Re: Driver support for TI's quadrature encoder module

On 09/09/11 13:05, AnilKumar, Chimata wrote:

Hi All,

Thanks for quick reply.

On Fri, Sep 09, 2011 at 17:09:09, Jonathan Cameron wrote:

quoted

On 09/08/11 17:50, Jonathan Cameron wrote:

quoted

On 09/08/11 15:54, Greg KH wrote:

quoted

On Thu, Sep 08, 2011 at 06:57:04PM +0530, AnilKumar, Chimata wrote:

quoted

Hi All,

I am developing a driver for TI's enhanced Quadrature Encoder Pulse
(eQEP) module. Details of the module at 
http://www.ti.com/lit/ug/sprufu8/sprufu8.pdf

Little description about the module:-

This module is used for direct interface with a linear or rotary 
incremental encoder.

Main features of the module:
a. Provides the direction of the motor (clock or anti-clock) b. 
Provides position count values, which can be used to compute speed
   of a machine (motor/generator)
c. Generate a sync output from position compare sub-module d. Can be 
used for High & Low speed measurements

This module has 4 inputs from external rotary/linear encoder.
a. Input A (Input from external rotary encoder, a pulse train) b. 
Input B (Input from external rotary encoder, a pulse train) c. Index 
Event (Input from external rotary encoder, a index event.
   This can be used as output, which can be generated after
   position compare)
d. Strobe Event (Input from external rotary encoder, a strobe event.
   This can be used as output, which can be generated after
   position compare)

The module provides pre-analyzed data based on the above four input 
signals. We can use the data directly for knowing the position and 
direction details.

I have few questions related to this:-

1. What subsystem this driver fit into? Input subsystem or UIO 
subsystem?

Why I think it may fit into UIO subsystem?
a. Same driver is used for lot of applications like motor control, 
volume control and menu navigation. This means we have to provide 
lot of flexibility to the user in configuring the hardware to work 
for different applications.

Does it use the UIO interface to userspace?  If not, then it's not a 
UIO driver.

quoted

Why I think it may fit into Input subsystem?
a. The module is receiving the signals from external encoder. Based 
on the signals it will generate some events, which we have to handle 
at the user space.
b. I found some encoder drivers are already present under input 
subsystem (example drivers/input/misc/rotary-encoder.c)

But, the current rotary encoder drivers are developed for much 
simpler hardware.

It looks like we need some additional interfaces to cater to the 
hardware capabilities. I am thinking of sysfs interfaces which allow 
user to directly configure the hardware for position control, 
position compare, input source selection, capture timer period, 
speed and direction of rotation.

That sounds like you want the IIO interface in drivers/staging/iio/ 
right?  That api handles all sorts of stuff like this.

Certainly possible - always happy to support new device types.

We have some resolvers in there already, and based on a hundred mile 
view this is similar...  From what I recall those interfaces are 
pretty messy right now, but looking at it another way, that gives us 
more flexibility to mess around with them to ensure they cover your part as well.

I'll rephrase this having taken a look. The current resolver drivers conform to our standard interface in now way whatsoever.  So whatever you do, don't copy them!  Along with meters drivers, those are the areas we haven't cleaned up at all yet (also a smattering of other drivers in terrible state for one reason or another elsewhere)

Jonathan, Thanks for the pointers

I did not know about IIO interface so far.
Will study it and come up with some patches
for my device based on this interface.

Excellent.  Please do read the abi docs carefully in particular.
They don't cover this type of device, but will give you pointers
on what abi makes sense.  Posting that abi even before your driver
is ready to linux-iio will get that conversation underway (can often
take much longer to pin down interface naming etc
than reviewing/fixing the actual code!)

Thanks,

Jonathan