linux-kernel - Re: [PATCH] reset: core: reset-gpio: Suppress registration error for optional resets

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <bf3de9ecf10a3f6a63a6f2e235369519cb2f63e8.camel@pengutronix.de>
Date: Mon, 06 Oct 2025 15:09:23 +0200
From: Philipp Zabel <p.zabel@...gutronix.de>
To: Andre Przywara <andre.przywara@....com>, Krzysztof Kozlowski
	 <krzysztof.kozlowski@...aro.org>
Cc: linux-sunxi@...ts.linux.dev, linux-kernel@...r.kernel.org, 
	linux-arm-kernel@...ts.infradead.org
Subject: Re: [PATCH] reset: core: reset-gpio: Suppress registration error
 for optional resets

Hi Andre,

On Do, 2025-10-02 at 23:28 +0100, Andre Przywara wrote:
> For reset controllers that are marked as optional, we should skip errors
> during probing and return NULL, to avoid unnecessary failures. The
> reset-gpio controller does this mostly, but returns the true error in
> case the __reset_add_reset_gpio_device() call fails. Treat this call the
> same as the other registration errors, and consider the optional flag.
> 
> One could argue that at this point it's a proper error that should not
> be ignored anymore, but in case of the reset-gpio controller this is not
> entirely true, since the code at the moment does not support GPIO
> controllers with three #gpio-cells - there is a TODO comment about this
> in that said function. So to avoid unnecessary probe fails for devices
> using reset-gpios (it's an optional reset after all), let's treat an
> error as still optional at this point.
> 
> This fixes operation of WiFi chips on Allwinner boards, where some use
> reset-gpios, and which currently fail because all Allwinner SoCs use
> GPIO controllers with 3 cells.
> 
> Signed-off-by: Andre Przywara <andre.przywara@....com>
> ---
>  drivers/reset/core.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/reset/core.c b/drivers/reset/core.c
> index 22f67fc77ae53..c2ccd08fb36e1 100644
> --- a/drivers/reset/core.c
> +++ b/drivers/reset/core.c
> @@ -1044,7 +1044,7 @@ __of_reset_control_get(struct device_node *node, const char *id, int index,
>  
>  		ret = __reset_add_reset_gpio_device(&args);
>  		if (ret) {
> -			rstc = ERR_PTR(ret);
> +			rstc = optional ? NULL : ERR_PTR(ret);

Please move this into __reset_add_reset_gpio_device() and pass flags
into it. Only the -ENOENT return value mentioned in the commit
description should be suppressed by the optional flag, not any -ENOMEM.
And especially not the return value of __reset_add_reset_gpio_lookup(),
which may be -EPROBE_DEFER.

regards
Philipp