Re: [PATCH 3/5] tty/8250_dw: Add support for OCTEON UARTS.

[David Daney <ddaney.cavm@gmail.com>]

On 06/19/2013 07:10 AM, Heikki Krogerus wrote:
> On Tue, Jun 18, 2013 at 12:12:53PM -0700, David Daney wrote:
>> A few differences needed by OCTEON:
>>
>> o These are DWC UARTS, but have USR at a different offset.
>>
>> o OCTEON must have 64-bit wide register accesses, so we have OCTEON
>>    specific register accessors.
>>
>> o No UCV register, so we hard code some properties.
>>
>> Signed-off-by: David Daney <david.daney@cavium.com>
>
> <snip>
>
>> @@ -270,10 +301,8 @@ static int dw8250_probe(struct platform_device *pdev)
>>   	uart.port.serial_out = dw8250_serial_out;
>>   	uart.port.private_data = data;
>>
>> -	dw8250_setup_port(&uart);
>> -
>>   	if (pdev->dev.of_node) {
>> -		err = dw8250_probe_of(&uart.port);
>> +		err = dw8250_probe_of(&uart.port, data);
>>   		if (err)
>>   			return err;
>>   	} else if (ACPI_HANDLE(&pdev->dev)) {
>> @@ -284,6 +313,9 @@ static int dw8250_probe(struct platform_device *pdev)
>>   		return -ENODEV;
>>   	}
>>
>> +	if (!data->no_ucv)
>> +		dw8250_setup_port(&uart);
>
> Moving the dw8250_setup_port() call here breaks dw8250_probe_acpi(). It
> expects values from the CPR register for the DMA burst size calculation.
>
> The DMA support can be moved to a separate function. This way it can
> be called after this point, and it will then be available for both DT
> and ACPI. I can make a patch tomorrow. That should solve this issue.
>

I am reworking the patch because other problems were found.  I will try 
to get this part right in the next version.

David Daney