On 10/31/2013 06:57 AM, Chen Gang wrote:
> On 10/26/2013 09:18 PM, Chen Gang wrote:
>> On 10/25/2013 01:29 PM, Greg KH wrote:
>>> No, just use the platform-specific SERIAL_PORT_DNFS, instead of having a
>>> copy of it here in this driver, which is just wrong. So please remove
>>> this, and just rely on the system version of this, defining it to
>>> "nothing" if it isn't present, like the 8250 serial driver does it.
>>>
>> Hmm... excuse me, I am not quite familiar with it, I will consult
>> related member.
>>
>> Hello Vineet, I have 2 questions to consult you, please help check:
>>
>> - can arc support this driver? ("drivers/staging/speakup/serialio.c")
No I don' think it is relevant for ARC - but nevertheless as a general rule, it
should still build for ARC.
>> - if arc supports the driver, can arc let it be initialized statically?
>> (not dynamically, e.g. BASE_BAUD).
Like Greg said it needs to be fixed in the driver, just like 8250 does.
If this driver needs a specific definition of SERIAL_PORT_DFNS, that needs to come
from platform's asm/serial.h w/o littering all other arches with an irrelevant ifdef.
> After scan "arch" and "drivers" sub-directory:
>
> - arc and openrisc define BASE_BAUD as non-constant number, and like most of architectures, neither of them provide SERIAL_PORT_DFNS.
>
> - frv and parisc define SERIAL_PORT_DFNS as empty, for all other architectures who already define SERIAL_PORT_DFNS, they also use BASE_BAUD on it.
>
> - for all drivers, they treat SERIAL_PORT_DFNS as 'constant' to initialize static variables statically.
>
> So for me, we need defind SERIAL_PORT_DFNS as empty for arc and
> openrisc, since they can not treate BASE_BAUD as constant number.
No, please fix the driver instead.
-Vineet
_______________________________________________
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxx
http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
