Hello Nicolas

I have the feeling that moving the code chunk that uses this new
variable (rs485_enabled) here ...

quoted

      ret = uart_add_one_port(&atmel_uart, &port->uart);
      if (ret)
              goto err_add_port;

@@ -2574,7 +2572,7 @@ static int atmel_serial_probe(struct platform_device *pdev)
      device_init_wakeup(&pdev->dev, 1);
      platform_set_drvdata(pdev, port);

-     if (port->uart.rs485.flags & SER_RS485_ENABLED) {
+     if (rs485_enabled) {
              UART_PUT_MR(&port->uart, ATMEL_US_USMODE_NORMAL);
              UART_PUT_CR(&port->uart, ATMEL_US_RTSEN);
      }

... (this one ^^^) up where you can test the SER_RS485_ENABLED, can be
even simpler.

otherwise, it seems good so you can add my:

Acked-by: Nicolas Ferre <redacted>

Thanks.

I supposed that the code needs to be initialized with rs485 disabled
and then enable it. Also I did not want to change the behaviour a
driver in a patch that involves 3 other drivers. If you don't mind I
prefer that this is added to a separated patch.

Thanks