On Tue, Nov 08, 2016 at 02:40:06PM +0100, Arnd Bergmann wrote:
> The newly added acpi_gpiochip_scan_gpios function produces a few harmless
> warnings:
>
> drivers/gpio/gpiolib-acpi.c: In function ‘acpi_gpiochip_add’:
> drivers/gpio/gpiolib-acpi.c:925:7: error: ‘dflags’ may be used uninitialized
> in this function [-Werror=maybe-uninitialized]
> drivers/gpio/gpiolib-acpi.c:925:9: error: ‘lflags’ may be used uninitialized
> in this function [-Werror=maybe-uninitialized]
>
> The problem is that he compiler cannot know that a negative return value
> from fwnode_property_read_u32_array() or acpi_gpiochip_pin_to_gpio_offset()
> implies that the IS_ERR(gpio_desc) is true, as the value could in theory
> be below -MAX_ERRNO.
>
> The function already initializes its output values to zero, and moving
> that intialization a little higher up ensures that we can never have
> uninitialized data in the caller.
>
> Fixes: c80f1ba75df2 ("ACPI / gpio: Add hogging support")
> Signed-off-by: Arnd Bergmann <a...@arndb.de>
Acked-by: Mika Westerberg <mika.westerb...@linux.intel.com>
