lkml.org 
[lkml]   [2013]   [Oct]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH V3] clk: palmas: add clock driver for palmas
On 10/08/2013 12:08 PM, Stephen Warren wrote:
> On 10/08/2013 10:14 AM, Nishanth Menon wrote:
>> On 10/08/2013 09:39 AM, Laxman Dewangan wrote:
>>> Thanks Nishanth for review.
>>>
>>> On Tuesday 08 October 2013 06:59 PM, Nishanth Menon wrote:
>>>> On 10/08/2013 08:21 AM, Laxman Dewangan wrote:
>>>>> Palmas devices has two clock output CLK32K_KG and CLK32K_KG_AUDIO
>>>> not all palmas devices have 2 clocks - example: tps659038
>>>
>>> This is for generic palmas and I have seen it for TPS65913, TPS65914,
>>> TPS80036. If the generic one is not compatible then it need to add
>>> device specific and at that time, it is require to update the binding
>>> document accordingly.
>>
>> ?? you do have two clocks inside the device they should be represented
>> as two compatible entities - that simplifies everyone's life.
>
> I think the terminology you're using here is quite confusing.
>
> Are you talking about having two different compatible values for two
> different HW designs, where those different designs implement different
> sets of clocks (which makes sense), or two different DT nodes for two
> different clocks (which IMHO doesn't always, unless those different
> clocks *truly* are separate IP blocks with completely independent
> register regions, and where those IP blocks are likely to be re-used
> as-is in other chips).
>
clk32k and clk32k_audio are two different resources and since these
are two different resource instances - a "compatible" matching an
actual device is my suggestion.

clk32k and clk32k_audio are two different resources because they have
their specific set of controls registers and may even be independently
present in a Palmas variant.

To highlight this: The example of tps659038 where clk32k is not
present, but clk32k_audio is present (and happens to be disabled by
default - thanks to an OTP on the chip - on platform like DRA7-evm, it
is used to for 32k clk for wlan -currently hacked in u-boot using
plain i2c writes[1] - yes it is yucky).

Obviously, there are many ways to implement this. based on the current
implementation, it indicates that if i create a node with
"ti,palmas-clk" -i'd create two clocks - that is wrong for tps659038.

Now (with the current approach), if I have to create a one clock for
tps659038, i have to fix the for adding clock providers, add up
"ti,tps659038-clk" etc.. it is doable - but IMHO, I dont need to do it
with only the relevant nodes in dts.

Further, it has no way to indicate that device X uses clock Y using
clocks =<&xyz> either.

[1] DRA7-evm u-boot hack.
i2c dev 0
i2c probe
i2c mw 0x58 0xfb 0x2b (select secondary function for GPIO_5)
i2c mw 0x58 0xd5 0x01 (enable audio clock)

--
Regards,
Nishanth Menon


\
 
 \ /
  Last update: 2013-10-08 20:01    [W:0.113 / U:0.824 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site