Re: [PATCH 2/2] Input: tsc2007 - Add a z1_low_threshhold platform data parameter

[Dmitry Torokhov <dmitry.torokhov@gmail.com>]

On Thu, Dec 01, 2011 at 02:19:48PM +0800, Feng Tang wrote:
> On Thu, 1 Dec 2011 13:47:26 +0800
> Dmitry Torokhov <dmitry.torokhov@gmail.com> wrote:
> 
> > On Wed, Nov 30, 2011 at 10:34:18AM +0800, Feng Tang wrote:
> > > On Tue, 29 Nov 2011 17:23:10 +0800
> > > Dmitry Torokhov <dmitry.torokhov@gmail.com> wrote:
> > > 
> > > > On Tue, Nov 29, 2011 at 04:12:58PM +0800, Feng Tang wrote:
> > > > > This originates from a patch in Meego IVI kernel with the name
> > > > > 	linux-2.6.37-connext-0027-tsc2007.patch
> > > > > There is no author info excepte a line "From  MeeGo
> > > > > <kernel@meego.com>"
> > > > > 
> > > > > When integrating tsc2007 on Intel IVI platform, there are a lot
> > > > > of noise data whose z1 value is around 10 which is not a sane
> > > > > "z1". So add a "z1_low_threshhold" to filter those nosie data,
> > > > > to make the device work properly.
> > > > 
> > > > Sounds like a task for userspace to ignore pressure that is too
> > > > low. Bonus points for making it configurable so user can adjust
> > > > sensitivity.
> > > 
> > > Actually, there is one more point :), without this patch, the
> > > driver won't work on our platforms.
> > > 
> > > The tsc2007_soft_irq will keep reading the input data unless there
> > > is no valid pressure data. In our case, those noise data will be
> > > seen as valid data during tsc2007_calculate_pressure(), and the
> > > tsc2007_soft_irq will run endlessly.
> > 
> > Even if we add the pressure threshold would not that noise cause
> > endless stream of interrupts?
> 
> No, there is no endless interrupts for tsc2007. Without the z1
> threshold, the while circle in tsc2007_soft_irq will run endlessly
> as the noise data will be seen as a valid data:
> 
> 		rt = tsc2007_calculate_pressure(ts, &tc);
> 		if (rt == 0 && !ts->get_pendown_state) {
> 			/*
> 			 * If pressure reported is 0 and we don't have
> 			 * callback to check pendown state, we have to
> 			 * assume that pen was lifted up.
> 			 */
> 			break;
> 		}
> 
> With the z1 threshold check, the rt will be 0 for noise data, and the
> code flow broke out.

What I meant is with the threshold check we'll break out of the ISR but
why won't IRQ be raised again?

> 
> > 
> > Also, what kind of z2 is reported with low z1? And do you implement
> > get_pendown_state()?
> 
> z2 seems normal as some data between 3000-4000. We don't have a
> get_pendown_state().

OK, there is max_rt platform parameter. I think we should employ it
instead and break out if we get several incorrect samples in a row. Too
bad you do not have a dedicate method.

-- 
Dmitry