Re: [PATCH] omapdss: Add new panel driver for Topolly td028ttec1LCD.

From: Lars-Peter Clausen
Date: Thu Oct 10 2013 - 14:58:19 EST

On 10/10/2013 03:42 PM, Dr. H. Nikolaus Schaller wrote:

Am 10.10.2013 um 14:26 schrieb Lars-Peter Clausen:

On 10/10/2013 02:13 PM, Tomi Valkeinen wrote:
On 10/10/13 14:52, Dr. H. Nikolaus Schaller wrote:

Yes, I agree and I am willing to help if someone comes up with such a SoC.
At the moment we have connected it to the OMAP3 only.

True, but even without that kind of SoC, SPI bitbanging should be
handled by an SPI driver, not by the drivers that use the bus.

I.e. I want not to do a lot of work for others who we just guess about that they
might exist...

Yep. It's fine for me, it's not that much extra code in the panel driver.

The panel hardware has three wires, so the panel driver (if it does
handle the bus by bitbanging) can only refer to three gpios.

Hm. The panle hardware has 3 but the SoC (OMAP3) the driver
is running on has 4.

Right, but this panel driver is a driver for the panel hardware. Not for
the SoC, or the SoC+panel combination. So the panel driver must only use
resources as seen by the panel hardware.

So either
the bus details should be hidden by using the normal spi framework, or
if the driver does the bitbanging, use the gpios as specified in the
panel spec. The panel driver cannot contain any board specific stuff.

The bitbang driver shown below can handle either 3 or 4 gpios (except
for initialization).

It's not a bitbang driver, it's a panel driver. And anyway, if I
understood right, your use of 4 gpios was just a hack to try to make it
look like a normal 4-wire SPI bus. What you really have is 3 wires, 3
gpios. I don't see any reason to use 4 gpios, as two of them are the same.

There is a protection resistor in the one defined as "output" so that
protocol errors don't have two outputs work against each other.


Hmm, how does it work anyway. Did I understand it right, the panel's
'DIN' pin is connected to two gpios on the SoC, and one of those gpios
is set as output, and the other as input?

Yes.

So the SoC is always pulling
that line up or down, and the panel is also pulling it up or down when
it's sending data. I'm no HW guy but that sounds quite bad =).

Yes, that is the strange thing with this protocol. It does a direction reversal
after some time. I.e. the panel switches its input to output and the SoC has
to be fast enough to do the same. Therefore, we have one output GPIO
(protected by a resistor) and a separate input GPIO.

Sometimes interfacing hardware is indeed strange.


I've never written or studied a bitbanging driver, but shouldn't there
be just one gpio used on the SoC for DIN, and it would be set to either
output or input mode, depending on if we are reading or writing?

Back in the OpenMoko days we used the panel in normal 4-wire SPI mode with
the GPIO bitbang SPI master. The bitbang code in this driver also looks like
normal 4 wire.


Thanks for this hint!

Maybe using a standard 4-wire SPI simply works if we only write data and
make sure the panel is not sending anything...

According to the datasheet the the panel as a dedicated dout pin. Maybe you did not connect it in your design, which means you won't be able to read any data from the panel at all.

Also your custom bitbang code looks a bit strange:

gpio_set_value(data->dout_gpio, 1);
if (gpio_get_value(data->din_gpio) == 0)
return 1;

You set the state on the dout pin and then read the din pin, if both do not match you abort with an error. This suggest that, for whatever reason, you feed the dout pin back into the din pin in your design. Btw. this is also the only place where din is used in your driver.

- Lars
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/