Re: [PATCH v2 2/2] mmc: dw_mmc: Add the ability to set the ciu clock frequency

[Doug Anderson <dianders@chromium.org>]

Jaehoon,

On Mon, Jun 17, 2013 at 9:51 PM, Jaehoon Chung <jh80.chung@samsung.com> wrote:
> Hi Doug,
>
> I have one question for using <clock-frequency>.
> I found the fixed-rate-clocks feature.
> If we want to set <clock-frequency>, then can we use the fixed-rate-clocks?
> i'm not sure how use the fixed-rate-clocks. but it seems to set fixed-rate value for clock frequency.
>
> clk_set_rate() didn't ensure to set the <clock-frequency> value.

I'm not sure I understand the question.  I don't think that the
fixed-rate-clocks have a close relation to the clock-frequency or the
ciu clock.  The fixed-rate-clock entries for a board usually specify
the root clock source for a board.  For instance in exynos5250-snow
you can see:

fixed-rate-clocks {
  xxti {
    compatible = "samsung,clock-xxti";
    clock-frequency = <24000000>;
  };
};

Other clocks in the board are derived from this clock through PLLs,
muxes, dividers, gates, etc.  On 5250 we have:

 fin_pll (xxti) -> fout_mpll -> fout_mplldiv2 -> mout_mpll_fout ->
 sclk_mpll -> sclk_mpll_user -> mout_mmc1 -> div_mmc1
 div_mmc_pre1 -> sclk_mmc1

In 5250 the ciu clock for mmc1 is sclk_mmc1, which is a simple gate.
When you "enable" this clock it, ungates it.  The sclk_mmc1 has the
flag CLK_SET_RATE_PARENT on it.  That means when you try to set the
rate it will involve the parent clock (div_mmc_pre1).  The parent
clock also has CLK_SET_RATE_PARENT, so it can also involve div_mmc1.
I haven't dug through to see how the clock framework splits up divides
between div_mmc1 and div_mmc_pre1, but it's supposed to handle that.

We don't allow clk_set_rate to percolate any higher (no
CLK_SET_RATE_PARENT at mout_mmc1).

-Doug