Following are two patches:

 - Define and use platform_data for davinci_mmc, supporting
   the usual board-specific hooks for sensing card presence,
   the writeprotect switch, and saying if the slot is wired
   to support SD vs just MMC (or potentially 1-wire SDIO).

 - Use that in dm6446 EVM ... which AFAIK is the only board
   in the DaVinci tree with vaguely working MMC support.
   This asks the MSP430 for the relevant signals.

No DM355 EVM support quite yet, but that'll be simple;
I've already had an "unclean" version of it working.

One question I have relates to that second patch:  would
someone volunteer to convert it to something more like
the new drivers/mfd/dm355evm_msp.c code, including the
RTC and keyboard/IR-remote support?  I think this version
would be OK for a start, and I don't plan to write/test
that code myself.  (Been there, done that.)  Maybe you
have a co-worker who would benefit from such a project.


The more interesting question relates to the card detect
logic overall.  Frankly, what's there now seems deeply
broken ... not only does it not work right (at least for
me) even on a dm6446 evm, but it's replicating lots of
work the MMC core does.  And all those commands it's
issuing on its own will prevent the driver from merging
to mainline, too.

So what I'm suggesting is to build on these two patches
(and a third for dm355 evm) and ... rip out ALL the
current card detect logic in davinci_mmc, and rely on
the MMC_CAP_NEEDS_POLL support in the MMC core.

I figure once all that crap is gone, it'll be easier
to get what's left working correctly...

So, comments?  Anyone who's attached to that MMC code
and has a contrary opinion?  Anyone for whom that code
is actually working for card plug/unplug today?

- Dave


_______________________________________________
Davinci-linux-open-source mailing list
[email protected]
http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source

Reply via email to