Linux Watchdog driver development
 help / color / mirror / Atom feed
* [PATCH] watchdog: renesas_wdt: avoid sleeping clock calls in restart handler
@ 2026-11-17  2:05 Luo Yifan
  2026-11-17  2:20 ` sashiko-bot
  0 siblings, 1 reply; 3+ messages in thread
From: Luo Yifan @ 2026-11-17  2:05 UTC (permalink / raw)
  To: Wim Van Sebroeck, Guenter Roeck, linux-watchdog; +Cc: linux-kernel

rwdt_restart() runs in atomic context, as the comment above the function
points out, but it calls clk_prepare_enable().  clk_prepare() may sleep on
the clock mutex, which is not allowed from a restart handler.

Prepare the clock once in the probe function and only call clk_enable()
from the restart handler.  clk_enable() does not sleep, so the restart
path stays atomic.  Unprepare the clock in the probe error path and in
rwdt_remove().

Signed-off-by: Luo Yifan <luoyifan@cmss.chinamobile.com>
---
 drivers/watchdog/renesas_wdt.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)

diff --git a/drivers/watchdog/renesas_wdt.c b/drivers/watchdog/renesas_wdt.c
index d9760f2a3b..9630d05686 100644
--- a/drivers/watchdog/renesas_wdt.c
+++ b/drivers/watchdog/renesas_wdt.c
@@ -135,7 +135,7 @@ static int rwdt_restart(struct watchdog_device *wdev, unsigned long action,
 	struct rwdt_priv *priv = watchdog_get_drvdata(wdev);
 	u8 val;
 
-	clk_prepare_enable(priv->clk);
+	clk_enable(priv->clk);
 
 	/* Stop the timer before we modify any register */
 	val = readb_relaxed(priv->base + RWTCSRA) & ~RWTCSRA_TME;
@@ -233,6 +233,14 @@ static int rwdt_probe(struct platform_device *pdev)
 	if (IS_ERR(priv->clk))
 		return PTR_ERR(priv->clk);
 
+	/*
+	 * The restart handler runs in atomic context and can only enable the
+	 * clock, so prepare it here.
+	 */
+	ret = clk_prepare(priv->clk);
+	if (ret)
+		return ret;
+
 	pm_runtime_enable(dev);
 	pm_runtime_get_sync(dev);
 	priv->clk_rate = clk_get_rate(priv->clk);
@@ -290,6 +298,7 @@ static int rwdt_probe(struct platform_device *pdev)
 
  out_pm_disable:
 	pm_runtime_disable(dev);
+	clk_unprepare(priv->clk);
 	return ret;
 }
 
@@ -299,6 +308,7 @@ static void rwdt_remove(struct platform_device *pdev)
 
 	watchdog_unregister_device(&priv->wdev);
 	pm_runtime_disable(&pdev->dev);
+	clk_unprepare(priv->clk);
 }
 
 static int rwdt_suspend(struct device *dev)
-- 
2.24.4




^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [PATCH] watchdog: renesas_wdt: avoid sleeping clock calls in restart handler
  2026-11-17  2:05 [PATCH] watchdog: renesas_wdt: avoid sleeping clock calls in restart handler Luo Yifan
@ 2026-11-17  2:20 ` sashiko-bot
  2026-11-18  1:58   ` Guenter Roeck
  0 siblings, 1 reply; 3+ messages in thread
From: sashiko-bot @ 2026-11-17  2:20 UTC (permalink / raw)
  To: Luo Yifan, linux-watchdog

Thank you for your contribution! Sashiko AI review found 2 potential
issues to consider:
- [Low] watchdog: renesas_wdt: missing Fixes trailer for atomic sleep
  fix
- [Low] watchdog: renesas_wdt: use devm_clk_get_prepared() in
  rwdt_probe()
--

commit 4ed48dd9a6e0be55587f40b05e00cb1b9e5da877
Author: Luo Yifan <luoyifan@cmss.chinamobile.com>

> watchdog: renesas_wdt: avoid sleeping clock calls in restart handler
>
> rwdt_restart() runs in atomic context, as the comment above the function
> points out, but it calls clk_prepare_enable().  clk_prepare() may sleep on
> the clock mutex, which is not allowed from a restart handler.
>
> Prepare the clock once in the probe function and only call clk_enable()
> from the restart handler.  clk_enable() does not sleep, so the restart
> path stays atomic.  Unprepare the clock in the probe error path and in
> rwdt_remove().
>
> Signed-off-by: Luo Yifan <luoyifan@cmss.chinamobile.com>

[Severity: Low]
Since this fixes the sleeping clk_prepare_enable() call in rwdt_restart(),
should this include a Fixes tag?

  Fixes: fa01fa70039c ("watchdog: renesas_wdt: don't sleep in atomic context")

> diff --git a/drivers/watchdog/renesas_wdt.c b/drivers/watchdog/renesas_wdt.c
> index d9760f2a3b235..9630d05686461 100644
> --- a/drivers/watchdog/renesas_wdt.c
> +++ b/drivers/watchdog/renesas_wdt.c

[ ... ]

> @@ -233,6 +233,14 @@ static int rwdt_probe(struct platform_device *pdev)
>  	if (IS_ERR(priv->clk))
>  		return PTR_ERR(priv->clk);
>  
> +	/*
> +	 * The restart handler runs in atomic context and can only enable the
> +	 * clock, so prepare it here.
> +	 */
> +	ret = clk_prepare(priv->clk);
> +	if (ret)
> +		return ret;
> +
>  	pm_runtime_enable(dev);
>  	pm_runtime_get_sync(dev);
>  	priv->clk_rate = clk_get_rate(priv->clk);
> @@ -290,6 +298,7 @@ static int rwdt_probe(struct platform_device *pdev)
>  
>   out_pm_disable:
>  	pm_runtime_disable(dev);
> +	clk_unprepare(priv->clk);
>  	return ret;
>  }
>  
> @@ -299,6 +308,7 @@ static void rwdt_remove(struct platform_device *pdev)
>  
>  	watchdog_unregister_device(&priv->wdev);
>  	pm_runtime_disable(&pdev->dev);
> +	clk_unprepare(priv->clk);
>  }

[Severity: Low]
This isn't a bug, but could rwdt_probe() use devm_clk_get_prepared() instead
of devm_clk_get()?  That would prepare priv->clk on acquisition and
automatically unprepare it on probe failure or after rwdt_remove(), avoiding
the manual clk_prepare() and clk_unprepare() calls.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261117020530.19973-1-luoyifan@cmss.chinamobile.com?part=1

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH] watchdog: renesas_wdt: avoid sleeping clock calls in restart handler
  2026-11-17  2:20 ` sashiko-bot
@ 2026-11-18  1:58   ` Guenter Roeck
  0 siblings, 0 replies; 3+ messages in thread
From: Guenter Roeck @ 2026-11-18  1:58 UTC (permalink / raw)
  To: sashiko-reviews, Luo Yifan, linux-watchdog

On 11/16/26 18:20, sashiko-bot@kernel.org wrote:
> Thank you for your contribution! Sashiko AI review found 2 potential
> issues to consider:
> - [Low] watchdog: renesas_wdt: missing Fixes trailer for atomic sleep
>    fix
> - [Low] watchdog: renesas_wdt: use devm_clk_get_prepared() in
>    rwdt_probe()

Please address both.

Thanks,
Guenter

> --
> 
> commit 4ed48dd9a6e0be55587f40b05e00cb1b9e5da877
> Author: Luo Yifan <luoyifan@cmss.chinamobile.com>
> 
>> watchdog: renesas_wdt: avoid sleeping clock calls in restart handler
>>
>> rwdt_restart() runs in atomic context, as the comment above the function
>> points out, but it calls clk_prepare_enable().  clk_prepare() may sleep on
>> the clock mutex, which is not allowed from a restart handler.
>>
>> Prepare the clock once in the probe function and only call clk_enable()
>> from the restart handler.  clk_enable() does not sleep, so the restart
>> path stays atomic.  Unprepare the clock in the probe error path and in
>> rwdt_remove().
>>
>> Signed-off-by: Luo Yifan <luoyifan@cmss.chinamobile.com>
> 
> [Severity: Low]
> Since this fixes the sleeping clk_prepare_enable() call in rwdt_restart(),
> should this include a Fixes tag?
> 
>    Fixes: fa01fa70039c ("watchdog: renesas_wdt: don't sleep in atomic context")
> 
>> diff --git a/drivers/watchdog/renesas_wdt.c b/drivers/watchdog/renesas_wdt.c
>> index d9760f2a3b235..9630d05686461 100644
>> --- a/drivers/watchdog/renesas_wdt.c
>> +++ b/drivers/watchdog/renesas_wdt.c
> 
> [ ... ]
> 
>> @@ -233,6 +233,14 @@ static int rwdt_probe(struct platform_device *pdev)
>>   	if (IS_ERR(priv->clk))
>>   		return PTR_ERR(priv->clk);
>>   
>> +	/*
>> +	 * The restart handler runs in atomic context and can only enable the
>> +	 * clock, so prepare it here.
>> +	 */
>> +	ret = clk_prepare(priv->clk);
>> +	if (ret)
>> +		return ret;
>> +
>>   	pm_runtime_enable(dev);
>>   	pm_runtime_get_sync(dev);
>>   	priv->clk_rate = clk_get_rate(priv->clk);
>> @@ -290,6 +298,7 @@ static int rwdt_probe(struct platform_device *pdev)
>>   
>>    out_pm_disable:
>>   	pm_runtime_disable(dev);
>> +	clk_unprepare(priv->clk);
>>   	return ret;
>>   }
>>   
>> @@ -299,6 +308,7 @@ static void rwdt_remove(struct platform_device *pdev)
>>   
>>   	watchdog_unregister_device(&priv->wdev);
>>   	pm_runtime_disable(&pdev->dev);
>> +	clk_unprepare(priv->clk);
>>   }
> 
> [Severity: Low]
> This isn't a bug, but could rwdt_probe() use devm_clk_get_prepared() instead
> of devm_clk_get()?  That would prepare priv->clk on acquisition and
> automatically unprepare it on probe failure or after rwdt_remove(), avoiding
> the manual clk_prepare() and clk_unprepare() calls.
> 


^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2026-11-18  1:59 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-17  2:05 [PATCH] watchdog: renesas_wdt: avoid sleeping clock calls in restart handler Luo Yifan
2026-11-17  2:20 ` sashiko-bot
2026-11-18  1:58   ` Guenter Roeck

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox