Re: [PATCH v2 22/66] media: sun6i-csi: Use runtime pm for clocks and reset

[Paul Kocialkowski <paul.kocialkowski@bootlin.com>]

[-- Attachment #1: Type: text/plain, Size: 4103 bytes --]

Hi,

On Wed 09 Feb 22, 10:22, Maxime Ripard wrote:
> On Sat, Feb 05, 2022 at 07:53:45PM +0100, Paul Kocialkowski wrote:
> > Wrap the clock and reset preparation into runtime pm functions
> > for better organization of the code.
> > 
> > Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
> > ---
> >  .../platform/sunxi/sun6i-csi/sun6i_csi.c      | 94 ++++++++++++++-----
> >  1 file changed, 69 insertions(+), 25 deletions(-)
> > 
> > diff --git a/drivers/media/platform/sunxi/sun6i-csi/sun6i_csi.c b/drivers/media/platform/sunxi/sun6i-csi/sun6i_csi.c
> > index 2355088fdc37..b53b92b70a89 100644
> > --- a/drivers/media/platform/sunxi/sun6i-csi/sun6i_csi.c
> > +++ b/drivers/media/platform/sunxi/sun6i-csi/sun6i_csi.c
> > @@ -152,40 +152,18 @@ int sun6i_csi_set_power(struct sun6i_csi_device *csi_dev, bool enable)
> >  
> >  	if (!enable) {
> >  		regmap_update_bits(regmap, CSI_EN_REG, CSI_EN_CSI_EN, 0);
> > +		pm_runtime_put(dev);
> >  
> > -		clk_disable_unprepare(csi_dev->clk_ram);
> > -		clk_disable_unprepare(csi_dev->clk_mod);
> > -		reset_control_assert(csi_dev->reset);
> >  		return 0;
> >  	}
> >  
> > -	ret = clk_prepare_enable(csi_dev->clk_mod);
> > -	if (ret) {
> > -		dev_err(csi_dev->dev, "Enable csi clk err %d\n", ret);
> > +	ret = pm_runtime_resume_and_get(dev);
> > +	if (ret < 0)
> >  		return ret;
> > -	}
> > -
> > -	ret = clk_prepare_enable(csi_dev->clk_ram);
> > -	if (ret) {
> > -		dev_err(csi_dev->dev, "Enable clk_dram_csi clk err %d\n", ret);
> > -		goto clk_mod_disable;
> > -	}
> > -
> > -	ret = reset_control_deassert(csi_dev->reset);
> > -	if (ret) {
> > -		dev_err(csi_dev->dev, "reset err %d\n", ret);
> > -		goto clk_ram_disable;
> > -	}
> >  
> >  	regmap_update_bits(regmap, CSI_EN_REG, CSI_EN_CSI_EN, CSI_EN_CSI_EN);
> >  
> >  	return 0;
> > -
> > -clk_ram_disable:
> > -	clk_disable_unprepare(csi_dev->clk_ram);
> > -clk_mod_disable:
> > -	clk_disable_unprepare(csi_dev->clk_mod);
> > -	return ret;
> >  }
> >  
> >  static enum csi_input_fmt get_csi_input_format(struct sun6i_csi_device *csi_dev,
> > @@ -800,6 +778,65 @@ static irqreturn_t sun6i_csi_isr(int irq, void *private)
> >  	return IRQ_HANDLED;
> >  }
> >  
> > +static int sun6i_csi_suspend(struct device *dev)
> > +{
> > +	struct sun6i_csi_device *csi_dev = dev_get_drvdata(dev);
> > +
> > +	reset_control_assert(csi_dev->reset);
> > +	clk_disable_unprepare(csi_dev->clk_ram);
> > +	clk_disable_unprepare(csi_dev->clk_mod);
> > +	clk_disable_unprepare(csi_dev->clk_bus);
> > +
> > +	return 0;
> > +}
> > +
> > +static int sun6i_csi_resume(struct device *dev)
> > +{
> > +	struct sun6i_csi_device *csi_dev = dev_get_drvdata(dev);
> > +	int ret;
> > +
> > +	ret = reset_control_deassert(csi_dev->reset);
> > +	if (ret) {
> > +		dev_err(dev, "failed to deassert reset\n");
> > +		return ret;
> > +	}
> > +
> > +	ret = clk_prepare_enable(csi_dev->clk_bus);
> > +	if (ret) {
> > +		dev_err(dev, "failed to enable bus clock\n");
> > +		goto error_reset;
> > +	}
> > +
> > +	ret = clk_prepare_enable(csi_dev->clk_mod);
> > +	if (ret) {
> > +		dev_err(dev, "failed to enable module clock\n");
> > +		goto error_clk_bus;
> > +	}
> > +
> > +	ret = clk_prepare_enable(csi_dev->clk_ram);
> > +	if (ret) {
> > +		dev_err(dev, "failed to enable ram clock\n");
> > +		goto error_clk_mod;
> > +	}
> > +
> > +	return 0;
> 
> You've change the order of the reset vs clock initialization.

Oh right. Is there a preference regarding the order?
It probably makes more sense to release reset once all the clocks are
ready rather than the opposite (so that would match the previous order).

> Also, does that work with CONFIG_PM disabled?

It doesn't, but the driver should really depend on PM anyway (it doesn't yet).
That makes me realize that I didn't cleanup the Kconfig entry, which
should have been updated along the way.

Will fix both next time.

Thanks for the review!

Paul

-- 
Paul Kocialkowski, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]