Re: [RESEND PATCH v6 3/3] input: pm8xxx-vibrator: add new SPMI vibrator support

[Dmitry Torokhov <dmitry.torokhov@gmail.com>]

Hi Fenglin,

On Thu, Mar 28, 2024 at 02:52:32PM +0800, Fenglin Wu wrote:
> 
> 
> On 2023/10/1 0:17, Dmitry Torokhov wrote:
> > On Mon, Sep 25, 2023 at 10:54:45AM +0800, Fenglin Wu wrote:
> > > 
> > > 
> > > On 9/24/2023 3:07 AM, Dmitry Baryshkov wrote:
> > > > > +
> > > > > +       switch (vib->data->hw_type) {
> > > > > +       case SSBI_VIB:
> > > > >                   mask = SSBI_VIB_DRV_LEVEL_MASK;
> > > > >                   shift = SSBI_VIB_DRV_SHIFT;
> > > > > +               break;
> > > > > +       case SPMI_VIB:
> > > > > +               mask = SPMI_VIB_DRV_LEVEL_MASK;
> > > > > +               shift = SPMI_VIB_DRV_SHIFT;
> > > > > +               break;
> > > > > +       case SPMI_VIB_GEN2:
> > > > > +               mask = SPMI_VIB_GEN2_DRV_MASK;
> > > > > +               shift = SPMI_VIB_GEN2_DRV_SHIFT;
> > > > > +               break;
> > > > > +       default:
> > > > > +               return -EINVAL;
> > > > Could you please move the switch to the previous patch? Then it would
> > > > be more obvious that you are just adding the SPMI_VIB_GEN2 here.
> > > > 
> > > > Other than that LGTM.
> > > 
> > > Sure, I can move the switch to the previous refactoring patch.
> > 
> > Actually, the idea of having a const "reg" or "chip", etc. structure is
> > to avoid this kind of runtime checks based on hardware type and instead
> > use common computation. I believe you need to move mask and shift into
> > the chip-specific structure and avoid defining hw_type.
> > 
> > Thanks.
> 
> Hi Dmitry,
> 
> The v7 changes have been pending for a while, I am not sure if you are still
> insist on this. As I explained, I actually did it this way in v2 and it got
> updated to this by following other comments.
> 
> Can you respond and tell me if you prefer changes similar to v2? I can
> update and push v8 by following your suggestion.
> 
> v7: https://lore.kernel.org/linux-arm-msm/20231108-pm8xxx-vibrator-v7-0-632c731d25a8@quicinc.com/
> 
> v2: https://lore.kernel.org/linux-arm-msm/20230718062639.2339589-3-quic_fenglinw@quicinc.com/

Yes, I believe what you had in v2 was better, and Dmitry Baryshkov's
comments on v2 were also great.

You can have 2 styles of code - you have a hw type for each regulator
and then use it to do conditional logic in the code. If you do it this
way you and you need to add a new device type or model you have to go
through the code and validate all the checks. Or you could have a
structure that is defined flexibly enough to cover all existing
permutations, and you rely on the data in it to control the behavior.
You should not mix the 2 styles, as this just makes the code more
confusing.

Thanks.

-- 
Dmitry