Grant Likely wrote:
> On Wed, Apr 1, 2009 at 7:41 AM, Wolfgang Grandegger <wg@xxxxxxxxxxxxxx> wrote:
>> Grant Likely wrote:
>>> On Wed, Apr 1, 2009 at 1:51 AM, Wolfgang Grandegger <wg@xxxxxxxxxxxxxx> wrote:
>>>> Grant Likely wrote:
>>>>> The table definition is more verbose this way, but I think it results
>>>>> in more understandable and easier to extend code. It also adds lets
>>>>> the compiler do more type checking for you.
>>>> OK but I don't like the callback function to do the settings. We need
>>>> backward compatibility with old DTS files including the ugly "dfsrr"
>>>> property, right? Then it seems consequent to continue using i2c->flags
>>>> for that purpose and not to introduce another method. If we don't need
>>>> backward compatibility, we could drop the flags completely and just use
>>>> callback functions.
>>> I don't understand why you don't like it. It's an elegant solution
>>> and it simplifies the code somewhat. After grabbing the callback
>>> pointer the compatibility code can simply override it. But I won't
>>> belabor the point or oppose the patch if you stick with the flags
>>> pointer.
>> I changed my mind ;-). Have a look to PATCH v2 I sent out a few minutes ago.
>
> I saw and I like. :-)
>
>>> I've been thinking about this more. These tables are only ever going
>>> to be used by the i2c_mpc driver and so really they are a part of the
>>> i2c_mpc driver itself. Putting them into common code doesn't make any
>>> sense because it is not common code. I will not merge a patch that
>>> puts them into mpc5200 common code.
>> It's not common code, I agree. How about putting it into mpc52xx_i2c.c
>> and use:
>>
>> $ cat Makefile
>> obj-$(CONFIG_I2C_MPC) += mpc52xx_i2c.o
>>
>> Or it could be moved back to the i2c_mpc driver and we add stubs for the
>> functions to get the bus frequency, or use #ifdef's.
>
> I'm happy with either as long as it lives in the same directory as the
> i2c_mpc driver, and as long as the i2c maintainers are okay with it.
Ben, any preference here? See also below.
> Oh, and if you use a separate file it should be statically linked into
> the i2c_mpc module using the i2c_mpc-$(CONFIG_WHATEVER) trick. No
> EXPORT_SYMBOLS should be needed.
Yep, then we would have in "drivers/i2c/busses/Makefile"
obj-$(CONFIG_I2C_MPC) += i2c-mpc.o
i2c-mpc-y = i2c-mpc.o
i2c-mpc-$(CONFIG_FSL_SOC) += i2c-mpc-8xxx.o
i2c-mpc-$(CONFIG_PPC_52xx) += i2c-mpc-52xx.o
And a common include file i2c-mpc.h including the stubs for the arch
dependent functions. But we may get in trouble when we need support for
the MPC512x. I currently tend to add the FDR related code to an extra
file i2c-mpc-fdr.c, plus i2c-mpc.h and use the good old #ifdef's to
select the relevant code.
Wolfgang.
--
To unsubscribe from this list: send the line "unsubscribe linux-i2c" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
