Re: [PATCH] rtc: stmp3xxx: use devm_platform_ioremap_resource()

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:

Pre-existing issues:
- [High] The return value of `platform_get_irq()` is not checked for errors, breaking deferred probing and unnecessarily disrupting hardware state.
- [High] The child watchdog platform device is never unregistered when the RTC driver is unbound, leading to a leaked device and potential Use-After-Free.
--

--- Patch [1]: [PATCH] rtc: stmp3xxx: use devm_platform_ioremap_resource() ---
commit eb7d944b8c0d01b4294ccf13b092d4f30d47bb4e
Author: Rosen Penev <rosenp@gmail.com>

    rtc: stmp3xxx: use devm_platform_ioremap_resource()
    
    Replace the open-coded platform_get_resource() plus devm_ioremap()
    sequence with a single devm_platform_ioremap_resource() call, which folds
    the resource lookup and mapping into one step and returns an ERR_PTR on
    failure, checked with IS_ERR() and propagated via PTR_ERR().

> diff --git a/drivers/rtc/rtc-stmp3xxx.c b/drivers/rtc/rtc-stmp3xxx.c
> index 7afcd14aeee56..05f128ccf8277 100644
> --- a/drivers/rtc/rtc-stmp3xxx.c
> +++ b/drivers/rtc/rtc-stmp3xxx.c
> @@ -245,28 +245,21 @@ static void stmp3xxx_rtc_remove(struct platform_device *pdev)

[Severity: High]
This is a pre-existing issue, but I noticed that stmp3xxx_rtc_remove() doesn't
unregister the dynamically added child platform device created by
stmp3xxx_wdt_register() during probe.

If the RTC driver is unbound, the child watchdog device remains registered.
A subsequent watchdog operation like stmp3xxx_wdt_set_timeout() will call
dev_get_drvdata(dev->parent), which returns NULL because the parent's drvdata
is cleared during unbind. This leads to a NULL pointer dereference when
accessing rtc_data->io.

Could this be fixed by saving the child device pointer and unregistering it
in the remove function?

>  static int stmp3xxx_rtc_probe(struct platform_device *pdev)
>  {
>  	struct stmp3xxx_rtc_data *rtc_data;

[ ... ]

> -	rtc_data->io = devm_ioremap(&pdev->dev, r->start, resource_size(r));
> -	if (!rtc_data->io) {
> -		dev_err(&pdev->dev, "ioremap failed\n");
> -		return -EIO;
> -	}
> -
> +	rtc_data->io = io;
>  	rtc_data->irq_alarm = platform_get_irq(pdev, 0);

[Severity: High]
This isn't a bug introduced by this patch, but does this lack of error checking
break deferred probing?

If platform_get_irq() returns -EPROBE_DEFER, it is assigned directly to
rtc_data->irq_alarm. Later, this negative error code is passed to
devm_request_irq() in stmp3xxx_rtc_probe():

    err = devm_request_irq(&pdev->dev, rtc_data->irq_alarm,
            stmp3xxx_rtc_interrupt, 0, "RTC alarm", &pdev->dev);

Since devm_request_irq() expects an unsigned int for the IRQ number, the
negative value is cast and fails with -EINVAL. The driver then returns
-EINVAL instead of -EPROBE_DEFER, which permanently prevents the RTC and
watchdog from probing on systems that require deferred probing for the
interrupt controller.

Should we add a check for negative error codes from platform_get_irq() here?

>  
>  	rtc_stat = readl(rtc_data->io + STMP3XXX_RTC_STAT);

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260715021933.1551663-1-rosenp@gmail.com?part=1