We check the padding of other uAPI v2 structures but not that of line
config attributes. For used attributes: check if their padding is
zeroed, for unused: check if the entire structure is zeroed.

Fixes: 3c0d9c635ae2 ("gpiolib: cdev: support GPIO_V2_GET_LINE_IOCTL and GPIO_V2_LINE_GET_VALUES_IOCTL")
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
---
Changes in v2:
- Make checking even stricter: check if padding is zeroed for used
  attributes, for unused ones: check if the entire struct is zeroed
- Link to v1: https://patch.msgid.link/20260519-gpio-cdev-attr-padding-check-v1-1-a0c6d4a698bf@oss.qualcomm.com
---
 drivers/gpio/gpiolib-cdev.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/drivers/gpio/gpiolib-cdev.c b/drivers/gpio/gpiolib-cdev.c
index f36b7c06996d70b2286edbd181899e4c572b9086..edbcc86e4b26f88036ed12c13055bb2c371fb6a3 100644
--- a/drivers/gpio/gpiolib-cdev.c
+++ b/drivers/gpio/gpiolib-cdev.c
@@ -1194,6 +1194,15 @@ static int gpio_v2_line_config_validate(struct gpio_v2_line_config *lc,
 	if (!mem_is_zero(lc->padding, sizeof(lc->padding)))
 		return -EINVAL;
 
+	for (i = 0; i < lc->num_attrs; i++) {
+		if (lc->attrs[i].attr.padding != 0)
+			return -EINVAL;
+	}
+
+	if (!mem_is_zero(&lc->attrs[i],
+			 (GPIO_V2_LINE_NUM_ATTRS_MAX - lc->num_attrs) * sizeof(*lc->attrs)))
+		return -EINVAL;
+
 	for (i = 0; i < num_lines; i++) {
 		flags = gpio_v2_line_config_flags(lc, i);
 		ret = gpio_v2_line_flags_validate(flags);

---
base-commit: 6a50ba100ace43f43c87384367eb2d2605fcc16c
change-id: 20260519-gpio-cdev-attr-padding-check-7e52c98a3de7

Best regards,
-- 
Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>

On Wed, May 20, 2026 at 09:42:10AM +0200, Bartosz Golaszewski wrote:
> We check the padding of other uAPI v2 structures but not that of line
> config attributes. For used attributes: check if their padding is
> zeroed, for unused: check if the entire structure is zeroed.
> 
> Fixes: 3c0d9c635ae2 ("gpiolib: cdev: support GPIO_V2_GET_LINE_IOCTL and GPIO_V2_LINE_GET_VALUES_IOCTL")
> Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
> ---
> Changes in v2:
> - Make checking even stricter: check if padding is zeroed for used
>   attributes, for unused ones: check if the entire struct is zeroed
> - Link to v1: https://patch.msgid.link/20260519-gpio-cdev-attr-padding-check-v1-1-a0c6d4a698bf@oss.qualcomm.com
> ---
>  drivers/gpio/gpiolib-cdev.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/drivers/gpio/gpiolib-cdev.c b/drivers/gpio/gpiolib-cdev.c
> index f36b7c06996d70b2286edbd181899e4c572b9086..edbcc86e4b26f88036ed12c13055bb2c371fb6a3 100644
> --- a/drivers/gpio/gpiolib-cdev.c
> +++ b/drivers/gpio/gpiolib-cdev.c
> @@ -1194,6 +1194,15 @@ static int gpio_v2_line_config_validate(struct gpio_v2_line_config *lc,
>  	if (!mem_is_zero(lc->padding, sizeof(lc->padding)))
>  		return -EINVAL;
>  
> +	for (i = 0; i < lc->num_attrs; i++) {
> +		if (lc->attrs[i].attr.padding != 0)
> +			return -EINVAL;
> +	}
> +

That works for me - the explict test against 0 makes the intent as clear
as a mem_is_zero() call. For me anyway.


> +	if (!mem_is_zero(&lc->attrs[i],
> +			 (GPIO_V2_LINE_NUM_ATTRS_MAX - lc->num_attrs) * sizeof(*lc->attrs)))
> +		return -EINVAL;
> +

Dislike the use of i here.
lc->num_attrs would remove the implicit dependency on the preceeding loop.

I'm also uncomfortable with the lc->num_attrs == GPIO_V2_LINE_NUM_ATTRS_MAX
case as the pointer is out of range.  The size is 0 so it probably wont get
dereferenced, but just passing it around makes my skin crawl.
Perhaps put the size in a var and only call mem_is_zero() if size > 0?

Cheers,
Kent.