Peripheral crashes when Central powers off

RE: Peripheral crashes when Central powers off

quan1328 — Wed, 26 Jun 2024 16:17:50 GMT

I just got this resolve. Thanks, Vidar.

RE: Peripheral crashes when Central powers off

quan1328 — Wed, 26 Jun 2024 10:30:50 GMT

Hi Vidar,

Thanks for the link in your last reply, which hints me about the bug in `func_thread()`

When the connected central disconnects, BLE stack generates an IRQ, which calls to my on_disconnected() callback that was set up in the other module of my project?

In this link: https://docs.zephyrproject.org/latest/kernel/services/data_passing/message_queues.html

However, this can increase interrupt latency as interrupts are locked while a data item is written or read.

While the queue is read by `k_msgq_get()`, the Interrupt comes in, thus, the thread crashes.

What do you think about this crash flow?

By the way, I summarize my code below:

typedef struct esp_msg
{
	short len;
	uint8_t buf[512];
} esp_msg_t;

void func_thread1()
{
    // stack size: 4096
    // prio: 1
    
    while(1)
    {
        esp_msg_t msg;
		msg.len = 125 * 2 + 4;
		memcpy(msg.buf, ecg_send_pkt, msg.len);
		k_msgq_put(&uart_msgq, &msg, K_NO_WAIT);
		
		...
		// sleep
    }
}

void func_thread2()
{
    // stack size: 16024, increased from 4096
    // prio:  2
    
    esp_msg_t msg;

    while (1)
    {
        int status = k_msgq_get(&uart_msgq, &msg, K_MSEC(500));
		if (status != 0)
		{
			// k_yield(); no need due to K_MSEC(500)
			continue;
		}
		
		notify_user_data(msg.buf, msg.len);
    }
}

In `func_thread2()` I changed from `k_msgq_get(&uart_msgq, &msg, k_is_in_isr() ? K_NO_WAIT : K_FOREVER) == 0` to `

k_msgq_get(&uart_msgq, &msg, K_MSEC(500))` because I don't think it's needed. However, it still crashes.

Somehow, k_msgq_get() blocks the system when ISR comes in. That's why it crashes?

How should I solve this?

Regards

RE: Peripheral crashes when Central powers off

Vidar Berg — Fri, 21 Jun 2024 13:53:34 GMT

Hello,

It looks like the assert in the "ios" case is raised here because the connection object pointer is NULL: https://github.com/nrfconnect/sdk-zephyr/blob/52292768176a4372eaeb682d191dccd5c6064b66/subsys/bluetooth/host/gatt.c#L2090, which maybe makes sense given the connection has been terminated. However, I'm not sure what could be causing the busfault.

I recommend you build the project with CONFIG_RESET_ON_FATAL_ERROR=n, then try to debug your project in VS code. After the fault exception has been raised, you can click on the last stack frame for the 'uart_pro_task' thread to see where the program was running prior to the crash.

Best regards,

Vidar