5.4-stable review patch. If anyone has any objections, please let me know. ------------------ From: Rouven Czerwinski <r.czerwinski@xxxxxxxxxxxxxx> commit 23484d817082c3005252d8edfc8292c8a1006b5b upstream. Fix the undefined usage of the GPIO consumer API after retrieving the GPIO description with GPIO_ASIS. The API documentation mentions that GPIO_ASIS won't set a GPIO direction and requires the user to set a direction before using the GPIO. This can be confirmed on i.MX6 hardware, where rfkill-gpio is no longer able to enabled/disable a device, presumably because the GPIO controller was never configured for the output direction. Fixes: b2f750c3a80b ("net: rfkill: gpio: prevent value glitch during probe") Cc: stable@xxxxxxxxxxxxxxx Signed-off-by: Rouven Czerwinski <r.czerwinski@xxxxxxxxxxxxxx> Link: https://msgid.link/20231207075835.3091694-1-r.czerwinski@xxxxxxxxxxxxxx Signed-off-by: Johannes Berg <johannes.berg@xxxxxxxxx> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> --- net/rfkill/rfkill-gpio.c | 8 ++++++++ 1 file changed, 8 insertions(+) --- a/net/rfkill/rfkill-gpio.c +++ b/net/rfkill/rfkill-gpio.c @@ -116,6 +116,14 @@ static int rfkill_gpio_probe(struct plat return -EINVAL; } + ret = gpiod_direction_output(rfkill->reset_gpio, true); + if (ret) + return ret; + + ret = gpiod_direction_output(rfkill->shutdown_gpio, true); + if (ret) + return ret; + rfkill->rfkill_dev = rfkill_alloc(rfkill->name, &pdev->dev, rfkill->type, &rfkill_gpio_ops, rfkill);