Re: [PATCH] media: mediatek: vcodec: Convert mtk_vcodec_dec_hw platform remove callback

[Fei Shao]

Hi,

On Thu, May 11, 2023 at 1:50 PM Chen-Yu Tsai <wenst@xxxxxxxxxxxx> wrote:
>
> On Thu, May 11, 2023 at 12:43 AM Uwe Kleine-König
> <u.kleine-koenig@xxxxxxxxxxxxxx> wrote:
> >
> > [expanding the audience a bit for more expertise]
> >
> > Hello,
> >
> > On Wed, May 10, 2023 at 11:31:35PM +0800, Fei Shao wrote:
> > > This aligns with [1] and converts the platform remove callback to
> > > .remove_new(), which returns void.
> > >
> > > [1]: commit a3afc5b10661 ("media: mtk_vcodec_dec_drv: Convert to
> > >      platform remove callback returning void")
> > >
> > > Signed-off-by: Fei Shao <fshao@xxxxxxxxxxxx>
> > >
> > > ---
> > >
> > >  drivers/media/platform/mediatek/vcodec/mtk_vcodec_dec_hw.c | 6 ++----
> > >  1 file changed, 2 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/drivers/media/platform/mediatek/vcodec/mtk_vcodec_dec_hw.c b/drivers/media/platform/mediatek/vcodec/mtk_vcodec_dec_hw.c
> > > index b753bf54ebd9..bd5743723da6 100644
> > > --- a/drivers/media/platform/mediatek/vcodec/mtk_vcodec_dec_hw.c
> > > +++ b/drivers/media/platform/mediatek/vcodec/mtk_vcodec_dec_hw.c
> > > @@ -193,16 +193,14 @@ static int mtk_vdec_hw_probe(struct platform_device *pdev)
> > >       return ret;
> > >  }
> > >
> > > -static int mtk_vdec_hw_remove(struct platform_device *pdev)
> > > +static void mtk_vdec_hw_remove(struct platform_device *pdev)
> > >  {
> > >       pm_runtime_disable(&pdev->dev);
> > > -
> > > -     return 0;
> > >  }
> > >
> > >  static struct platform_driver mtk_vdec_driver = {
> > >       .probe  = mtk_vdec_hw_probe,
> > > -     .remove = mtk_vdec_hw_remove,
> > > +     .remove_new = mtk_vdec_hw_remove,
> > >       .driver = {
> > >               .name   = "mtk-vdec-comp",
> > >               .of_match_table = mtk_vdec_hw_match,
> >
> > While the patch looks fine, I wonder if having a remove callback just to
> > do pm_runtime_disable() is worth keeping it. Doesn't the core care for
> > things like that? I grepped a bit around, device_unbind_cleanup() is
> > called after device_remove() which calls pm_runtime_reinit(). Does that
> > mean calling pm_runtime_disable in .remove() is useless? In that case,
> > you could just drop the .remove() callback.

Thanks for the feedback.

I wonder if the core would handle that for drivers... if I understand
it correctly, pm_runtime_reinit() does not disable runtime PM for a
device, otherwise pm_runtime_remove() wouldn't need to bother
calling __pm_runtime_disable() before pm_runtime_reinit().

In fact, from the runtime_pm documentation [1] I read the following:
"Drivers in ->remove() callback should undo the runtime PM changes
done in ->probe(). Usually this means calling pm_runtime_disable(),
pm_runtime_dont_use_autosuspend() etc."

Based on the above I assume it's still necessary given pm_runtime_enable()
was called in ->probe().

[1]: https://www.kernel.org/doc/Documentation/power/runtime_pm.txt

>
> Maybe just switch to devm_pm_runtime_enable() on the enable side?

That sounds like a good alternative, I'll revise and send a new patch, thanks.

Regards,
Fei

>
> ChenYu