twi busy error 17

[schef]

Hello guys, i randomly get twi error busy 17.

i have #define TWI_DEFAULT_CONFIG_IRQ_PRIORITY 7 in sdk_config.h and .interrupt_priority = APP_IRQ_PRIORITY_LOW in init function.

This is my code:

void twi_init(void)
{
  if(twi_init_status) return;
  ret_code_t err_code;

  const nrf_drv_twi_config_t twi_mma_7660_config = {
    .scl                = SCL,
    .sda                = SDA,
    .frequency          = NRF_TWI_FREQ_100K,
    .interrupt_priority = APP_IRQ_PRIORITY_LOW
  };

  err_code = nrf_drv_twi_init(&twi_instance, &twi_mma_7660_config, twi_handler, NULL);
  APP_ERROR_CHECK(err_code);

  nrf_drv_twi_enable(&twi_instance);
  nrf_delay_ms(200);
  twi_init_status = true;
}

void twi_handler(nrf_drv_twi_evt_t const * p_event, void * p_context)
{
  switch(p_event->type)
  {
    case NRF_DRV_TWI_EVT_DONE:
      m_xfer_done = true;
      break;
    default:
      break;
  }
}

and this are my read and write functions:

void twi_write(uint8_t dev_addres, uint8_t reg_addres, uint8_t data[], uint8_t length)
{
  uint8_t buffer[length];
  buffer[0] = reg_addres;
  for(int i = 0; i < length; i++) buffer[i+1] = data[i];

  m_xfer_done = false;
  ret_code_t err_code = nrf_drv_twi_tx(&twi_instance, dev_addres, buffer, sizeof(buffer) + 1, false);
  APP_ERROR_CHECK(err_code);
  while (!m_xfer_done);
}

void twi_write_no_register(uint8_t dev_addres, uint8_t data[], uint8_t length)
{
  m_xfer_done = false;
  ret_code_t err_code = nrf_drv_twi_tx(&twi_instance, dev_addres, data, length, false);
  APP_ERROR_CHECK(err_code);
  while (!m_xfer_done);
}

void twi_read(uint8_t dev_addres, uint8_t reg_addres, uint8_t *data_pointer, uint8_t length)
{
  ret_code_t err_code;
  uint8_t buffer[length];
  buffer[0] = reg_addres;

  m_xfer_done = false;
  err_code = nrf_drv_twi_tx(&twi_instance, dev_addres, buffer, 1, false); // sends ADDR+W, dev_addres, reg_addres
  APP_ERROR_CHECK(err_code);
  while (!m_xfer_done);

  m_xfer_done = false;
  err_code = nrf_drv_twi_rx(&twi_instance, dev_addres, buffer, length); // reads from reg_addres #(length) of bytes
  APP_ERROR_CHECK(err_code);
  while (!m_xfer_done);

  for(int i=0;i<length;i++)
  {
    *data_pointer=buffer[i];
    data_pointer++;
  }
}

I was reading posts about this problem but coudn't find a solution. I tried to put delays after twi functions but it didn't resolve the issue. Is this something i should handle in my code (if yes how would be helpfull)? I have also tried blocking mode and the error is the same. Sometimes it works for hours and sometime error occures right away.

Thanks.

[MartinBL]

Hi,

1. Have you declared m_xfer_done as 'volatile'?
2. Do you have any other interrupts going on? Are you calling any TWI read/write functions from other interrupts?
3. Do you mind showing the code you use to initialize the TWI?

[schef]

I have edited the anwser to add the init code. I have static volatile bool m_xfer_done = false. I have alot of code going on but i think i don't have any. Is there a simple way to find out?

[MartinBL]

If you call one of your functions above inside an interrupt/event handler I think you would know. Are you doing TWI operations based on any Softdevice events?

Is it one particular line that returns busy all the time?

I think it is very strange that you see the busy error in blocking mode too. I didn't think that was possible.

I guess it could be smart to implement some code for NRF_DRV_TWI_EVT_ADDRESS_NACK and NRF_DRV_TWI_EVT_DATA_NACK as well as the NRF_DRV_TWI_EVT_DONE in your event handler to catch such errors. I can't see how that could be relevant to the busy error though.

I think it might be useful to see the rest of your code. You can create a support ticket at nordicsemi.com if you want confidentiality.

[schef]

I don't do TWI in combination with softdevice. I use it for reading temperature, gps, rtc. Every time the same line returns busy error, the nrf_drv_twi_rx command. Would it be usefull if i pass the error code, make some logic outside, and then try to read again?

[MartinBL]

I guess you can do something like:

uint32_t timeout = some timeout value;
// Do TWI transfer here
while((!m_xfer_done) && --timeout)
    ;
if(!timeout) return NRF_ERROR_TIMEOUT;

and then handle the timeout in your code, but it would be better to find the root cause as well, wouldn't it?