[Letux-kernel] [PATCH] ARM: dts: am33xx-l4: Add missing touchscreen clock properties

[H. Nikolaus Schaller]

Hi Miquel,

> Am 08.03.2022 um 10:48 schrieb Miquel Raynal <miquel.raynal at bootlin.com>:
> 
> Hi H.,
> 
> hns at goldelico.com wrote on Mon, 7 Mar 2022 20:19:48 +0100:
> 
>> Hi Miquel,
>> some tiny typos...
>> 
>>> Am 07.03.2022 um 12:14 schrieb Miquel Raynal <miquel.raynal at bootlin.com>:
>>> 
>>> When adding support for TI magadc, the MFD driver (common to the  
>> 
>> "magadc"?
> 
> It's actually the name of the hardware block. It stands for 'magnetic
> stripe reader and/or ADC', very much like the first ADC which has a
> specific Touchscreen hardware feature as well. You can wire X lines to
> a touchscreen, and TOTAL - X lines to the ADC, same applies to the
> magnetic stripe reader.
> 
> I can s|magadc|Magnetic Stripe Reader/second ADC| to clarify.

Ok, I see. I had googled and grepped through the kernel logs and found
no hint about "magadc". But if support is newly added, this explains.

> 
>>> touchscreen and the ADC) got updated to ease the insertion of a new DT
>>> node for the ADC, with its own compatible, clocks, etc. Commit
>>> 235a96e92c16 ("mfd: ti_am335x_tscadc: Don't search the tree for our
>>> clock") removed one compatible specific information which was the clock
>>> name, because the clock was looked up from scratch in the DT while this
>>> hardware block was only fed by a single clock, already defined and
>>> properly filled in the DT.
>>> 
>>> Problem is, this change was only validated with an am437x-based board,
>>> where the clocks are effectively correctly defined and referenced. But
>>> on am33xx, the ADC clock is also correctly defined but is not referenced
>>> with a clock phandle as it out to be.  
>> 
>> maybe you mean "ought to be"?
> 
> I knew there was something wrong with it, but not what exactly :)

That is what peer review is good for...

> 
>>> The touchscreen bindings clearly state that the clocks/clock-names
>>> properties are mandatory, but they have been forgotten in one DTSI. This
>>> was probably not noticed in the first place because of the clock
>>> actually existing and the clk_get() call going through all the tree
>>> anyway.
>>> 
>>> Add the missing clock phandles in the am33xx touchscreen description.  
>> 
>> Yes, makes touch on BeagleBoard Black with Chipsee 4"3 panel work again!
>> 
>>> 
>>> Reported-by: H. Nikolaus Schaller <hns at goldelico.com>
>>> Fixes: 235a96e92c16 ("mfd: ti_am335x_tscadc: Don't search the tree for our clock")  
>> Tested-by: H. Nikolaus Schaller <hns at goldelico.com>
> 
> Thanks!
> Miquèl

BR and thanks,
Nikolaus