Re: [PATCH] mmc: move regulator handling to core

[Daniel Mack <daniel@caiaq.de>]

On Thu, Dec 03, 2009 at 04:27:39PM +0200, Adrian Hunter wrote:
> gDaniel Mack wrote:

[...]

> > drivers/mmc/core/core.c   |   36 ++++++++++++++++++++----------------
> > drivers/mmc/core/host.c   |    3 +++
> > drivers/mmc/host/mmci.c   |   28 ++++++++++++----------------
> > drivers/mmc/host/mmci.h   |    1 -
> > drivers/mmc/host/pxamci.c |   20 ++++++++------------
> > include/linux/mmc/host.h  |   10 ++++++----
> 
> What about arch/arm/mach-omap2/mmc-twl4030.c ?

Argh, missed that one. And this particular case doesn't fit to my
modifications. I don't know the code well ...  We would need to
have a struct mmc_host * in all the functions there calling
mmc_regulator_{set,get}_ocr. Any idea how to resolve that?

> >--- a/drivers/mmc/core/host.c
> >+++ b/drivers/mmc/core/host.c
> >@@ -18,6 +18,7 @@
> > #include <linux/leds.h>
> > #include <linux/mmc/host.h>
> >+#include <linux/regulator/consumer.h>
> > #include "core.h"
> > #include "host.h"
> >@@ -154,6 +155,8 @@ void mmc_remove_host(struct mmc_host *host)
> > 	mmc_remove_host_debugfs(host);
> > #endif
> >+	regulator_put(host->vcc);
> >+
> 
> If the core is doing a 'regulator_put()' shouldn't it also be doing
> a 'regulator_get()'?  Why not leave it to the drivers?

Yes, I can change the patch to do that, no problem. The major reason why
I didn't put the regulator_get() to the mmc core is that I need to have
the platform_device to obtain its name.

Daniel