Re: [PATCH] drivers: thermal: tsens: Work with old DTBs

[Amit Kucheria]

On Wed, Dec 11, 2019 at 10:43 PM Stephan Gerhold <stephan@xxxxxxxxxxx> wrote:
>
> Hi Amit,
>
> Thanks for the patch!
>
> On Wed, Dec 11, 2019 at 03:28:33PM +0530, Amit Kucheria wrote:
> > In order for the old DTBs to continue working, the new interrupt code
> > must not return an error if interrupts are not defined.
> >
> > Fixes: 634e11d5b450a ("drivers: thermal: tsens: Add interrupt support")
> > Signed-off-by: Amit Kucheria <amit.kucheria@xxxxxxxxxx>
> > ---
> >  drivers/thermal/qcom/tsens.c | 5 +++--
> >  1 file changed, 3 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/thermal/qcom/tsens.c b/drivers/thermal/qcom/tsens.c
> > index 015e7d2015985..d8f51067ed411 100644
> > --- a/drivers/thermal/qcom/tsens.c
> > +++ b/drivers/thermal/qcom/tsens.c
> > @@ -109,7 +109,7 @@ static int tsens_register(struct tsens_priv *priv)
> >
> >       irq = platform_get_irq_byname(pdev, "uplow");
> >       if (irq < 0) {
> > -             ret = irq;
> > +             dev_warn(&pdev->dev, "Missing uplow irq in DT\n");
> >               goto err_put_device;
> >       }
>
> platform_get_irq_byname() already logs an error if the IRQ cannot be
> found: qcom-tsens 4a9000.thermal-sensor: IRQ uplow not found
>
> To replace that error with a warning (not sure if that is worth it),
> we would need to replace the call with platform_get_irq_byname_optional().

I didn't know about optional, interesting.

> >
> > @@ -118,7 +118,8 @@ static int tsens_register(struct tsens_priv *priv)
> >                                       IRQF_TRIGGER_HIGH | IRQF_ONESHOT,
> >                                       dev_name(&pdev->dev), priv);
> >       if (ret) {
> > -             dev_err(&pdev->dev, "%s: failed to get irq\n", __func__);
> > +             dev_warn(&pdev->dev, "%s: failed to get uplow irq\n", __func__);
> > +             ret = 0;
> >               goto err_put_device;
>
> In case of the old DT, platform_get_irq_byname() will return -ENXIO,
> because no interrupt is specified in the device tree.
> So we should have already run into the error earlier,
> and jumped to "err_put_device".
>
> Is this hunk really necessary?

You're right. Just checking for ENXIO should be enough for old DTs.

> In other words, wouldn't it be enough to do something like
>
> @@ -110,6 +110,8 @@ static int tsens_register(struct tsens_priv *priv)
>         irq = platform_get_irq_byname(pdev, "uplow");
>         if (irq < 0) {
>                 ret = irq;
> +               if (ret == -ENXIO)
> +                       ret = 0;
>                 goto err_put_device;
>         }
>
> ... to essentially ignore only the "IRQ does not exist" condition
> for old device trees?

Thanks for the review.