Hello Marek, > On Thu, 26 Oct 2023 18:17:59 +0200 > Marek Behún <kabel@xxxxxxxxxx> wrote: > >> + mcu->gc.request = omnia_gpio_request; >> + mcu->gc.get_direction = omnia_gpio_get_direction; >> + mcu->gc.direction_input = omnia_gpio_direction_input; >> + mcu->gc.direction_output = omnia_gpio_direction_output; >> + mcu->gc.get = omnia_gpio_get; >> + mcu->gc.get_multiple = omnia_gpio_get_multiple; >> + mcu->gc.set = omnia_gpio_set; >> + mcu->gc.set_multiple = omnia_gpio_set_multiple; >> + mcu->gc.init_valid_mask = omnia_gpio_init_valid_mask; >> + mcu->gc.can_sleep = true; >> + mcu->gc.names = omnia_mcu_gpio_templates; >> + mcu->gc.base = -1; >> + mcu->gc.ngpio = ARRAY_SIZE(omnia_gpios); >> + mcu->gc.label = "Turris Omnia MCU GPIOs"; >> + mcu->gc.parent = dev; >> + mcu->gc.owner = THIS_MODULE; > > Sigh. I will need to add a custom of_xlate here that accepts 3 cells > instead of 2. The reason is that Pali prepared U-Boot code last year to > patch reset-gpios into pcie nodes if MCU node is present in the > devicetree, and back then we decided that the binding should be > 3-celled: > reset-gpios = <&mcu BANK GPIO FLAGS>; > > So now the pcie controllers do not work because OF cannot give proper > GPIOs. > > I will fix this in v5. I was reviewing the seris in order to apply it, but I didn't find your v5 in the mainling list. Did you send it ? Regards, Gregory -- Gregory Clement, Bootlin Embedded Linux and Kernel engineering http://bootlin.com