Re: [PATCH 1/3] drivers: thermal: tsens: Add critical interrupt support

[Amit Kucheria]

On Fri, Nov 29, 2019 at 3:13 AM Bjorn Andersson
<bjorn.andersson@xxxxxxxxxx> wrote:
>
> On Thu 28 Nov 10:46 PST 2019, Amit Kucheria wrote:
>
> > On Wed, Nov 13, 2019 at 1:08 AM Bjorn Andersson
> > <bjorn.andersson@xxxxxxxxxx> wrote:
> > >
> > > On Mon 11 Nov 11:21 PST 2019, Amit Kucheria wrote:
> > >
> > > > TSENS IP v2.x adds critical threshold interrupt support for each sensor
> > > > in addition to the upper/lower threshold interrupt. Add support in the
> > > > driver.
> > > >
> > > > Signed-off-by: Amit Kucheria <amit.kucheria@xxxxxxxxxx>
> > > > ---
> > > >  drivers/thermal/qcom/tsens-common.c | 129 ++++++++++++++++++++++++++--
> > > >  drivers/thermal/qcom/tsens-v2.c     |   8 +-
> > > >  drivers/thermal/qcom/tsens.c        |  21 +++++
> > > >  drivers/thermal/qcom/tsens.h        |  73 ++++++++++++++++
> > > >  4 files changed, 220 insertions(+), 11 deletions(-)
> > > >
> > > > diff --git a/drivers/thermal/qcom/tsens-common.c b/drivers/thermal/qcom/tsens-common.c
> > > > index 4359a4247ac3..2989cb952cdb 100644
> > > > --- a/drivers/thermal/qcom/tsens-common.c
> > > > +++ b/drivers/thermal/qcom/tsens-common.c
> > > > @@ -23,6 +23,10 @@
> > > >   * @low_thresh:     lower threshold temperature value
> > > >   * @low_irq_mask:   mask register for lower threshold irqs
> > > >   * @low_irq_clear:  clear register for lower threshold irqs
> > > > + * @crit_viol:      critical threshold violated
> > >
> > > "violated" as in "temperature is above crit_thresh"?
> >
> > Yes.
> >
> > >
> > > > + * @crit_thresh:    critical threshold temperature value
> > > > + * @crit_irq_mask:  mask register for critical threshold irqs
> > > > + * @crit_irq_clear: clear register for critical threshold irqs
> > > >   *
> > > [..]
> > > > diff --git a/drivers/thermal/qcom/tsens.c b/drivers/thermal/qcom/tsens.c
> > > > index 7d317660211e..784c4976c4f9 100644
> > > > --- a/drivers/thermal/qcom/tsens.c
> > > > +++ b/drivers/thermal/qcom/tsens.c
> > > > @@ -121,6 +121,27 @@ static int tsens_register(struct tsens_priv *priv)
> > > >
> > > >       enable_irq_wake(irq);
> > > >
> > > > +     if (tsens_version(priv) > VER_1_X) {
> > > > +             irq = platform_get_irq_byname(pdev, "critical");
> > > > +             if (irq < 0) {
> > >
> > > Treating this as a fatal error breaks backwards compatibility with
> > > current devicetree; and even within your patch series, tsens should fail
> > > to probe between this patch and the application of patch 3.
> >
> > Good catch.
> >
> > > Please flip this around and do:
> > >
> > > irq = platform_get_irq_byname(pdev, "critical");
> > > if (irq >= 0 && tsens_version(priv) > VER_1_X) {
> > >         request_irq()...
> > > }
> >
> > Won't this still break with current devicetree since irq < 0 until
> > patch 3? Or are you saying we shouldn't check for
> > platform_get_irq_byname() failure?
> >
>
> I'm trying to say that dtsi without "critical" defined should cause the
> driver to simply skip this segment, not fail to initialize.
>
> > I can see two ways out:
> > 1. We patch the dtsi before the code change.
>
> You're expected to maintain backwards compatibility with existing dtb
> files out there. The support for critical interrupt is an additional
> feature, so you should be able to do this by detecting if "critical" is
> defined (e.g. by checking the return value of
> platform_get_irq_byname()).
>
> > 2. We make critical interrupt failure non-fatal by just printing some
> > messages and still returning success.
> >
>
> Try to make it as specific as possible (without adding a bunch of code)
> and throw in a dev_info() if no "critical" is found.

I believe I have now addressed the problem in v2 explicitly overriding
the return value in case of failure in the critical interrupt irq
setup path and simply printing a warning.

In hindsight, critical interrupt support should've been added in the
same series as uplow interrupt support so avoid having to support
"intermediate" DTS file state for one kernel version.

Regards,
Amit