Re: [PATCH fix] Revert "USB: bcma: Add a check for devm_gpiod_get"

[Geert Uytterhoeven]

	Hi Rafał,

On Tue, 31 Aug 2021, Rafał Miłecki wrote:
> From: Rafał Miłecki <rafal@xxxxxxxxxx>
>
> This reverts commit f3de5d857bb2362b00e2a8d4bc886cd49dcb66db.
>
> That commit broke USB on all routers that have USB always powered on and
> don't require toggling any GPIO. It's a majority of devices actually.
>
> The original code worked and seemed safe: vcc GPIO is optional and
> bcma_hci_platform_power_gpio() takes care of checking the pointer before
> using it.
>
> This revert fixes:
> [   10.801127] bcma_hcd: probe of bcma0:11 failed with error -2
>
> Cc: Chuhong Yuan <hslester96@xxxxxxxxx>
> Signed-off-by: Rafał Miłecki <rafal@xxxxxxxxxx>

Thanks for your patch, which is being backported to stable.

> --- a/drivers/usb/host/bcma-hcd.c
> +++ b/drivers/usb/host/bcma-hcd.c
> @@ -406,12 +406,9 @@ static int bcma_hcd_probe(struct bcma_device *core)
> 		return -ENOMEM;
> 	usb_dev->core = core;
>
> -	if (core->dev.of_node) {
> +	if (core->dev.of_node)
> 		usb_dev->gpio_desc = devm_gpiod_get(&core->dev, "vcc",
> 						    GPIOD_OUT_HIGH);


> -		if (IS_ERR(usb_dev->gpio_desc))
> -			return PTR_ERR(usb_dev->gpio_desc);
> -	}

This means real errors and probe deferral are no longer handled.
What about using devm_gpiod_get_optional() instead?
After that, the check in bcma_hci_platform_power_gpio() can be removed,
as gpiod_set_value() handles NULL pointers fine.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds