Hi Prabhkar,
> Subject: Re: [PATCH 1/2] mmc: renesas_sdhi: Jump to error path instead of
> returning directly
>
> Hi Biju,
>
> Thank you for the review.
>
> On Mon, Apr 4, 2022 at 7:02 PM Biju Das <[email protected]>
> wrote:
> >
> > Hi Prabhakar and Pavel,
> >
> > Thanks for the patch.
> >
> > > Subject: [PATCH 1/2] mmc: renesas_sdhi: Jump to error path instead
> > > of returning directly
> > >
> > > Jump to error path "edisclk" instead of returning directly in case
> > > of
> > > devm_reset_control_get_optional_exclusive() failure.
> > >
> > > Fixes: b4d86f37eacb7 ("mmc: renesas_sdhi: do hard reset if
> > > possible")
> > > Reported-by: Pavel Machek <[email protected]>
> > > Signed-off-by: Lad Prabhakar
> > > <[email protected]>
> > > ---
> > > drivers/mmc/host/renesas_sdhi_core.c | 6 ++++--
> > > 1 file changed, 4 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/drivers/mmc/host/renesas_sdhi_core.c
> > > b/drivers/mmc/host/renesas_sdhi_core.c
> > > index 2797a9c0f17d..cddb0185f5fb 100644
> > > --- a/drivers/mmc/host/renesas_sdhi_core.c
> > > +++ b/drivers/mmc/host/renesas_sdhi_core.c
> > > @@ -1033,8 +1033,10 @@ int renesas_sdhi_probe(struct platform_device
> > > *pdev,
> > > goto efree;
> > >
> > > priv->rstc =
> > > devm_reset_control_get_optional_exclusive(&pdev->dev,
> > > NULL);
> > > - if (IS_ERR(priv->rstc))
> > > - return PTR_ERR(priv->rstc);
> > > + if (IS_ERR(priv->rstc)) {
> > > + ret = PTR_ERR(priv->rstc);
> > > + goto edisclk;
> > > + }
> >
> > Why can't devm_reset_control_get_optional_exclusive to be moved up
> before devm_clk_get?
> >
> In that case we will have to jump to the "efree" label Or if you don't
> want goto at all this can be moved to the very beginning of the probe.
I guess it has to move up, first get reset handle and clock handle and return error
directly in case of error, Then do clk/reset ops.
>
> Wolfram, what is your preference on the above?
>
> Cheers,
> Prabhakar
Hi Biju,
On Mon, Apr 4, 2022 at 7:12 PM Biju Das <[email protected]> wrote:
>
> Hi Prabhkar,
>
> > Subject: Re: [PATCH 1/2] mmc: renesas_sdhi: Jump to error path instead of
> > returning directly
> >
> > Hi Biju,
> >
> > Thank you for the review.
> >
> > On Mon, Apr 4, 2022 at 7:02 PM Biju Das <[email protected]>
> > wrote:
> > >
> > > Hi Prabhakar and Pavel,
> > >
> > > Thanks for the patch.
> > >
> > > > Subject: [PATCH 1/2] mmc: renesas_sdhi: Jump to error path instead
> > > > of returning directly
> > > >
> > > > Jump to error path "edisclk" instead of returning directly in case
> > > > of
> > > > devm_reset_control_get_optional_exclusive() failure.
> > > >
> > > > Fixes: b4d86f37eacb7 ("mmc: renesas_sdhi: do hard reset if
> > > > possible")
> > > > Reported-by: Pavel Machek <[email protected]>
> > > > Signed-off-by: Lad Prabhakar
> > > > <[email protected]>
> > > > ---
> > > > drivers/mmc/host/renesas_sdhi_core.c | 6 ++++--
> > > > 1 file changed, 4 insertions(+), 2 deletions(-)
> > > >
> > > > diff --git a/drivers/mmc/host/renesas_sdhi_core.c
> > > > b/drivers/mmc/host/renesas_sdhi_core.c
> > > > index 2797a9c0f17d..cddb0185f5fb 100644
> > > > --- a/drivers/mmc/host/renesas_sdhi_core.c
> > > > +++ b/drivers/mmc/host/renesas_sdhi_core.c
> > > > @@ -1033,8 +1033,10 @@ int renesas_sdhi_probe(struct platform_device
> > > > *pdev,
> > > > goto efree;
> > > >
> > > > priv->rstc =
> > > > devm_reset_control_get_optional_exclusive(&pdev->dev,
> > > > NULL);
> > > > - if (IS_ERR(priv->rstc))
> > > > - return PTR_ERR(priv->rstc);
> > > > + if (IS_ERR(priv->rstc)) {
> > > > + ret = PTR_ERR(priv->rstc);
> > > > + goto edisclk;
> > > > + }
> > >
> > > Why can't devm_reset_control_get_optional_exclusive to be moved up
> > before devm_clk_get?
> > >
> > In that case we will have to jump to the "efree" label Or if you don't
> > want goto at all this can be moved to the very beginning of the probe.
>
> I guess it has to move up, first get reset handle and clock handle and return error
> directly in case of error, Then do clk/reset ops.
>
Fine by me.
Cheers,
Prabhakar
> >
> > Wolfram, what is your preference on the above?
> >
> > Cheers,
> > Prabhakar
> I guess it has to move up, first get reset handle and clock handle and return error
> directly in case of error, Then do clk/reset ops.
>
> >
> > Wolfram, what is your preference on the above?
Yes, moving up makes sense. First check all the handles before we
actually initialize the hardware.
Thanks for pointing all this out.