Quoting Sai Prakash Ranjan (2019-01-17 07:19:42)
> diff --git a/drivers/watchdog/qcom-wdt.c b/drivers/watchdog/qcom-wdt.c
> index 780971318810..5dfd604477a4 100644
> --- a/drivers/watchdog/qcom-wdt.c
> +++ b/drivers/watchdog/qcom-wdt.c
> @@ -245,6 +245,28 @@ static int qcom_wdt_remove(struct platform_device *pdev)
> return 0;
> }
>
> +static int __maybe_unused qcom_wdt_suspend(struct device *dev)
> +{
> + struct qcom_wdt *wdt = dev_get_drvdata(dev);
> +
> + if (watchdog_active(&wdt->wdd))
> + qcom_wdt_stop(&wdt->wdd);
> +
> + return 0;
> +}
> +
> +static int __maybe_unused qcom_wdt_resume(struct device *dev)
> +{
> + struct qcom_wdt *wdt = dev_get_drvdata(dev);
> +
> + if (watchdog_active(&wdt->wdd))
> + qcom_wdt_start(&wdt->wdd);
> +
> + return 0;
> +}
This looks fairly generic. For example, the Mediatek driver also stops
and starts (but also pings after starting). Grepping for 'active' in
drivers/watchdog/ finds more examples. Could there be some functions in
watchdog core that do the common things like watchdog_stop() and
watchdog_start() and watchdog_start_and_ping()? Or maybe a bit can be
set during registration so that the 'struct class watchdog_class' can
get PM ops to stop and start on suspend/resume of the watchdog character
device?
Nothing is wrong with the patch, I'm just bemoaning the amount of code
duplication here.
