Using the nRF9151 with nRF Connect SDK v3.1.0 and modem firmware 2.0.2, we're seeing an issue where the modem works fine for days, then it is suddenly unable to open any sockets, with attempts failing with NRF_ETIMEDOUT. We've seen this happen on two out of a handful of devices in initial trials.
Specifically, we connect to a cloud server every 10 minutes and send a report. This starts with a call to zsock_connect(). In problem devices, this begins to return -1, with errno set to NRF_ETIMEDOUT.
In tracing this call, we found that zsock_connect() calls nrf9x_socket_offload_connect(), as the sockets are offloaded to the modem. Then, this function calls nrf_connect() on line 462 of nrf9x_sockets.c (as it is an IPv4 address we're connecting to). It is this function that sets errno to NRF_ETIMEDOUT, but as this function is in the modem firmware we cannot inspect it. The function waits approximately 25 seconds before returning, so it does seem like the modem is trying to do something before timing out.
Do you have any theories as to why this would happen? If the connection were poor, we'd expect occasional timeouts, but in this specific issue, everything is fine, until all of a sudden, every connection attempt times out. Power-cycling the system seems to cause it to start working again, so it doesn't seem to be an issue of poor signal.
For reference, lte_lc_conn_eval_params_get() returns the following on a unit in the problem state:
struct lte_lc_conn_eval_params {
enum lte_lc_rrc_mode rrc_state = LTE_LC_RRC_MODE_IDLE;
enum lte_lc_energy_estimate energy_estimate = LTE_LC_ENERGY_CONSUMPTION_NORMAL;
enum lte_lc_tau_triggered tau_trig = LTE_LC_CELL_IN_TAI_LIST;
enum lte_lc_ce_level ce_level = LTE_LC_CE_LEVEL_0;
int earfcn = 5035;
int16_t dl_pathloss = 105; //105 dB
int16_t rsrp = 54; // -87 dBm
int16_t rsrq = 14; // -13.0 dB
int16_t tx_rep = 1;
int16_t rx_rep = 8;
int16_t phy_cid = 204;
int16_t band = 12;
int16_t snr = 25; // 1 dB
int16_t tx_power = 1; // 1 dBm
int mcc = 310;
int mnc = 260;
uint32_t cell_id = <redacted>;
};
We'd appreciate any insight you can provide into the causes of this error.
