2020-03-26 15:07:37

by Stefan Riedmüller

Subject: [PATCH 1/3] watchdog: da9062: Use pre-configured timeout until userspace takes over

If the watchdog is already running during probe read back its
pre-configured timeout (set e.g. by the bootloader) and use it to ping
the watchdog until userspace takes over. Otherwise the default timeout
set before might not result in a fast enough ping.

Signed-off-by: Stefan Riedmueller <[email protected]>
---
drivers/watchdog/da9062_wdt.c | 16 ++++++++++++++++
1 file changed, 16 insertions(+)

diff --git a/drivers/watchdog/da9062_wdt.c b/drivers/watchdog/da9062_wdt.c
index 0ad15d55071c..6d81b1276b87 100644
--- a/drivers/watchdog/da9062_wdt.c
+++ b/drivers/watchdog/da9062_wdt.c
@@ -35,6 +35,14 @@ struct da9062_watchdog {
bool use_sw_pm;
};

+static unsigned int da9062_wdt_read_timeout(struct da9062_watchdog *wdt)
+{
+ int val;
+
+ regmap_read(wdt->hw->regmap, DA9062AA_CONTROL_D, &val);
+ return wdt_timeout[val & DA9062AA_TWDSCALE_MASK];
+}
+
static unsigned int da9062_wdt_timeout_to_sel(unsigned int secs)
{
unsigned int i;
@@ -184,6 +192,7 @@ static int da9062_wdt_probe(struct platform_device *pdev)
{
struct device *dev = &pdev->dev;
int ret;
+ int timeout;
struct da9062 *chip;
struct da9062_watchdog *wdt;

@@ -213,6 +222,13 @@ static int da9062_wdt_probe(struct platform_device *pdev)
watchdog_set_drvdata(&wdt->wdtdev, wdt);
dev_set_drvdata(dev, &wdt->wdtdev);

+ timeout = da9062_wdt_read_timeout(wdt);
+ if (timeout > 0) {
+ wdt->wdtdev.timeout = timeout;
+ set_bit(WDOG_HW_RUNNING, &wdt->wdtdev.status);
+ dev_info(wdt->hw->dev, "watchdog is running (%u s)", timeout);
+ }
+
ret = devm_watchdog_register_device(dev, &wdt->wdtdev);
if (ret < 0)
return ret;
--
2.23.0

2020-03-31 16:03:33

by Guenter Roeck

Subject: Re: [PATCH 1/3] watchdog: da9062: Use pre-configured timeout until userspace takes over

On Thu, Mar 26, 2020 at 04:02:01PM +0100, Stefan Riedmueller wrote:
> If the watchdog is already running during probe read back its
> pre-configured timeout (set e.g. by the bootloader) and use it to ping
> the watchdog until userspace takes over. Otherwise the default timeout
> set before might not result in a fast enough ping.
>
> Signed-off-by: Stefan Riedmueller <[email protected]>
> ---
> drivers/watchdog/da9062_wdt.c | 16 ++++++++++++++++
> 1 file changed, 16 insertions(+)
>
> diff --git a/drivers/watchdog/da9062_wdt.c b/drivers/watchdog/da9062_wdt.c
> index 0ad15d55071c..6d81b1276b87 100644
> --- a/drivers/watchdog/da9062_wdt.c
> +++ b/drivers/watchdog/da9062_wdt.c
> @@ -35,6 +35,14 @@ struct da9062_watchdog {
> bool use_sw_pm;
> };
>
> +static unsigned int da9062_wdt_read_timeout(struct da9062_watchdog *wdt)
> +{
> + int val;
> +
> + regmap_read(wdt->hw->regmap, DA9062AA_CONTROL_D, &val);
> + return wdt_timeout[val & DA9062AA_TWDSCALE_MASK];
> +}
> +
> static unsigned int da9062_wdt_timeout_to_sel(unsigned int secs)
> {
> unsigned int i;
> @@ -184,6 +192,7 @@ static int da9062_wdt_probe(struct platform_device *pdev)
> {
> struct device *dev = &pdev->dev;
> int ret;
> + int timeout;
> struct da9062 *chip;
> struct da9062_watchdog *wdt;
>
> @@ -213,6 +222,13 @@ static int da9062_wdt_probe(struct platform_device *pdev)
> watchdog_set_drvdata(&wdt->wdtdev, wdt);
> dev_set_drvdata(dev, &wdt->wdtdev);
>
> + timeout = da9062_wdt_read_timeout(wdt);
> + if (timeout > 0) {
> + wdt->wdtdev.timeout = timeout;
> + set_bit(WDOG_HW_RUNNING, &wdt->wdtdev.status);
> + dev_info(wdt->hw->dev, "watchdog is running (%u s)", timeout);

The user won't know what "(%u s)" means.e that "%u" reflects the timeout.
Also, the newline at the end is missing.

Personally, I think the message is just noise and should be dropped entirely.

Guenter

> + }
> +
> ret = devm_watchdog_register_device(dev, &wdt->wdtdev);
> if (ret < 0)
> return ret;

2020-03-31 16:08:04

by Guenter Roeck

Subject: Re: [PATCH 1/3] watchdog: da9062: Use pre-configured timeout until userspace takes over

On Thu, Mar 26, 2020 at 04:02:01PM +0100, Stefan Riedmueller wrote:
> If the watchdog is already running during probe read back its
> pre-configured timeout (set e.g. by the bootloader) and use it to ping
> the watchdog until userspace takes over. Otherwise the default timeout
> set before might not result in a fast enough ping.
>
> Signed-off-by: Stefan Riedmueller <[email protected]>
> ---
> drivers/watchdog/da9062_wdt.c | 16 ++++++++++++++++
> 1 file changed, 16 insertions(+)
>
> diff --git a/drivers/watchdog/da9062_wdt.c b/drivers/watchdog/da9062_wdt.c
> index 0ad15d55071c..6d81b1276b87 100644
> --- a/drivers/watchdog/da9062_wdt.c
> +++ b/drivers/watchdog/da9062_wdt.c
> @@ -35,6 +35,14 @@ struct da9062_watchdog {
> bool use_sw_pm;
> };
>
> +static unsigned int da9062_wdt_read_timeout(struct da9062_watchdog *wdt)
> +{
> + int val;
> +
> + regmap_read(wdt->hw->regmap, DA9062AA_CONTROL_D, &val);
> + return wdt_timeout[val & DA9062AA_TWDSCALE_MASK];
> +}
> +
> static unsigned int da9062_wdt_timeout_to_sel(unsigned int secs)
> {
> unsigned int i;
> @@ -184,6 +192,7 @@ static int da9062_wdt_probe(struct platform_device *pdev)
> {
> struct device *dev = &pdev->dev;
> int ret;
> + int timeout;
> struct da9062 *chip;
> struct da9062_watchdog *wdt;
>
> @@ -213,6 +222,13 @@ static int da9062_wdt_probe(struct platform_device *pdev)
> watchdog_set_drvdata(&wdt->wdtdev, wdt);
> dev_set_drvdata(dev, &wdt->wdtdev);
>
> + timeout = da9062_wdt_read_timeout(wdt);
> + if (timeout > 0) {
> + wdt->wdtdev.timeout = timeout;
> + set_bit(WDOG_HW_RUNNING, &wdt->wdtdev.status);
> + dev_info(wdt->hw->dev, "watchdog is running (%u s)", timeout);
> + }

Oh, and I agree with the comment made for the da9063 driver: This driver
should really implement watchdog_init_timeout() and set the default timeout
accordingly if specified in dt (or, in other words, follow the da9063
implementation).

Guenter

> +
> ret = devm_watchdog_register_device(dev, &wdt->wdtdev);
> if (ret < 0)
> return ret;