client can be used independently by multiple entities.

After an LE connection was created using the GAP LE API, you can query for the connection MTU with gatt_client_get_mtu.

Multiple GATT queries to the same GATT Server cannot be interleaved. Therefore, you can either use a state machine or similar to perform the queries in sequence, or you can check if you can perform a GATT query on a particular connection right now using gatt_client_is_ready, and retry later if it is not ready. As a result to a GATT query, zero to many GATT_EVENT_Xs are returned before a GATT_EVENT_QUERY_COMPLETE event completes the query.

For more details on the available GATT queries, please consult GATT Client API.

Authentication

By default, the GATT Server is responsible for security and the GATT Client does not enforce any kind of authentication. If the GATT Client accesses Characteristic that require encrytion or authentication, the remote GATT Server will return an error, which is returned in the att status of the GATT_EVENT_QUERY_COMPLETE.

You can define ENABLE_GATT_CLIENT_PAIRING to instruct the GATT Client to trigger pairing in this case and to repeat the request.

This model allows for an attacker to spoof another device, but don't require authentication for the Characteristics. As a first improvement, you can define ENABLE_LE_PROACTIVE_AUTHENTICATION in btstack_config.h. When defined, the GATT Client will request the Security Manager to re-encrypt the connection if there is stored bonding information available. If this fails, the GATT_EVENT_QUERY_COMPLETE will return with the att status ATT_ERROR_BONDING_INFORMATION_MISSING.

With ENABLE_LE_PROACTIVE_AUTHENTICATION defined and in Central role, you need to delete the local bonding information if the remote lost its bonding information, e.g. because of a device reset. See example/sm_pairing_central.c.

Even with the Proactive Authentication, your device may still connect to an attacker that provides the same advertising data as your actual device. If the device that you want to connect requires pairing, you can instruct the GATT Client to automatically request an encrypted connection before sending any GATT Client request by calling gatt_client_set_required_security_level(). If the device provides sufficient IO capabilities, a MITM attack can then be prevented. We call this 'Mandatory Authentication'.

The following diagrams provide a detailed overview about the GATT Client security mechanisms in different configurations:

GATT Server

The GATT server stores data and accepts GATT client requests, commands and confirmations. The GATT server sends responses to requests and when configured, sends indication and notifications asynchronously to the GATT client.

To save on both code space and memory, BTstack does not provide a GATT Server implementation. Instead, a textual description of the GATT profile is directly converted into a compact internal ATT Attribute database by a GATT profile compiler. The ATT protocol server - implemented by and - answers incoming ATT requests based on information provided in the compiled database and provides read- and write-callbacks for dynamic attributes.

GATT profiles are defined by a simple textual comma separated value (.csv) representation. While the description is easy to read and edit, it is compact and can be placed in ROM.

The current format is shown in Listing below.

// import service_name
#import <service_name.gatt>

PRIMARY_SERVICE, {SERVICE_UUID}
CHARACTERISTIC, {ATTRIBUTE_TYPE_UUID}, {PROPERTIES}, {VALUE}
CHARACTERISTIC, {ATTRIBUTE_TYPE_UUID}, {PROPERTIES}, {VALUE}
...
PRIMARY_SERVICE, {SERVICE_UUID}
CHARACTERISTIC, {ATTRIBUTE_TYPE_UUID}, {PROPERTIES}, {VALUE}
...

UUIDs are either 16 bit (1800) or 128 bit (00001234-0000-1000-8000-00805F9B34FB).

Value can either be a string (“this is a string”), or, a sequence of hex bytes (e.g. 01 02 03).

Properties can be a list of properties combined using '|'

Reads/writes to a Characteristic that is defined with the DYNAMIC flag, are forwarded to the application via callback. Otherwise, the Characteristics cannot be written and it will return the specified constant value.

Adding NOTIFY and/or INDICATE automatically creates an additional Client Configuration Characteristic.

Property Description
READ Characteristic can be read
WRITE Characteristic can be written using Write Request
WRITE_WITHOUT_RESPONSE Characteristic can be written using Write Command
NOTIFY Characteristic allows notifications by server
INDICATE Characteristic allows indication by server
DYNAMIC Read or writes to Characteristic are handled by application

To require encryption or authentication before a Characteristic can be accessed, you can add one or more of the following properties:

Property Description
AUTHENTICATION_REQUIRED Read and Write operations require Authentication
READ_ENCRYPTED Read operations require Encryption
READ_AUTHENTICATED Read operations require Authentication
WRITE_ENCRYPTED Write operations require Encryption
WRITE_AUTHENTICATED Write operations require Authentication
ENCRYPTION_KEY_SIZE_X Require encryption size >= X, with W in [7..16]

For example, Volume State Characteristic (Voice Control Service) requires: - Mandatory Properties: Read, Notify - Security Permissions: Encryption Required

In addition, its read is handled by application. We can model this Characteristic as follows:

CHARACTERISTIC, ORG_BLUETOOTH_CHARACTERISTIC_VOLUME_STATE, DYNAMIC | READ | NOTIFY | ENCRYPTION_KEY_SIZE_16

To use already implemented GATT Services, you can import it using the #import command. See list of provided services.

BTstack only provides an ATT Server, while the GATT Server logic is mainly provided by the GATT compiler. While GATT identifies Characteristics by UUIDs, ATT uses Handles (16 bit values). To allow to identify a Characteristic without hard-coding the attribute ID, the GATT compiler creates a list of defines in the generated *.h file.

Similar to other protocols, it might be not possible to send any time. To send a Notification, you can call att_server_request_to_send_notification to request a callback, when yuo can send the Notification.

If your application cannot handle an ATT Read Request in the att_read_callback in some situations, you can enable support for this by adding ENABLE_ATT_DELAYED_RESPONSE to btstack_config.h. Now, you can store the requested attribute handle and return ATT_READ_RESPONSE_PENDING instead of the length of the provided data when you don't have the data ready. For ATT operations that read more than one attribute, your att_read_callback might get called multiple times as well. To let you know that all necessary attribute handles have been 'requested' by the att_server, you'll get a final att_read_callback with the attribute handle of ATT_READ_RESPONSE_PENDING. When you've got the data for all requested attributes ready, you can call att_server_response_ready, which will trigger processing of the current request. Please keep in mind that there is only one active ATT operation and that it has a 30 second timeout after which the ATT server is considered defunct by the GATT Client.

Implementing Standard GATT Services

Implementation of a standard GATT Service consists of the following 4 steps:

  1. Identify full Service Name
  2. Use Service Name to fetch XML definition at Bluetooth SIG site and convert into generic .gatt file
  3. Edit .gatt file to set constant values and exclude unwanted Characteristics
  4. Implement Service server, e.g., battery_service_server.c

Step 1:

To facilitate the creation of .gatt files for standard profiles defined by the Bluetooth SIG, the tool/convert_gatt_service.py script can be used. When run without a parameter, it queries the Bluetooth SIG website and lists the available Services by their Specification Name, e.g., org.bluetooth.service.battery_service.

$ tool/convert_gatt_service.py
Fetching list of services from https://www.bluetooth.com/specifications/gatt/services

Specification Type                                     | Specification Name            | UUID
-------------------------------------------------------+-------------------------------+-----------
org.bluetooth.service.alert_notification               | Alert Notification Service    | 0x1811
org.bluetooth.service.automation_io                    | Automation IO                 | 0x1815
org.bluetooth.service.battery_service                  | Battery Service               | 0x180F
...
org.bluetooth.service.weight_scale                     | Weight Scale                  | 0x181D

To convert a service into a .gatt file template, please call the script again with the requested Specification Type and the output file name
Usage: tool/convert_gatt_service.py SPECIFICATION_TYPE [service_name.gatt]

Step 2:

To convert service into .gatt file, call *tool/convert_gatt_service.py with the requested Specification Type and the output file name.

$ tool/convert_gatt_service.py org.bluetooth.service.battery_service battery_service.gatt
Fetching org.bluetooth.service.battery_service from
https://www.bluetooth.com/api/gatt/xmlfile?xmlFileName=org.bluetooth.service.battery_service.xml

Service Battery Service
- Characteristic Battery Level - properties ['Read', 'Notify']
-- Descriptor Characteristic Presentation Format       - TODO: Please set values
-- Descriptor Client Characteristic Configuration

Service successfully converted into battery_service.gatt
Please check for TODOs in the .gatt file

Step 3:

In most cases, you will need to customize the .gatt file. Please pay attention to the tool output and have a look at the generated .gatt file.

E.g. in the generated .gatt file for the Battery Service

// Specification Type org.bluetooth.service.battery_service
// https://www.bluetooth.com/api/gatt/xmlfile?xmlFileName=org.bluetooth.service.battery_service.xml

// Battery Service 180F
PRIMARY_SERVICE, ORG_BLUETOOTH_SERVICE_BATTERY_SERVICE
CHARACTERISTIC, ORG_BLUETOOTH_CHARACTERISTIC_BATTERY_LEVEL, DYNAMIC | READ | NOTIFY,
// TODO: Characteristic Presentation Format: please set values
#TODO CHARACTERISTIC_FORMAT, READ, _format_, _exponent_, _unit_, _name_space_, _description_
CLIENT_CHARACTERISTIC_CONFIGURATION, READ | WRITE,

you could delete the line regarding the CHARACTERISTIC_FORMAT, since it's not required if there is a single instance of the service. Please compare the .gatt file against the Adopted Specifications.

Step 4:

As described above all read/write requests are handled by the application. To implement the new services as a reusable module, it's necessary to get access to all read/write requests related to this service.

For this, the ATT DB allows to register read/write callbacks for a specific handle range with att_server_register_service_handler().

Since the handle range depends on the application's .gatt file, the handle range for Primary and Secondary Services can be queried with gatt_server_get_get_handle_range_for_service_with_uuid16.

Similarly, you will need to know the attribute handle for particular Characteristics to handle Characteristic read/writes requests. You can get the attribute value handle for a Characteristics gatt_server_get_value_handle_for_characteristic_with_uuid16().

In addition to the attribute value handle, the handle for the Client Characteristic Configuration is needed to support Indications/Notifications. You can get this attribute handle with gatt_server_get_client_configuration_handle_for_characteristic_with_uuid16()

Finally, in order to send Notifications and Indications independently from the main application, att_server_register_can_send_now_callback can be used to request a callback when it's possible to send a Notification or Indication.

To see how this works together, please check out the Battery Service Server in src/ble/battery_service_server.c.

GATT Database Hash

When a GATT Client connects to a GATT Server, it cannot know if the GATT Database has changed and has to discover the provided GATT Services and Characteristics after each connect.

To speed this up, the Bluetooth specification defines a GATT Service Changed Characteristic, with the idea that a GATT Server would notify a bonded GATT Client if its database changed. However, this is quite fragile and it is not clear how it can be implemented in a robust way.

The Bluetooth Core Spec 5.1 introduced the GATT Database Hash Characteristic, which allows for a simple robust mechanism to cache a remote GATT Database. The GATT Database Hash is a 16-byte value that is calculated over the list of Services and Characteristics. If there is any change to the database, the hash will change as well.

To support this on the GATT Server, you only need to add a GATT Service with the GATT Database Characteristic to your .gatt file. The hash value is then calculated by the GATT compiler.

PRIMARY_SERVICE, GATT_SERVICE
CHARACTERISTIC, GATT_DATABASE_HASH, READ,

Note: make sure to install the PyCryptodome python package as the hash is calculated using AES-CMAC, e.g. with:

pip install pycryptodomex