Hi
I am running on a nRF52832 chip with following software stacks
- Mesh SDK v3.1.0
- nRF SDK v15.2.0
- Softdevice v6.1.0
I am working on a use-case where I am processing some GATT messages. Sometimes I need to create a mesh message based upon the GATT message. The message contains a target mesh address. This address is either a unicast or a group address. I need to publish a message on a model handle to a specific target address.
What I am trying to do this basically this: status = access_model_publish_to(p_client->model_handle, &p_client->access_message.message, target_address);
Currently I am using the unacknowledged publish with:
status = access_model_publish(p_client->model_handle, &p_client->access_message.message);
Just before creating and publishing the message I am using the following code to change the target address
static void publish_address_change(const access_model_handle_t model_handle, const uint16_t address)
{
// Add and set new target
dsm_handle_t publish_address_handle = DSM_HANDLE_INVALID;
NRF_MESH_ERROR_CHECK(dsm_address_publish_add(actual_address, &publish_address_handle));
NRF_MESH_ERROR_CHECK(access_model_publish_address_set(model_handle, publish_address_handle));
}
My code is based upon the Hung Bui's reply here: https://devzone.nordicsemi.com/f/nordic-q-a/35931/multiple-server-on-one-client---light-switch-demo
The code works well until the m_addresses array in device_state_manager.c have been filled up.
I can increase the size of the array with DSM_NONVIRTUAL_ADDR_MAX, but that only postpone the problem.
I have tried to delete the old address when I change to a new address. To accomplish this, I have the following code:
static void publish_address_change(const access_model_handle_t model_handle, const uint16_t address)
{
// Convert GATT target address to mesh target address
uint16_t actual_address = process_target_address(address);
// Remove old publish address if one exists
dsm_handle_t old_publish_address_handle = DSM_HANDLE_INVALID;
if (access_model_publish_address_get(model_handle, &old_publish_address_handle) == NRF_SUCCESS)
{
nrf_mesh_address_t publish_address_stored = {0};
if (dsm_address_get(old_publish_address_handle, &publish_address_stored) == NRF_SUCCESS &&
publish_address_stored.value == address)
{
// No need to change publish address, return
return;
}
else
{
dsm_address_publish_remove(old_publish_address_handle);
}
}
// Add and set new target
dsm_handle_t publish_address_handle = DSM_HANDLE_INVALID;
NRF_MESH_ERROR_CHECK(dsm_address_publish_add(actual_address, &publish_address_handle));
NRF_MESH_ERROR_CHECK(access_model_publish_address_set(model_handle, publish_address_handle));
}
Unfortunately, this doesn't work. In a test setup with 6 nodes they eventually end up in a state where. when turning power on and off, the dsm addresses and access_model handles go out of sync.
Specifically I get a mesh assert error when it tries to add an address handle that points to a 0x0000 address in dsm_address_publish_add_handle(dsm_handle_t address_handle)
This happens in mesh_stack_init -> access_flash_config_load
What am I doing wrong?
Is there a recommended way of implementing a gateway feature?
Thanks,
Thomas
