protential bug in id_manager.c

[nine-fox]

Hi Nordic,

I am using SDK 15.3. While I am working on a central, which connects 2 periperial devices simultaneously, I found a id_manager.c issue, protentially a bug I guess. Any suggestions would be great, thanks.

1. variable not initialised.

See the following declaration in id_manager.c:

static im_connection_t m_connections[IM_MAX_CONN_HANDLES];

I can not find initialization code for it. As a result, if I call the following code in BLE_GAP_EVT_CONNECTED event, status.bonded will return true.

err_code = pm_conn_sec_status_get(p_gap_evt->conn_handle, &status);

if (status.bonded)
{
NRF_LOG_DEBUG("incorrect condition!");
}

By initializing m_connections.peer_id to PM_PEER_ID_INVALID in pm_init() resolves the problem.

2. invalid status

When I have first periperal device connected(conn_handle = 0x00, peer_id = 0x00), then disconnect, then second periperal device connects(conn_handle = 0x00), m_connections contains invalid data, index 0 value is not PM_PEER_ID_INVALID, as a result, Peer Manager still think it is a bonded peer(peer_id = 0). Call pm_conn_secure(conn_handle,false) will do a encryption procedure not a bonding procedure.

I am willing to do a value set(m_connections[0].peer_id = PM_PEER_ID_INVALID; in BLE_GAP_EVT_DISCONNECTED event. But not sure whether it is the right way. Would you please suggest? thanks!

NINE-FOX

[Kenneth]

Hi,

I believe this have been fixed in later release of nRF5 SDK, the latest is v17.1.0:
https://www.nordicsemi.com/Products/Development-software/nRF5-SDK/Download#infotabs

Can you please download the latest and confirm?

Best regards,
Kenneth

[nine-fox]

Well, I just tried tonight, and the problem still exists

[Kenneth]

Can the problem be related to the observer priority of the ble events?

I am looking at the implementation of ble_evt_handler() in peer_manager.c, where I can find that this seems to be handled:

ble_evt_handler()->
im_ble_evt_handler->
...
bonded_matching_peer_id = PM_PEER_ID_INVALID;
...
m_connections[gap_evt.conn_handle].peer_id = bonded_matching_peer_id;

NRF_SDH_BLE_OBSERVER(m_ble_evt_observer, PM_BLE_OBSERVER_PRIO, ble_evt_handler, NULL);

Best regards,
Kenneth

[nine-fox]

Hi thanks for your reminding! Right, I updated to make lower the proiorty of pm than application, it works!

Thanks!