[PATCH v2 0/3] Input: rotary-encoder - use more than two gpios

Daniel Mack daniel at zonque.org
Tue Feb 2 04:08:07 PST 2016


Hi Uwe,

Thanks for the update.

On 02/02/2016 11:24 AM, Uwe Kleine-König wrote:
> Some time ago I sent a v1 of this, now after testing the changes more
> deeply patch 3 changed a bit. The old series started with
> 
> 	Date: Wed,  2 Dec 2015 11:07:11 +0100                                                                                               
> 	From: Uwe Kleine-König <u.kleine-koenig at pengutronix.de>                                                                             
> 	Subject: [PATCH RFC 0/3] input: rotary_encoder: use more than two gpios as input                                                    
> 	Message-Id: <1449050834-31779-1-git-send-email-u.kleine-koenig at pengutronix.de>                                                      
> 
> The two first patches are just preparation for the third patch.
> 
> There is an obvious improvement that allows detection of quick changes
> more reliably with >2 gpios, but I didn't implement this yet. (With 4
> GPIOs you can distinguish a counter clockwise movement of three states
> from a clock wise movement of a single state. Still the patch is useful
> as it makes these devices work at all.
> 
> My test device looks as follows:
> 
>         rotary at 0 {
>                 compatible = "rotary-encoder";
>                 gpios = <&gpio4 12 GPIO_ACTIVE_HIGH>, <&gpio4 11 GPIO_ACTIVE_HIGH>, <&gpio4 10 GPIO_ACTIVE_HIGH>, <&gpio4 9 GPIO_ACTIVE_HIGH>;
> 
>                 rotary-encoder,steps = <16>;
> 		rotary-encoder,steps-per-period = <16>;
>         };
> 
> While Daniel Mack and Rojhalat Ibrahim agreed that this device is an
> absolute encoder and should be supported by a simpler logic, I still
> consider it worthwhile to get these patches in as a first step. Also the
> binding looks right, so IMHO the comments shouldn't stop this series
> from going in.

I still don't understand why this is implemented that way, rather than
going for a much simpler logic of interpretation that also allows users
to read out the absolute position.

The code to read the value would be really just as simple as reading all
GPIOs and or-ing their values into the result, and skip the state
machine completely. This code would be active if a new attribute
(something like 'rotary-encoder,hardware-absolute') is set, or even
implicitly, when more than 2 GPIOs are specified.

Is there any reason for not doing that?


Thanks,
Daniel




More information about the linux-arm-kernel mailing list