Re: [PATCH] platform/x86/amd/pmc: Fix RTC device leak in amd_pmc_verify_czn_rtc()

[Ilpo Järvinen]

On Thu, 17 Sep 2026, Wentao Liang wrote:

> rtc_class_open() takes a reference to the RTC device, but all paths
> that return after it succeeded, apart from the final one where the
> alarm is programmed, leave the function without dropping it. Route
> them through a common exit that calls rtc_class_close().
> 
> Fixes: 59348401ebed ("platform/x86: amd-pmc: Add special handling for timer based S0i3 wakeup")
> Cc: stable@xxxxxxxxxxxxxxx
> Signed-off-by: Wentao Liang <vulab@xxxxxxxxxxx>
> ---
>  drivers/platform/x86/amd/pmc/pmc.c | 21 ++++++++++++++-------
>  1 file changed, 14 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/platform/x86/amd/pmc/pmc.c b/drivers/platform/x86/amd/pmc/pmc.c
> index cae3fcafd4d7..f40e8d9b124a 100644
> --- a/drivers/platform/x86/amd/pmc/pmc.c
> +++ b/drivers/platform/x86/amd/pmc/pmc.c
> @@ -569,32 +569,39 @@ static int amd_pmc_verify_czn_rtc(struct amd_pmc_dev *pdev, u32 *arg)
>  		return 0;
>  	rc = rtc_read_alarm(rtc_device, &alarm);
>  	if (rc)
> -		return rc;
> +		goto out;
>  	if (!alarm.enabled) {
>  		dev_dbg(pdev->dev, "alarm not enabled\n");
> -		return 0;
> +		rc = 0;
> +		goto out;
>  	}
>  	rc = rtc_read_time(rtc_device, &tm);
>  	if (rc)
> -		return rc;
> +		goto out;
>  	then = rtc_tm_to_time64(&alarm.time);
>  	now = rtc_tm_to_time64(&tm);
>  	duration = then-now;
>  
>  	/* in the past */
> -	if (then < now)
> -		return 0;
> +	if (then < now) {
> +		rc = 0;
> +		goto out;
> +	}
>  
>  	/* will be stored in upper 16 bits of s0i3 hint argument,
>  	 * so timer wakeup from s0i3 is limited to ~18 hours or less
>  	 */
> -	if (duration <= 4 || duration > U16_MAX)
> -		return -EINVAL;
> +	if (duration <= 4 || duration > U16_MAX) {
> +		rc = -EINVAL;
> +		goto out;
> +	}
>  
>  	*arg |= (duration << 16);
>  	rc = rtc_alarm_irq_enable(rtc_device, 0);
>  	pm_pr_dbg("wakeup timer programmed for %lld seconds\n", duration);
>  
> +out:
> +	rtc_class_close(rtc_device);
>  	return rc;
>  }

I was expecting 2 patch series, with one which adds the DEFINE_FREE() into 
the public rtc header, not going back to the old approach. You also seemed 
to forget to increase the version number in the subject.

-- 
 i.