Re: [PATCHv5] staging/iio/adc: change the MXS touchscreen driver implementation

[Marek Vasut <marex@denx.de>]

Dear Jürgen Beisert,

> Hi Marek,
> 
> On Monday 23 September 2013 09:55:58 Jürgen Beisert wrote:
> > On Monday 23 September 2013 01:55:04 Marek Vasut wrote:
> > > > The following series replaces the current busy loop touchscreen
> > > > implementation for i.MX28/i.MX23 SoCs by a fully interrupt driven
> > > > implementation.
> > > > 
> > > > Since i.MX23 and i.MX28 silicon differs, the existing implementation
> > > > can be used for the i.MX28 SoC only.
> > > > 
> > > > So, the first two patches of this series move the i.MX28 specific
> > > > definitions out of the way. The third patch simplifies the register
> > > > access to make it easier to add the i.MX23 support. Then the i.MX23
> > > > specific definitions are added, also the code to distinguish both
> > > > SoCs at run-time. Up to here the existing touchscreen driver will
> > > > now run on an i.MX23 Soc as well.
> > > > 
> > > > When the i.MX23 SoC is running from battery it seems not to be a good
> > > > idea to run a busy loop to detect touches and their location. The
> > > > fourth patch adds a fully interrupt driven implementation which makes
> > > > use of the built-in delay and multiple sample features of the
> > > > touchscreen controller. This will reduce the interrupt load to a
> > > > minimum.
> > > > 
> > > > The next to last patch in this series just removes the existing busy
> > > > loop implementation.
> > > > 
> > > > The last patch adds a devicetree bindings proposal (to be discussed).
> > > > 
> > > > Some restrictions/questions:
> > > > 
> > > > - the touchscreen part is yet tested on i.MX23 SoC only
> > > > - has someone a good idea how to implement a reliable pressure
> > > > measurement if the resistances of the touchscreen's plates are
> > > > unknown?
> > > > 
> > > > Changes since v4:
> > > > 
> > > > - honor Jonathan's comments about function names
> > > > - honor Dmitry's comments about workqueue canceling and interrupts
> > > > - adding devicetree bindings proposal
> > > > 
> > > > Changes since v3:
> > > > 
> > > > - split adding register access functions and i.MX23 support into two
> > > > patches
> > > > 
> > > > Changes since v2:
> > > > 
> > > > - useless debug output removed
> > > > 
> > > > Changes since v1:
> > > > 
> > > > - adding register access functions to make the existing code more
> > > > readable - adding some functions to distinguish the SoCs at run-time
> > > > to avoid if-else contructs whenever differences in the register
> > > > layout between i.MX23 and i.MX28 must be handled
> > > 
> > > Just tested this on M28EVK, the touchscreen is dead. I do not get any
> > > touch events when I use ts_calibrate and when I hd /dev/input/eventX ,
> > > I get nothing either.
> > > 
> > > Right now, I'm somehow on a tight schedule, but I'd like to see this
> > > resolved ASAP. Do you have any hint for me ?
> > 
> > Seems the very first touch detection does not work. A colleague just told
> > me, we have an i.MX28 based hardware here and maybe a touchscreen from a
> > different platform which maybe can be used with the i.MX28 platform. Stay
> > tuned...
> 
> I also found an MX28EVK with a touchscreen here and: yes it does not work
> with the modified driver. But now I'm confused, because of Lothar's
> statement about the TX28 platform.
> 
> @Lothar: What kernel revision did you use? I had to go back to 3.9 to get a
> working kernel on the MX28EVK. More recent vanilla kernels stop working
> (MMC and network fail here).

Can you fix the SD and Network or at least show the fail log ?

btw. I use 3.12-rc1 (next/master) to test the touchscreen on M28EVK. I don't 
have the LCD module attached to MX28EVK, so I can't test there.

Best regards,
Marek Vasut