[PATCH] mmc: mmci: Improve runtime PM support

Linus Walleij linus.walleij at linaro.org
Tue Oct 25 04:53:44 EDT 2011


On Tue, Oct 25, 2011 at 10:05 AM, Adrian Hunter <adrian.hunter at intel.com> wrote:

> The spec does say that it is OK to turn off the clock, but it
> must be turned back on to release the busy signal.  That will always
> happen with the block driver because it always sends a status command
> when waiting.

Aha, thanks for clearing this up!

> Maybe it is an issue with SDIO but currently SDIO is using
> MMC_QUIRK_BROKEN_CLK_GATING by default (although wl1271 is
> using it)

IIRC SDIO is supposed to allow their clock to be cut, if it is a
peripheral that needs constant clocking it should have its own
clockline/crystal or whatever, basically.

However most SDIO device engineers just see the MCLK
and like "hey, clock, I'll just lock my PLL onto that sweet
clock line and save the crystal..." violating the spec. In worst
cases they even drive the entire logic inside the device from
the MCLK line and count on it being perpetually enabled.
Atleast that's how I think about it.

Yours,
Linus Walleij



More information about the linux-arm-kernel mailing list