[PATCHv2 3/7] ARM: imx: add support code for IMX50 based machines

[Greg Ungerer]

Hi Shawn,

On 01/11/13 18:21, Shawn Guo wrote:
> On Fri, Nov 01, 2013 at 11:59:10AM +1000, Greg Ungerer wrote:
>> On 31/10/13 16:10, Shawn Guo wrote:
>>> On Tue, Oct 29, 2013 at 03:15:53PM +1000, gerg at uclinux.org wrote:
>>>> +DT_MACHINE_START(IMX50_DT, "Freescale i.MX50 (Device Tree Support)")
>>>> +	.map_io		= mx53_map_io,
>>>> +	.init_early	= imx53_init_early,
>>>> +	.init_irq	= mx53_init_irq,
>>>> +	.handle_irq	= imx53_handle_irq,
>>>> +	.init_machine	= imx50_dt_init,
>>>> +	.init_late	= imx53_init_late,
>>>> +	.dt_compat	= imx50_dt_board_compat,
>>>> +	.restart	= mxc_restart,
>>>> +MACHINE_END
>>>
>>> The imx53 is a platform that was converted from non-DT to DT.  Some of
>>> these imx53 hooks may not even necessary for imx53 now, since imx53
>>> supports DT boot only today.  So I do not prefer to reuse these imx53
>>> hooks on the brand new imx50 DT machine.  I would suggest to maintain a
>>> mach-imx50.c and machine descriptor for imx50 with only the necessary
>>> hooks implemented as local functions inside mach-imx50.c.  After imx53
>>> machine descriptor and hooks gets cleaned up, we can look at how we can
>>> consolidate these descriptors if they are pretty much in common.
>>
>> With the current support I could drop .init_early and .init_late
>> I believe. All the other mx53 functions are needed to successfully
>> boot.
>>
>> imx53_handle_irq is easy to fix, it is just defined to be
>> tzic_handle_irq anyway.
>>
>> mx53_map_io and mx53_init_irq are coded in mm-imx5.c. They look strait
>> forward. Do you want me to create imx50 specific versions in there?
> 
> Oh, right.  I overlooked the fact that the static mapping created in
> mx53_map_io() is still used by clock driver everywhere.  It does not
> make too much sense to clean them up before imx51 gets converted to be
> DT only.
> 
> Okay, I'm fine with reusing some of the imx53 hooks as required.  But
> at least imx53_init_early() shouldn't be reused on imx50, because it
> sets cpu type specifically for imx53, and calls mxc_iomux_v3_init()
> which is not needed for imx50 (even for imx53 as well today).

Yep, that sounds right. I'll modify that patch and resend.

Thanks
Greg