Re: [PATCH 5/8] board-rx51-peripherals: split vaux3 and vmmc2 supplies

[Tony Lindgren <tony@atomide.com>]

* Mark Brown <broonie@opensource.wolfsonmicro.com> [091015 02:01]:
> On Wed, Oct 14, 2009 at 10:15:48AM -0700, Tony Lindgren wrote:
> > * Mark Brown <broonie@opensource.wolfsonmicro.com> [091012 02:18]:
> > > On Mon, Oct 12, 2009 at 11:08:58AM +0300, Eduardo Valentin wrote:
> 
> > > > I'm afraid using dev_name is not that easy. The mmc driver generates device
> > > > name at runtime. That's why this board file setups .dev at runtime as well.
> 
> ...
> 
> > > > So, changing this supply to something static using .dev_name it is not
> > > > possible with current code. That would need refactoring the whole mmc and
> > > > hsmmc setup. And the device naming procedure is dependent on cpu as well.
> > > > Check arch/arm/mach-omap2/device.c:omap2_init_mmc.
> 
> > > same answer each time it's run?  How does this work with the clock API?
> 
> > The clocks are matched using clkdev. Basically the driver just requests
> > functional clock (fck) and interface clock (ick):
> 
> > $ grep mmci arch/arm/*omap*/clock*.c
> > arch/arm/mach-omap1/clock.c:	CLK("mmci-omap.0", "fck",	&mmc1_ck,	CK_16XX | CK_1510 | CK_310),
> > arch/arm/mach-omap1/clock.c:	CLK("mmci-omap.0", "ick",	&armper_ck.clk,	CK_16XX | CK_1510 | CK_310),
> 
> So this is using the standard dev_name based clkdev matching which
> Eduardo said was impossible for the regulators.  Is it just that this
> will actually work fine for the regulators or is there some other magic
> in the OMAP code that joins things up?

Well the mmc regulators are just passed from board-*.c files to
mmc-twl4030.c which does all the low-level init needed. No other
special magic going on.

Eduardro, care to check the dev_name issue one more time?

Regards,

Tony