[PATCH] ARM:MXS: Initial support for the Crystalfontz CFA-10036

[maxime.ripard@free-electrons.com (Maxime Ripard)]

Hi everyone,

This is an attempt for a initial support for the Crystalfontz CFA
10036 module.

The module basically has a iMX28 SoC, plus a micro-SD slot, 3 pins to
access the DUART, a USB OTG port, a OLED screen using the Solomon
SSD1307 controller over I2C and a 200-pins SODIMM port to plug an
expansion board.

Support for this board and the additional devices will come
eventually.

For now, only the DUART and the MMC controller are used.

On the MMC part though, Something odd is going on.  On the first
patches I made against 3.4, everything was working fine. With 3.5-rc2
though, the MMC never replies to the first CMD52 command sent to
it. The pin muxing hasn't changed and has been checked, the card is
well detected through the card detect pin, the power enable pin is
correctly set to a low level, and so the SD card is correctly
powered. The MMC works fine with the exact same kernel on the
imx28-evk though.

What seems odd to me, is that when booting the board with the 3.4
kernel, at boot, the mxs-mmc driver reports a SSP clock of 288MHz,
while on 3.5, it reports a 480MHz SSP clock. By looking at the clk
driver for the iMX28, there is a comment saying that the 480MHz clock
seems to high for SD controlers, and that it should be 288MHz
instead. Given that the clock part has changed quite a lot from 3.4 to
3.5, I think it might be a good lead, but hardcoding the SSP clock to
288MHz in the mxs-mmc driver instead of using the clk_get_rate
function does not resolve the problem.

Also, this problem is seen with both DT-probing and traditionnal
probing.

Do someone have an idea on what could possibly happen here ?

Thanks,
Maxime