Re: [PATCH v6 6/6] PCI: uniphier: Add error message when failed to get phy

[Rob Herring <robh+dt@kernel.org>]

On Fri, Aug 21, 2020 at 1:05 AM Kunihiko Hayashi
<hayashi.kunihiko@socionext.com> wrote:
>
> On 2020/08/18 1:39, Rob Herring wrote:
> > On Fri, Aug 7, 2020 at 4:25 AM Kunihiko Hayashi
> > <hayashi.kunihiko@socionext.com> wrote:
> >>
> >> Even if phy driver doesn't probe, the error message can't be distinguished
> >> from other errors. This displays error message caused by the phy driver
> >> explicitly.
> >>
> >> Signed-off-by: Kunihiko Hayashi <hayashi.kunihiko@socionext.com>
> >> ---
> >>   drivers/pci/controller/dwc/pcie-uniphier.c | 8 ++++++--
> >>   1 file changed, 6 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/drivers/pci/controller/dwc/pcie-uniphier.c b/drivers/pci/controller/dwc/pcie-uniphier.c
> >> index 93ef608..7c8721e 100644
> >> --- a/drivers/pci/controller/dwc/pcie-uniphier.c
> >> +++ b/drivers/pci/controller/dwc/pcie-uniphier.c
> >> @@ -489,8 +489,12 @@ static int uniphier_pcie_probe(struct platform_device *pdev)
> >>                  return PTR_ERR(priv->rst);
> >>
> >>          priv->phy = devm_phy_optional_get(dev, "pcie-phy");
> >
> > The point of the optional variant vs. devm_phy_get() is whether or not
> > you get an error message. So shouldn't you switch to devm_phy_get
> > instead?
> >
> >> -       if (IS_ERR(priv->phy))
> >> -               return PTR_ERR(priv->phy);
> >> +       if (IS_ERR(priv->phy)) {
> >> +               ret = PTR_ERR(priv->phy);
> >> +               if (ret != -EPROBE_DEFER)
> >> +                       dev_err(dev, "Failed to get phy (%d)\n", ret);
> >> +               return ret;
> >> +       }
>
> The 'phys' property is optional, so if there isn't 'phys' in the PCIe node,
> devm_phy_get() returns -ENODEV, and devm_phy_optional_get() returns NULL.
>
> When devm_phy_optional_get() replaces devm_phy_get(),
> condition for displaying an error message changes to:
>
>     (ret != -EPROBE_DEFER && ret != -ENODEV)
>
> This won't be simple, but should it be replaced?

Nevermind. I was thinking we had some error prints for the optional
vs. non-optional variants.

Rob