[Letux-kernel] [PATCH v4 1/8] drivers:input:tsc2007: add new common binding names, pre-calibration, flipping and rotation

H. Nikolaus Schaller hns at goldelico.com
Fri Nov 11 20:01:42 CET 2016


Hi Rob,
finally I found time to catch up. Sorry for the delay.

> Am 31.10.2016 um 04:39 schrieb Rob Herring <robh+dt at kernel.org>:
> 
> On Tue, Oct 18, 2016 at 12:27 PM, H. Nikolaus Schaller
> <hns at goldelico.com> wrote:
>> Hi Rob,
>> 
>>> Am 18.10.2016 um 18:22 schrieb Rob Herring <robh+dt at kernel.org>:
>>> 
>>> On Mon, Oct 17, 2016 at 8:57 AM, H. Nikolaus Schaller <hns at goldelico.com> wrote:
>>>> commit b98abe52fa8e ("Input: add common DT binding for touchscreens")
>>>> introduced common DT bindings for touchscreens [1] and a helper function to
>>>> parse the DT.
>>>> 
>>>> commit ed7c9870c9bc ("Input: of_touchscreen - add support for inverted / swapped axes")
>>>> added another helper for parsing axis inversion and swapping
>>>> and applying them to x and y coordinates.
>>>> 
>>>> Both helpers have been integrated to accommodate any orientation of the
>>>> touch panel in relation to the LCD.
>>> 
>>> Please add the explanation of why this is a compatible change here.
>> 
>> Can you please describe in more detail what you are missing here?
> 
> You stop handling ti,fuzzx for example. So the default driver value
> will be used and this is okay because...

Ok, now I see. Thanks!

> 
> Seems like you would have a regression in behavior if the fuzz values
> specified in an existing DT are ignored.

Well, is there any existing DT using this ti,fuzz property for a tsc2007?

I did not find any use of it in arch/arm/boot/dts.

Since we mainly want to fix arch/arm/boot/dts/omap3-gta04.dts we do not
want to introduce the old ti,fuzzx any more. Rather we want to use the
new "touchscreen-fuzz-x".

Therefore, I thought that it is safe to remove them - and forgot to
mention that anywhere.

If someone shows a device where the ti,fuzz is still used, we can still
add code to override what has been parsed by the common bindings. Or he/she
can change the DT of the device.

> 
>> The patch simply makes use of the generic helper function introduced by the
>> two patches cited above and is therefore automatically compatible if none of
>> the new properties is defined.
>> 
>> The tsc2007 didn't have these features before. So what should we say
>> about compatibility?
> 
> It certainly had the fuzz features before.

> 
> 
>>>> -- ti,max-rt: maximum pressure.
>>>> -- ti,fuzzx: specifies the absolute input fuzz x value.
>>>> -  If set, it will permit noise in the data up to +- the value given to the fuzz
>>>> -  parameter, that is used to filter noise from the event stream.
>>>> -- ti,fuzzy: specifies the absolute input fuzz y value.
>>>> -- ti,fuzzz: specifies the absolute input fuzz z value.

I have now added a note in the commit message that the ti,fuzz properties are
removed as well by this patch.

>>>> +- ti,max-rt: maximum pressure resistance above which samples are ignored
>>>> +  (default: 4095).
>>>> +- ti,report-resistance: report resistance (no pressure = max_rt) instead
>>>> +  of pressure (no pressure = 0).
>>>> +- ti,min-x: minimum value reported by X axis ADC (default 0).
>>>> +- ti,max-x: maximum value reported by X axis ADC (default 4095).
>>>> +- ti,min-y: minimum value reported by Y axis ADC (default 0).
>>>> +- ti,max-y: maximum value reported by Y axis ADC (default 4095).
>>> 
>>> I thought these were going to be common? I think they should be.
>> 
>> Yes, we had discussed that before.
>> 
>> In a second thought, I found that they are not even existing for the different
>> touch chip drivers that are around. And some chip might not need them (because
>> e.g. min/max are constants or defined my means outside the DT or the chip itself
>> can store them in NV memory).
>> 
>> And, since they define ADC raw-values, they are very closely related to the
>> individual chip, so that I am not sure if it is really necessary to make
>> them common. At least if they are properly documented in the bindings for
>> the specific chip they can have different names without disturbing each
>> other.
> 
> They wouldn't disturb each other, but then there is no chance to have
> common parsing either.

What do we need common parsing for these, if they are really specific to the
combination of some touch and exactly this chip?

This is different from specifying a touch dimension or fuzz value in pixels.
Those are passed to the input layer and must be common. But these here are
not and it is the purpose of the driver to hide them from user space (unless
they are used as "reasonable" defaults because DT isn't complete or not up
to date).

> 
>> So I think it is perfectly reasonable to define common bindings towards
>> the input event layer (e.g. size, fuzz, swapping, inversion) and support
>> them by common code which bakes raw coordinates into input events. The latest
>> helper functions already do most of that fully automatic.
> 
> I could argue that size in pixels has no business being in DT.

The touchscreen common bindings define the units as "size in pixels".

> That's a property of the panel the touchscreen is glued to. I guess in some
> cases, the controller is internally scaling the ADC values.

Yes and that scaling is what the tsc2007 driver is doing here because the
controller doesn't internally like for e.g. USB connected touch screens.

> 
>> Next, I had simply copied them from the ads7846 driver where they
>> exist for a long time. So this is sort of "common", at least for two different
>> chips now. See also patch 6/8 of this series which adds the common properties
>> to the ads7846 as well.
>> 
>> For more reference about the existing bindings:
>> 
>> Documentation/devicetree/bindings/input/ads7846.txt
>> 
>> (btw I think someone should move them to bindings/input/touchscreen).
> 
> Patches welcome.

I have added a patch to [PATCH v7].

> 
>> It appears as if there is some overlap of the new generic properties (for x/y
>> swapping) with the old ads7846 properties, but that is something you had
>> already proposed a while ago to make me the driver recognize both properties to
>> stay compatible with older DT files.
>> 
>> So if we now rename the min/max-x/y properties for the tsc2007 we have to rename
>> them for the ads7846 (and maybe others) as well, which might break out-of-tree
>> ads7846 devices or leads to more complex code for handling of property aliases.
> 
> We don't have to rename them for ads7846. That's already baked.

And the longer I think about it, the more I think it is well baked as it is.

> But there's a trend here.

Which trend do you see?

IMHO these adc-min/max properties are very chip specific and there have not been
many new resistive touch controllers announced recently. So I do not see anything
which could become a trend.

> You're defining the same property. Use a common
> name, so when the 3rd binding comes along needing the same thing, we
> already have a common binding.

Why not take the already existing 1st binding of the ads7846 and let the tsc2007
come along as the 2nd one which needs the same thing?

Then we have a common binding between these two now and don't need to wait for
a 3rd one to come (in some uncertain future).

Therefore, I have not yet touched code or description in this area because I don't
know how to do it really better than proposed in [PATCH v4].

[Patch v7] for the whole series will come right after this mail.

BR and thanks,
Nikolaus



More information about the Letux-kernel mailing list