Re: [PATCH] watchdog: ath79_wdt: select OF and COMPILE_TEST

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [High] Permanent failure of deferred probing and device re-binding due to uncleared global `wdt_base` pointer on error paths.
- [Medium] Mismatch between Kconfig changes and commit message regarding OF dependency.
--

commit 279311df3b831cb8a967974beeea1a51baecd0ac
Author: Rosen Penev <rosenp@gmail.com>

watchdog: ath79_wdt: select OF and COMPILE_TEST

This has always been an OF driver. Select OF in Kconfig.

Add COMPILE_TEST for all platforms, not just ARM.

> watchdog: ath79_wdt: select OF and COMPILE_TEST
> 
> This has always been an OF driver. Select OF in Kconfig.

There is a discrepancy between the commit message and the implemented code.
The commit message states it will "select OF", but the patch implements
"depends on OF". Could the commit message be updated to match the
implementation?

> diff --git a/drivers/watchdog/Kconfig b/drivers/watchdog/Kconfig
> index dc78729ba2a5d..1590f82941ba8 100644
> --- a/drivers/watchdog/Kconfig
> +++ b/drivers/watchdog/Kconfig
> @@ -1811,7 +1811,8 @@ config M54xx_WATCHDOG
>  
>  config ATH79_WDT
>  	tristate "Atheros AR71XX/AR724X/AR913X hardware watchdog"
> -	depends on ATH79 || (ARM && COMPILE_TEST)
> +	depends on ATH79 || COMPILE_TEST
> +	depends on OF

[ ... ]

While reviewing this patch, I noticed an existing issue in the probe
function's error handling. The driver uses a global pointer to enforce a
singleton instance:

drivers/watchdog/ath79_wdt.c:ath79_wdt_probe() {
    ...
	if (wdt_base)
		return -EBUSY;

	wdt_base = devm_platform_ioremap_resource(pdev, 0);
	if (IS_ERR(wdt_base))
		return PTR_ERR(wdt_base);

	wdt_clk = devm_clk_get_enabled(&pdev->dev, "wdt");
	if (IS_ERR(wdt_clk))
		return PTR_ERR(wdt_clk);
    ...
}

If devm_platform_ioremap_resource() returns an error pointer, or if a
later step like devm_clk_get_enabled() returns -EPROBE_DEFER, the global
wdt_base pointer is never cleared.

Does this permanently break deferred probing and device re-binding? On a
second probe attempt, the uncleared wdt_base pointer would still evaluate
to true, causing the probe to immediately fail with -EBUSY.

Similarly, since ath79_wdt_remove() does not clear wdt_base, will unloading
and reloading the driver also fail?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260518210957.10545-1-rosenp@gmail.com?part=1