Re: [PATCH v2] clk: qcom: lpasscc: Re-configure the PLL in case lost

From: Doug Anderson
Date: Wed Oct 14 2020 - 22:00:00 EST


Hi,

On Wed, Oct 14, 2020 at 10:21 AM Taniya Das <tdas@xxxxxxxxxxxxxx> wrote:
>
> Thanks Doug for the patch.
>
> On 10/14/2020 9:28 PM, Douglas Anderson wrote:
> > From: Taniya Das <tdas@xxxxxxxxxxxxxx>
> >
> > In the case where the PLL configuration is lost, then the pm runtime
> > resume will reconfigure before usage.
> >
> > Fixes: edab812d802d ("clk: qcom: lpass: Add support for LPASS clock controller for SC7180")
> > Signed-off-by: Taniya Das <tdas@xxxxxxxxxxxxxx>
> > Signed-off-by: Douglas Anderson <dianders@xxxxxxxxxxxx>
> > ---
> > I took the liberty of fixing my own nits that I had with Taniya's
> > patch, AKA:
> >
> > https://lore.kernel.org/r/1602614008-2421-2-git-send-email-tdas@xxxxxxxxxxxxxx
> >
> > Changes in v2:
> > - Don't needlessly have a 2nd copy of dev_pm_ops and jam it in.
> > - Check the return value of pm_clk_resume()
> > - l_val should be unsigned int.
> >
> > drivers/clk/qcom/lpasscorecc-sc7180.c | 23 ++++++++++++++++++++++-
> > 1 file changed, 22 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/clk/qcom/lpasscorecc-sc7180.c b/drivers/clk/qcom/lpasscorecc-sc7180.c
> > index 228d08f5d26f..ee23eb5b9bf2 100644
> > --- a/drivers/clk/qcom/lpasscorecc-sc7180.c
> > +++ b/drivers/clk/qcom/lpasscorecc-sc7180.c
> > @@ -356,6 +356,25 @@ static const struct qcom_cc_desc lpass_audio_hm_sc7180_desc = {
> > .num_gdscs = ARRAY_SIZE(lpass_audio_hm_sc7180_gdscs),
> > };
> >
> > +static int lpass_core_cc_pm_clk_resume(struct device *dev)
> > +{
> > + struct regmap *regmap = dev_get_drvdata(dev);
> > + unsigned int l_val;
> > + int ret;
> > +
> > + ret = pm_clk_resume(dev);
> > + if (ret)
> > + return ret;
> > +
> > + /* Read PLL_L_VAL */
> > + regmap_read(regmap, 0x1004, &l_val);
> > + if (!l_val)
> > + clk_fabia_pll_configure(&lpass_lpaaudio_dig_pll, regmap,
> > + &lpass_lpaaudio_dig_pll_config);
> > +
> > + return 0;
> > +}
> > +
> > static int lpass_core_cc_sc7180_probe(struct platform_device *pdev)
> > {
> > const struct qcom_cc_desc *desc;
> > @@ -373,6 +392,8 @@ static int lpass_core_cc_sc7180_probe(struct platform_device *pdev)
> > if (IS_ERR(regmap))
> > return PTR_ERR(regmap);
> >
> > + dev_set_drvdata(&pdev->dev, regmap);
> > +
> > /*
> > * Keep the CLK always-ON
> > * LPASS_AUDIO_CORE_SYSNOC_SWAY_CORE_CLK
> > @@ -449,7 +470,7 @@ static int lpass_core_sc7180_probe(struct platform_device *pdev)
> > }
> >
> > static const struct dev_pm_ops lpass_core_cc_pm_ops = {
> > - SET_RUNTIME_PM_OPS(pm_clk_suspend, pm_clk_resume, NULL)
> > + SET_RUNTIME_PM_OPS(pm_clk_suspend, lpass_core_cc_pm_clk_resume, NULL)
>
> There are two devices and "lpass_hm_core" and the PLL is not part of the
> HM_CORE, thus was the reason to separate out the pm_ops.

Oh, that's really weird / unexpected. I've tried to disentangle this
in a v3 patch series so I'd be curious to see what people think.
Though it's probably fine to jam the "pm" value like your v1 did I
think it violates the "principle of least surprise" a bit.

-Doug