Re: [Gumstix-users] Bug in drivers/serial/pxa.c patch

[Craig H. <cr...@gu...>]

On Jul 2, 2006, at 12:35 PM, Dave Hylands wrote:

> Hi Darren,
>
>> The thing I've been confused about from the gitgo, is what exactly
>> serial_pxa_verify_port() is supposed to be verifying.  pxa.c is not
>> exactly bursting with comments.  In the upatched version,
>> serial_pxa_verify_port() always returns -EINVAL which was causing my
>> custom_divisor to be rejected.  I had originally changed it to always
>> return 0.  Craig patched it to do the baud rate checks, but that
>> broke the custom_divisor again.  Should I go back to just returning
>> zero?
>
> So I poked around a few other drivers to see what they're verify_port
> routines do, and basically, it just seems to be doing a few basic
> validity checks.
>
> These checks are typically:
> - Verify that the port type has been filled in properly
> - Verify that the IRQ is valid
> - Verify that baud base is valid
> - Verify that the port is setup as memory mapped, if only memory
> mapped support is available
>
> I think that the reason for verify_port is when a platform can have
> multuple types of serial ports, or variable multiples of the same
> serial port.
>
> So, I think that returning -EINVAL is wrong, and I think trying to
> check baud rates is wrong. Checking uartclk/16 == baud_base seems
> fine.
>
> I'll ask on the arm-linux list.

My intention was precisely to deal with the difference between FFUART/ 
STUART vs BTUART/HWUART which are subtly different on the PXA -- I  
wanted to check that if you are setting the baud rate on a UART, that  
it's in range for that UART.

C