Faults on Odroid-C2 during shutdown and fix proposal

Faults on Odroid-C2 during shutdown and fix proposal

[Heiner Kallweit]

Neil,

during shutdown I experience faults as described in the commit message of
339cd0ea0822 "soc: amlogic: meson-gx-pwrc-vpu: fix power-off when powered by bootloader"

My best guess:
I operate the board headless, therefore the domain is never powered on, leaving the
clocks disabled. The shutdown function then tries to disable the already disabled
clocks.
The following change fixes it for me, however there may be scenarios I miss,
therefore I'd like to hear your opinion.

 static void meson_gx_pwrc_vpu_shutdown(struct platform_device *pdev)
 {
-       meson_gx_pwrc_vpu_power_off(&vpu_hdmi_pd.genpd);
+       bool powered_off;
+
+       powered_off = meson_gx_pwrc_vpu_get_power(&vpu_hdmi_pd);
+       if (!powered_off)
+               meson_gx_pwrc_vpu_power_off(&vpu_hdmi_pd.genpd);
 }

And something completely different:
I get a misleading error message "failed to get reset lines", caused by the fact
that devm_reset_control_array_get() returns -EPROBE_DEFER. So I think we should
print the error message only if the error is not caused by a probe deferral.

Kind regards,
Heiner

Faults on Odroid-C2 during shutdown and fix proposal

[Neil Armstrong]

On 16/12/2017 00:19, Heiner Kallweit wrote:
> Neil,
> 
> during shutdown I experience faults as described in the commit message of
> 339cd0ea0822 "soc: amlogic: meson-gx-pwrc-vpu: fix power-off when powered by bootloader"
> 
> My best guess:
> I operate the board headless, therefore the domain is never powered on, leaving the
> clocks disabled. The shutdown function then tries to disable the already disabled
> clocks.
> The following change fixes it for me, however there may be scenarios I miss,
> therefore I'd like to hear your opinion.
> 
>  static void meson_gx_pwrc_vpu_shutdown(struct platform_device *pdev)
>  {
> -       meson_gx_pwrc_vpu_power_off(&vpu_hdmi_pd.genpd);
> +       bool powered_off;
> +
> +       powered_off = meson_gx_pwrc_vpu_get_power(&vpu_hdmi_pd);
> +       if (!powered_off)
> +               meson_gx_pwrc_vpu_power_off(&vpu_hdmi_pd.genpd);
>  }
> 
> And something completely different:
> I get a misleading error message "failed to get reset lines", caused by the fact
> that devm_reset_control_array_get() returns -EPROBE_DEFER. So I think we should
> print the error message only if the error is not caused by a probe deferral.

Feel free to provide a patch.

> 
> Kind regards,
> Heiner
> 

Hi Heiner,

Indeed, it's another corner case, please provide a patch.

Thanks,
Neil