Re: [PATCH 2/3] serial: mxs-auart: use devm resources for iomem and GPIO IRQs

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Missing `clk_disable_unprepare()` for `s->clk` in `mxs_auart_remove()` causes a clock enable count leak.
--

commit bb27ea241e1f8b8f030c765ac616b8c5f0b20802
Author: Rosen Penev <rosenp@gmail.com>

serial: mxs-auart: use devm resources for iomem and GPIO IRQs

This commit replaces platform_get_resource and ioremap with
devm_platform_get_and_ioremap_resource. It also converts GPIO IRQ requests to
use devm_request_irq, eliminating the need for manual cleanup and simplifying
the remove function.

> diff --git a/drivers/tty/serial/mxs-auart.c b/drivers/tty/serial/mxs-auart.c
> --- a/drivers/tty/serial/mxs-auart.c
> +++ b/drivers/tty/serial/mxs-auart.c

[ ... ]

> @@ -1683,11 +1656,7 @@ static void mxs_auart_remove(struct platform_device *pdev)
>  	struct mxs_auart_port *s = platform_get_drvdata(pdev);
>  
>  	uart_remove_one_port(&auart_driver, &s->port);
> -	auart_port[pdev->id] = NULL;
> -	mxs_auart_free_gpio_irq(s);
> -	iounmap(s->port.membase);
> -	if (is_asm9260_auart(s))
> -		clk_disable_unprepare(s->clk);

[Severity: Medium]
Does removing clk_disable_unprepare() here cause a clock enable count leak
when the driver is unbound?

Looking at mxs_auart_probe(), the driver acquires s->clk using devm_clk_get(),
but it is explicitly prepared and enabled via clk_prepare_enable(s->clk).
Because it cannot use devm_clk_get_enabled() (its rate must be set while the
clock is disabled), its lifecycle is unmanaged.

While subsequent patches might disable the clock for non-console ports at the
end of probe(), console ports leave it enabled. Won't deleting this manual
cleanup leave the clock's prepare and enable counts unbalanced on unbind?

> +	auart_port[s->port.line] = NULL;
>  }

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260603025857.287148-1-rosenp@gmail.com?part=2