Re: [PATCH V4 2/2] net: ethernet: ravb: Enable optional refclk

[Geert Uytterhoeven]

Hi Adam,

On Wed, Apr 14, 2021 at 3:08 PM Adam Ford <aford173@xxxxxxxxx> wrote:
> On Tue, Apr 13, 2021 at 2:33 AM Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
> > On Mon, Apr 12, 2021 at 3:27 PM Adam Ford <aford173@xxxxxxxxx> wrote:
> > > For devices that use a programmable clock for the AVB reference clock,
> > > the driver may need to enable them.  Add code to find the optional clock
> > > and enable it when available.
> > >
> > > Signed-off-by: Adam Ford <aford173@xxxxxxxxx>
> > > Reviewed-by: Andrew Lunn <andrew@xxxxxxx>
> > >
> > > ---
> > > V4:  Eliminate the NULL check when disabling refclk, and add a line
> > >      to disable the refclk if there is a failure after it's been
> > >      initialized.
> >
> > Thanks for the update!
> >
> > > --- a/drivers/net/ethernet/renesas/ravb_main.c
> > > +++ b/drivers/net/ethernet/renesas/ravb_main.c
> > > @@ -2148,6 +2148,13 @@ static int ravb_probe(struct platform_device *pdev)
> > >                 goto out_release;
> > >         }
> > >
> > > +       priv->refclk = devm_clk_get_optional(&pdev->dev, "refclk");
> > > +       if (IS_ERR(priv->refclk)) {
> > > +               error = PTR_ERR(priv->refclk);
> > > +               goto out_release;
> >
> > Note that this will call clk_disable_unprepare() in case of failure, which is
> > fine, as that function is a no-op in case of a failed clock.
>
> Geert,
>
> A bot reported that if I jump to out_release may try to free a clock
> if some instances where priv isn't defined.

As priv is allocated using alloc_etherdev_mqs(), priv->refclk is
NULL initially, but priv itself may indeed not be initialized when the first
"goto out_release" is taken.  Sorry for missing that.

> Currently, the priv->clk isn't freed either.  I have heard some
> back-and-forth discussions in other threads on whether or not devm
> functions auto free or not.

The devm_clk_get_optional() will be undone automatically, so there
is no need to handle that explicitly.

> I'm fine with sending a V5 to make the free for the refclock happen
> only when the priv has successfully initialized.  Should I also add

As this patch has been applied to net-next, you''ll have to send
a follow-up fix patch, not a v5.

> one for freeing priv->clk and change all the other goto out_release
> commands to point to this new section?

No, not for priv->clk, due to devm_*().

> I am thinking it would like something like:
>
> free_refclk:
>     clk_disable_unprepare(priv->refclk);

OK.

> free_clk;
>     clk_disable_unprepare(priv->clk);

NAK, as priv->clk is not enabled in ravb_probe().

> out_release:
>     free_netdev(ndev);
>     ....

OK.

Thanks!

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds