Re: [RFC patch] sci_tx_empty for scif serial device

[Michael Trimarchi <trimarchi@gandalf.sssup.it>]

Hi,

Paul Mundt wrote:
> On Tue, Apr 21, 2009 at 02:08:47PM +0200, Michael Trimarchi wrote:
>   
>> Add the sci_tx_empty function for scif serial device.
>>
>> Signed-off-by: Michael Trimarchi <trimarchi@gandalf.sssup.it>
>> ---
>> diff --git a/drivers/serial/sh-sci.c b/drivers/serial/sh-sci.c
>> index dbf5357..6cea2c4 100644
>> --- a/drivers/serial/sh-sci.c
>> +++ b/drivers/serial/sh-sci.c
>> @@ -794,6 +794,20 @@ static void sci_free_irq(struct sci_port *port)
>>  
>>  static unsigned int sci_tx_empty(struct uart_port *port)
>>  {
>> +	int txroom = SCIF_TXROOM_MAX;
>> +
>> +#if defined(CONFIG_CPU_SUBTYPE_SH7763)
>> +	if (!(port->mapbase = 0xffe00000) &&
>> +	    !(port->mapbase = 0xffe08000)) {
>> +		txroom = SCIF2_TXROOM_MAX)
>> +	}
>> +#endif
>>     
>
> Uhm.. your version of gcc must be orders of magnitude more magical than
> mine, as I see no way for this to compile :-)
>   
Of course is wrong here, I don't try the SH7763 configuration. Sorry...
>   
>> +	if (port->type = PORT_SCIF) {
>> +		if (scif_txroom(port) = txroom)
>> +			return TIOCSER_TEMT;
>> +		return 0;
>> +	}
>> +
>>     
> This looks ok anyways, is there some reason this can't be applied to
> SH7763, too?
>
>   
Apart the compilation issue, the sh7763 has the SCIF2 port has a txroom 
different.

I would like to know if it is correct that the fifosize of serial device 
may equal to the
tx buffer size. In this case I can change the txroom and tx_empty using 
this info from
the port strucy for scif device. For the sci device I don't think that 
it is useful to read the tx status register to check if it's empty or 
full, maybe no.

Michael