Re: [PATCH] gpiolib: acpi: Add ACPI device NULL check to acpi_can_fallback_to_crs()

From: Andy Shevchenko
Date: Fri May 10 2024 - 10:46:31 EST


On Thu, May 09, 2024 at 12:46:05PM +0200, Laura Nao wrote:
> Check ACPI device for NULL inside acpi_can_fallback_to_crs(), so callers
> won't need to.

Thank you for the patch, one change seems good to have along this.

..

> /* Never allow fallback if the device has properties */
> - if (acpi_dev_has_props(adev) || adev->driver_gpios)
> + if (!adev || acpi_dev_has_props(adev) || adev->driver_gpios)

Right, since it was adev || _crs() combined.

> return false;

Now we may remove that check from __acpi_find_gpio():

--- a/drivers/gpio/gpiolib-acpi.c
+++ b/drivers/gpio/gpiolib-acpi.c
@@ -988,10 +988,10 @@ __acpi_find_gpio(struct fwnode_handle *fwnode, const char *con_id, unsigned int
}

/* Then from plain _CRS GPIOs */
- if (!adev || !can_fallback)
- return ERR_PTR(-ENOENT);
+ if (can_fallback)
+ return acpi_get_gpiod_by_index(adev, NULL, idx, info);

- return acpi_get_gpiod_by_index(adev, NULL, idx, info);
+ return ERR_PTR(-ENOENT);
}

struct gpio_desc *acpi_find_gpio(struct fwnode_handle *fwnode,


As a side effect it will make the comment better to understand.

With above suggestion applied, feel free to add mine
Reviewed-by: Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>

You might need to rephrase the commit message to say that

"We also move the check in additional to the moving the function call
outside of __acpi_find_gpio()."

or something similar, up to you.

--
With Best Regards,
Andy Shevchenko