[PATCH 2/2] tty/serial: at91: add a fallback option to determine uart/usart property

[Nicolas Ferre]

On 16/10/2013 22:14, Greg Kroah-Hartman :
> On Tue, Oct 15, 2013 at 11:19:18AM +0200, Nicolas Ferre wrote:
>> On 14/10/2013 15:59, Jean-Christophe PLAGNIOL-VILLARD :
>>> On 10:43 Thu 10 Oct     , Nicolas Ferre wrote:
>>>> On older SoC, the "name" field is not filled in the register map.
>>>> Fix the way to figure out if the serial port is an uart or an usart for these
>>>> older products (with corresponding properties).
>>>>
>>>> Signed-off-by: Nicolas Ferre <nicolas.ferre at atmel.com>
>>>> ---
>>>>   drivers/tty/serial/atmel_serial.c | 19 ++++++++++++++++++-
>>>>   include/linux/atmel_serial.h      |  1 +
>>>>   2 files changed, 19 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/tty/serial/atmel_serial.c b/drivers/tty/serial/atmel_serial.c
>>>> index 6b0f75e..c7d99af 100644
>>>> --- a/drivers/tty/serial/atmel_serial.c
>>>> +++ b/drivers/tty/serial/atmel_serial.c
>>>> @@ -99,6 +99,7 @@ static void atmel_stop_rx(struct uart_port *port);
>>>>   #define UART_PUT_RTOR(port,v)	__raw_writel(v, (port)->membase + ATMEL_US_RTOR)
>>>>   #define UART_PUT_TTGR(port, v)	__raw_writel(v, (port)->membase + ATMEL_US_TTGR)
>>>>   #define UART_GET_IP_NAME(port)	__raw_readl((port)->membase + ATMEL_US_NAME)
>>>> +#define UART_GET_IP_VERSION(port) __raw_readl((port)->membase + ATMEL_US_VERSION)
>>>>
>>>>    /* PDC registers */
>>>>   #define UART_PUT_PTCR(port,v)	__raw_writel(v, (port)->membase + ATMEL_PDC_PTCR)
>>>> @@ -1503,6 +1504,7 @@ static void atmel_get_ip_name(struct uart_port *port)
>>>>   {
>>>>   	struct atmel_uart_port *atmel_port = to_atmel_uart_port(port);
>>>>   	int name = UART_GET_IP_NAME(port);
>>>> +	u32 version;
>>>>   	int usart, uart;
>>>>   	/* usart and uart ascii */
>>>>   	usart = 0x55534152;
>>>> @@ -1517,7 +1519,22 @@ static void atmel_get_ip_name(struct uart_port *port)
>>>>   		dev_dbg(port->dev, "This is uart\n");
>>>>   		atmel_port->is_usart = false;
>>>>   	} else {
>>>> -		dev_err(port->dev, "Not supported ip name, set to uart\n");
>>>> +		/* fallback for older SoCs: use version field */
>>>> +		version = UART_GET_IP_VERSION(port);
>>>> +		switch (version) {
>>>> +		case 0x302:
>>>> +		case 0x10213:
>>>> +			dev_dbg(port->dev, "This version is usart\n");
>>>> +			atmel_port->is_usart = true;
>>>> +			break;
>>>> +		case 0x203:
>>>> +		case 0x10202:
>>>> +			dev_dbg(port->dev, "This version is uart\n");
>>>> +			atmel_port->is_usart = false;
>>>> +			break;
>>>> +		default:
>>>> +			dev_err(port->dev, "Not supported ip name nor version, set to uart\n");
>>>
>>> it's not really an error a dev_warn is more oppropriate
>>
>> As we are already in -rc5 and that these fixes are critical for at91
>> platforms, I will not re-spin another patch just for this.
>>
>> Moreover, I have the feeling that if we end up in this case, it
>> means that we are in big troubles because the usart/uart included in
>> the product triggering this log is not known (I recall that newer
>> products do not have to hit these lines of code).
>>
>> With these 2 reasons, I prefer to keep my patch like it is.
>>
>> Greg, can you consider taking these two patches as regression fixes
>> for 3.12 (with Tested-by tag from Thomas)?
>
> Is this really a regression from 3.11?

Yes it is. Commit id that I am referring to in patch 1/2 
(055560b04a8cd063aea916fd083b7aec02c2adb8) hit the mainline in 3.12-rc 
time-frame.

> What's the worry about waiting
> for 3.13-rc1, getting this correct, and then backporting them to the
> 3.12-stable trees?

It will break all older at91 in 3.12-final. Moreover, I do think that 
the actual patches are bringing an incorrect solution and I do not plan 
to have a better one (which one?) for 3.13...

> I'd prefer that, so, please clean this up properly and resend it, with
> the tested-by: lines and I'll queue them up for 3.13-rc1.

I do not know what to cleanup. Anyway, tell me if you want that I resend 
the series of 2 patches with the "Tested-by" tag included.

Bye,
-- 
Nicolas Ferre