Addy,

On Wed, Aug 13, 2014 at 6:57 PM, Addy [off-list ref] wrote:

quoted

I think maybe it is suitable as follows:
mmc0 = &sdmmc
mmc1 = &sdio0
mmc2 = &sdio1
mmc3 = &emmc

Right, except the only ones that have landed in Heiko's tree are sdmmc
and emmc, so we can't do sdio0 and sdio1 yet.  You could post support
for sdio0 and adio1?

yes, I will post it today.

Also: it's really handy if emmc is 0.  See below: I don't think it's
great to use the ID to find the sysconfig registers.

quoted

So we can get ctrl_id:
ctrl_id = of_alias_get_id(host->dev->of_node, "mshc");

Somehow I hadn't realized that was there.  I guess we could use that
too.  I'd vote to remove that and use the standard "mmc" numbering
(and get some momentum to land those patches).  If you want I'll
repost using the mshc stuff, though.

quoted

and can get offset of registers:
offset = 0x200 + ctrl_id * 8 + 4 * drive_or_sample

I thought the plan was to actually implement the phase stuff as a clock driver.

...even if we didn't, I'd rather not rely on ID like this to find the
right address.  It's really non-obvious.