[PATCH] spi: pl022: Add clk_{un}prepare() support in runtime PM

On Wed, Sep 19, 2012 at 5:31 AM, viresh kumar [off-list ref] wrote:

On Tue, Sep 18, 2012 at 5:20 PM, Linus Walleij [off-list ref] wrote:

quoted

On Tue, Sep 18, 2012 at 6:09 AM, viresh kumar [off-list ref] wrote:

quoted

quoted

The amba layer is taking care of interface clock only and not
functional clock. So
i believe that's not the magic code. :)

This clock is the one for the external bus. In some designs these two
clocks are one and the same, and these won't currently get into any clock
disabled states, sadly. (We need to fix that some day.)

I went through the code and found following in amba/bus.c:


static int amba_pm_runtime_suspend(struct device *dev)
{
        struct amba_device *pcdev = to_amba_device(dev);
        int ret = pm_generic_runtime_suspend(dev);

        if (ret == 0 && dev->driver)
                clk_disable(pcdev->pclk);

        return ret;
}

static int amba_pm_runtime_resume(struct device *dev)
{
        struct amba_device *pcdev = to_amba_device(dev);
        int ret;

        if (dev->driver) {
                ret = clk_enable(pcdev->pclk);
                /* Failure is probably fatal to the system, but... */
                if (ret)
                        return ret;
        }

        return pm_generic_runtime_resume(dev);
}

If i am not wrong, these routines also get called with runtiime suspend/resume
of pl022?

Maybe. And that is part of the problem. Check this in
drivers/base/power/runtime.c, rpm_suspend():

        if (dev->pm_domain)
                callback = dev->pm_domain->ops.runtime_suspend;
        else if (dev->type && dev->type->pm)
                callback = dev->type->pm->runtime_suspend;
        else if (dev->class && dev->class->pm)
                callback = dev->class->pm->runtime_suspend;
        else if (dev->bus && dev->bus->pm)
                callback = dev->bus->pm->runtime_suspend;
        else
                callback = NULL;

So as you see the AMBA bus-level runtime PM callbacks will be
called UNLESS there is a class and UNLESS there is a type
and UNLESS there is a voltage domain for this device.

In Ux500 we solved this by calling the AMBA bus-level code
from the voltage domain so it's not completely overridden,
but the semantics are complex here. :-/

(And then we have yet to bring common suspend/resume
into the picture ... which makes is ever more complex.)

If the SPEAr is not using any voltage domains for example,
it'll be unaffected and work fine.

If that is the case, the even the interface clock of pl022 is getting
disabled when not in used.

Yes, hopefully...

And so for Architectures like SPEAr (where functional and interface
clock are controlled
by a single bit), we don't need anything else for power saving, with
respect to clocks.
Isn't it so?

If you have no power domains I hope the ref goes down to
zero and gates off the clock, so yes!

Yours,
Linus Walleij