[PATCH 2/2] serial: amba-pl011: complete support to ZTE uart

[Jun Nie]

2016-03-22 16:20 GMT+08:00 G Gregory <graeme.gregory at linaro.org>:
> On 22 March 2016 at 02:17, Jun Nie <jun.nie at linaro.org> wrote:
>> 2016-03-22 10:06 GMT+08:00 Timur Tabi <timur at codeaurora.org>:
>>> Jun Nie wrote:
>>>>
>>>> That's a fair comment. I suppose all platform need to enable clock for
>>>> the device. For device tree case, we need these code for clock
>>>> enabling.
>>>
>>>
>>> But why do you need to enable the clock?  The pl011 driver works fine for
>>> every platform without your patch.  If the ZTE platform needs to call
>>> clk_get() in order to work, than that is something different about your
>>> platform.
>>
>> ARM/ST pl011 is amba device and use pl011_probe() for initialization,
>> which call these clock enabling code for every device. You introduced
>> platform device driver entry for pl011 and only sbsa use platform
>> device probe method currently. I am trying to reuse this probe
>> function, so need to add clock enabling code in this function.
>>
>> If you did not find any issue in sbsa device tree case, I guess the
>> clock is always enabled. For mobile platform, clock is auto gated for
>> power saving in most of cases.
>>
>>>
>>>> For ACPI case, it may be handled automatically. If you agree
>>>> my opinion, I will move these lines to patch 1. Please correct me if I
>>>> am wrong.
>>>
>>>
>>> I think you need to create a separate patch that adds that call, and make
>>> sure it's tested by other platforms before it gets accepted.
>>
>> These code only impact sbsa device as only sbsa device is initialized
>> via platform driver entry. So only your flavor is need for testing the
>> code. Thanks! :-)
>>
> In SBSA mode the UART clock is always enabled by the firmware and
> there are no baud rate registers so we don't need to know the value of
> the clock. In ACPI there is no method to represent the clock. So if
> you add lines to unconditionally require a clock you will break SBSA
> mode in ACPI.
>
> Graeme

Right, Timur also already point out breakage of ACPI case. So I plan
to move clocking enabling code to device tree case handling. Do you
still have any other concern?

Jun