When calling bt_le_adv_update_data random assert in bt_hci_cmd_send_sync

[byte]

I have code that updates the advertising data once per second using bt_le_adv_update_data. This works fine but every 30 minutes or so I get an assert (see below) in bt_hci_cmd_send_sync waiting to take a k_sem. I can only assume the second core with the Bluetooth subsystem failed because the k_sem_take has a 10 second timeout which seems to be plenty of time to handle the hci command. I'm using nRF Connect v2.3.0. I'm new to multi-core embedded apps. Is there a way to monitor the second cores behavior, or debug it separately? I believe I am using the provided SoftDevice controller on the net core which is handling the bluetooth communications.

int bt_hci_cmd_send_sync(uint16_t opcode, struct net_buf *buf,

             struct net_buf **rsp)

{

    struct k_sem sync_sem;

    uint8_t status;

    int err;

    if (!buf) {

        buf = bt_hci_cmd_create(opcode, 0);

        if (!buf) {

            return -ENOBUFS;

        }

    }

    LOG_DBG("buf %p opcode 0x%04x len %u", buf, opcode, buf->len);

    k_sem_init(&sync_sem, 0, 1);

    cmd(buf)->sync = &sync_sem;

    net_buf_put(&bt_dev.cmd_tx_queue, net_buf_ref(buf));

    err = k_sem_take(&sync_sem, HCI_CMD_TIMEOUT);  // <<<<<<<<<<<<< this times out

    BT_ASSERT_MSG(err == 0, "k_sem_take failed with err %d", err);

    status = cmd(buf)->status;

    if (status) {

        LOG_WRN("opcode 0x%04x status 0x%02x", opcode, status);

        net_buf_unref(buf);

        switch (status) {

        case BT_HCI_ERR_CONN_LIMIT_EXCEEDED:

            return -ECONNREFUSED;

        default:

            return -EIO;

        }

    }

    LOG_DBG("rsp %p opcode 0x%04x len %u", buf, opcode, buf->len);

    if (rsp) {

        *rsp = buf;

    } else {

        net_buf_unref(buf);

    }

    return 0;

}

[Susheel Nuguru]

It is not possible to debug the Softdevice Controller as this is given as a binary and the symbol table is not part of it.

The error that you see in taking the semaphore is on the host side and should be built into the application on the app core. So you should be debugging the app core not the net core. Or did I understand your problem incorrectly?

[byte]

I think the semaphore is being used to wait for a net core response, is that correct? In this failure case the net core never responds, or takes longer than the timeout of 10 seconds. I have been able to catch the issue in the debugger but I'm not sure what would cause the net core to fail to respond.

[Susheel Nuguru]

I am sorry for the late reply. I am not familiar with this kind of error where a common semaphore I will try to use the code snippet you provided to try to reproduce this and give you a proper explanation.

[byte]

The code I copied above is part of Zephyr so you already have it. I am calling bt_le_adv_update_data which ultimately calls bt_hci_cmd_send_sync which is where the assert is. bt_hci_cmd_send_sync is in the file C:\ncs\v2.3.0\zephyr\subsys\bluetooth\host\hci_core.c. 

[Susheel Nuguru]

I did not manage to reproduce the assert in our latest (v2.3.0) trying to update advertising data once a second. Maybe you have a same issue like the one mentioned here and the main thing that might concern you from that github issue comments is 

bt_le_adv_update_data is called from timer ISR context, which is not supported, use delayed work instead

Check if you are calling the update from the timer ISR and if so delay the execution into a lower priority work queue.

[byte]

Thanks, I've tried putting all my BT calls in the main thread and I still have the problem. I was never running in a IRS context though, just in a separate thread. I also now have a test app that runs on a dev kit but it does not have the problem so I must have an issue but I'm not sure what to try next to track it down. 

[Susheel Nuguru]

Then I am not able to reproduce it calling it in the thread context and not the ISR context. Can you please attach a simplistic project for me to reproduce and debug this at my desk?

[erickcinco]

https://github.com/zephyrproject-rtos/zephyr/issues/52491

Possibly related to this?

[byte]

Thanks, It doesn't look like the same assert location but I'll look more closely at this issue. It seems to happen every 33 minutes no matter what we do to make it break. We've changed heap sizes, advertising update times, advertising interval times. Still 33 minutes.