Re: [PATCH 1/2] pinctrl: mediatek: Add mtk_eint_pm_ops to common-v2

[Nicolas Boichat]

On Sat, May 4, 2019 at 2:09 AM Sean Wang <sean.wang@xxxxxxxxxx> wrote:
>
> Hi, Nicolas
>
> On Thu, May 2, 2019 at 5:53 PM Nicolas Boichat <drinkcat@xxxxxxxxxxxx> wrote:
> >
> > On Thu, May 2, 2019 at 9:48 PM Yingjoe Chen <yingjoe.chen@xxxxxxxxxxxx> wrote:
> > >
> > > On Mon, 2019-04-29 at 11:25 +0800, Nicolas Boichat wrote:
> > > > pinctrl variants that include pinctrl-mtk-common-v2.h (and not
> > > > pinctrl-mtk-common.h) also need to use mtk_eint_pm_ops to setup
> > > > wake mask properly, so copy over the pm_ops to v2.
> > > >
> > > > It is not easy to merge the 2 copies (or move
> > > > mtk_eint_suspend/resume to mtk-eint.c), as we need to
> > > > dereference pctrl->eint, and struct mtk_pinctrl *pctl has a
> > > > different structure definition for v1 and v2.
> > > >
> > > > Signed-off-by: Nicolas Boichat <drinkcat@xxxxxxxxxxxx>
> > > > Reviewed-by: Chuanjia Liu <Chuanjia.Liu@xxxxxxxxxxxx>
> > > > ---
> > > >  .../pinctrl/mediatek/pinctrl-mtk-common-v2.c  | 19 +++++++++++++++++++
> > > >  .../pinctrl/mediatek/pinctrl-mtk-common-v2.h  |  1 +
> > > >  2 files changed, 20 insertions(+)
> > > >
> > > > diff --git a/drivers/pinctrl/mediatek/pinctrl-mtk-common-v2.c b/drivers/pinctrl/mediatek/pinctrl-mtk-common-v2.c
> > > > index 20e1c890e73b30c..7e19b5a4748eafe 100644
> > > > --- a/drivers/pinctrl/mediatek/pinctrl-mtk-common-v2.c
> > > > +++ b/drivers/pinctrl/mediatek/pinctrl-mtk-common-v2.c
> > > > @@ -723,3 +723,22 @@ int mtk_pinconf_adv_drive_get(struct mtk_pinctrl *hw,
> > > >
> > > >       return 0;
> > > >  }
> > > > +
> > > > +static int mtk_eint_suspend(struct device *device)
> > > > +{
> > > > +     struct mtk_pinctrl *pctl = dev_get_drvdata(device);
> > > > +
> > > > +     return mtk_eint_do_suspend(pctl->eint);
> > > > +}
> > > > +
> > > > +static int mtk_eint_resume(struct device *device)
> > > > +{
> > > > +     struct mtk_pinctrl *pctl = dev_get_drvdata(device);
> > > > +
> > > > +     return mtk_eint_do_resume(pctl->eint);
> > > > +}
> > > > +
> > > > +const struct dev_pm_ops mtk_eint_pm_ops = {
> > > > +     .suspend_noirq = mtk_eint_suspend,
> > > > +     .resume_noirq = mtk_eint_resume,
> > > > +};
> > >
> > > This is identical to the one in pinctrl-mtk-common.c and will have name
> > > clash if both pinctrl-mtk-common.c and pinctrl-mtk-common-v2.c are
> > > built.
> > >
> > > It would be better if we try to merge both version into mtk-eint.c, this
> > > way we could also remove some global functions.
> >
> > Argh, I didn't think about the name clash, you're right. I guess the
> > easy way is to rename this one mtk_eint_pm_ops_v2 ...
> >
> > As highlighted in the commit message, it's tricky to merge the 2 sets
> > of functions, they look identical, but they actually work on struct
> > mtk_pinctrl that are defined differently (in
> > pinctrl-mtk-common[-v2].h), so the ->eint member is at different
> > addresses...
> >
> > I don't really see a way around this... Unless we want to change
> > platform_set_drvdata(pdev, pctl); to pass another type of structure
> > that could be shared (but I think that'll make the code fairly
> > verbose, with another layer of indirection). Or just assign struct
> > mtk_eint to that, since that contains pctl so we could get back the
> > struct mtk_pinctrl from that, but that feels ugly as well...
> >
>
> I agree on renaming would make the thing simple. but I wouldn't like
> to rename to mtk_eint_pm_ops_v2 since this would make people
> misunderstand that is mtk_eint_v2.
>
> How about renaming to mtk_paris_pinctrl_pm_ops and then place related
> logic you added into pinctrl-paris.c? Because I prefer to keep pure
> pinctrl hardware operations in pinctrl-mtk-common-v2.c, and for
> relevant to other modules (mtk eint) or others subsystem (device tree
> binding, GPIO subsytem, PM something like that) they should be moved
> to pinctrl-paris.c or pinctrl-moore.c

Sounds reasonable. I uploaded a v2 that does just that.

Note that we'd still have to duplicate this code between paris and
moore, if we wanted to implement pm_ops in moore as well, but maybe
that's ok for now.

>      Sean
>
> > >
> > > Joe.C
> > >
> > >
> > >
> > > _______________________________________________
> > > Linux-mediatek mailing list
> > > Linux-mediatek@xxxxxxxxxxxxxxxxxxx
> > > http://lists.infradead.org/mailman/listinfo/linux-mediatek