[PATCH 09/19] drivers: serial: add support for Samsung S5PC110 SoC uart

[jassi brar]

On Thu, Nov 19, 2009 at 9:09 PM, Marek Szyprowski
<m.szyprowski at samsung.com> wrote:
> Hello,
>
> On Thursday, November 19, 2009 12:49 PM Mark Brown wrote:
>
>> On Thu, Nov 19, 2009 at 11:38:41AM +0000, Russell King - ARM Linux wrote:
>>
>> > I'm not sure what you're commenting on precisely, but the Samsung code as
>> > a whole doesn't use the clk API very well, and I suspect that is starting
>> > to cause people to have to pass clock names around.
>>
>> That's what I'd thought originally but it's not quite the problem here.
>> The issue is that Samsung serial port clock is chosen from a mux and the
>> logic to select which of the inputs to that mux should be used is in the
>> serial driver.  This means that if the set of parent clocks changes then
>> the serial driver needs to be told what they are.
>>
>> My suggestion was to push this logic down into the clock API so the
>> serial driver just requests a rate and then clock API picks the best
>> option from the mux.  As well as being nicer from the clock API point of
>> view this would also allow other drivers to use the same logic since
>> these muxes are a standard idiom for Samsung SoCs.
>
> What if a device does not operate properly when clocked from one of the
> available clock sources (lets say machine specific problem)? There must
> be a way of setting which clock(s) should not be considered when calculating
> best choice for particular device.
That shudn't be a concern. for example, if some machine doesn't provide clock
at some PAD, which could be used as a source, then the clock API wud read the
clock as 0 and the calculations wudn't pick that source up. Likewise, any such
'unusable' clock in the machine wud be marked by having invalid(0) rates.
The problem is moving this uart-mux into the the Clock API.