Re: [PATCH 2/2] ASoC: tegra: probe deferral error reporting

From: Dmitry Osipenko
Date: Sat Jul 21 2018 - 08:16:07 EST


On Saturday, 21 July 2018 15:03:57 MSK Dmitry Osipenko wrote:
> On Saturday, 21 July 2018 14:55:21 MSK Marcel Ziswiler wrote:
> > On Sat, 2018-07-21 at 14:17 +0300, Dmitry Osipenko wrote:
> > > On Saturday, 21 July 2018 12:56:15 MSK Mark Brown wrote:
> > > > On Fri, Jul 20, 2018 at 12:31:07PM +0000, Marcel Ziswiler wrote:
> > > > > On Fri, 2018-07-20 at 13:16 +0100, Mark Brown wrote:
> > > > > > > ac97->sync_gpio = of_get_named_gpio(pdev-
> > > > > > > >
> > > > > > > >dev.of_node,
> > > > > > > >
> > > > > > > "nvidia,codec-
> > > > > > >
> > > > > > > sync-
> > > > > > >
> > > > > > > gpio", 0);
> > > > > > >
> > > > > > > if (!gpio_is_valid(ac97->sync_gpio)) {
> > > > > > >
> > > > > > > - dev_err(&pdev->dev, "no codec-sync GPIO
> > > > > > > supplied\n");
> > > > > > > + ret = ac97->sync_gpio;
> > > > > > > + dev_err(&pdev->dev, "no codec-sync GPIO
> > > > > > > supplied:
> > > > > > > %d\n", ret);
> > > > > > >
> > > > > > > goto err_clk_put;
> > > > > > >
> > > > > > > }
> > > > > >
> > > > > > This isn't reporting an error code associated with the attempt
> > > > > > to
> > > > > > find a
> > > > > > codec-sync GPIO, it's the result of some other operation.
> > > > >
> > > > > What exactly is then the of_get_named_gpio() above please doing
> > > > > if
> > > > > not getting the codec sync GPIO? I am not following you, sorry.
> > > >
> > > > It's not in any way involved in setting the value of ret, whatever
> > > > value
> > > > that has it's nothing to do with that operation.
> > >
> > > The comment to gpio_is_valid() says that it "Returns GPIO number to
> > > use with
> > > Linux generic GPIO API, or one of the errno value on the error
> > > condition".
> > > Comment doesn't explicitly states that the returned GPIO number is
> > > always
> > > valid, but it is kinda implied.
> >
> > Do you mean I should be assigning the return value of gpio_is_valid()
> > to ret and use that instead?
>
> No, gpio_is_valid() returns a boolean. I think your patch is fine as it is
> is.
>
> Probably Mark meant something like this:
>
> ac97->sync_gpio = of_get_named_gpio(pdev->dev.of_node,
> "nvidia,codec-sync-gpio", 0);
> if (ac97->sync_gpio < 0) {
> ret = ac97->sync_gpio;
> dev_err(&pdev->dev, "no codec-sync GPIO supplied: %d\n",
> ret); goto err_clk_put;
> }
>
> if (!gpio_is_valid(ac97->sync_gpio)) {
> ret = -EINVAL;
> goto err_clk_put;
> }
>
> But that is not needed because of_get_named_gpio() returns either a valid
> GPIO number or a error code.

Also note that tegra20_ac97 code doesn't check the returned value of:

gpio_request(workdata->sync_gpio, "codec-sync");

That is a bit fragile. Probably would be better to move the GPIO requesting to
the drivers probe function using the devm_gpio_request_one() and fail the
driver probing if requesting fails. That should be a distinct patch if you'll
want to implement that.