| Message ID | 20190529101342.30673-1-ard.biesheuvel@linaro.org |
|---|---|
| State | New |
| Headers | show |
| Series | [RFC] gpio/acpi: set direction for ACPI OpRegion GPIOs | expand |
On Wed, May 29, 2019 at 12:13:42PM +0200, Ard Biesheuvel wrote: > The ACPI GPIO OpRegion code never sets the direction for GPIOs that > it configures from scratch, leaving unconfigured GPIOs to be left > at the default direction, which is typically input not output. The code does: enum gpiod_flags flags = acpi_gpio_to_gpiod_flags(agpio); before it requests the GPIO. Isn't that sufficient?
On Wed, 29 May 2019 at 12:27, Mika Westerberg <mika.westerberg@linux.intel.com> wrote: > > On Wed, May 29, 2019 at 12:13:42PM +0200, Ard Biesheuvel wrote: > > The ACPI GPIO OpRegion code never sets the direction for GPIOs that > > it configures from scratch, leaving unconfigured GPIOs to be left > > at the default direction, which is typically input not output. > > The code does: > > enum gpiod_flags flags = acpi_gpio_to_gpiod_flags(agpio); > > before it requests the GPIO. Isn't that sufficient? Yes, it works if I set the PullXxx value explicitly in the GpioIo() resource.
diff --git a/drivers/gpio/gpiolib-acpi.c b/drivers/gpio/gpiolib-acpi.c index c9fc9e232aaf..02a2bc7fe18b 100644 --- a/drivers/gpio/gpiolib-acpi.c +++ b/drivers/gpio/gpiolib-acpi.c @@ -966,6 +966,12 @@ acpi_gpio_adr_space_handler(u32 function, acpi_physical_address address, goto out; } + if (function == ACPI_WRITE) + gpiod_direction_output(desc, + !!((1 << i) & *value)); + else + gpiod_direction_input(desc); + conn->pin = pin; conn->desc = desc; list_add_tail(&conn->node, &achip->conns);
The ACPI GPIO OpRegion code never sets the direction for GPIOs that it configures from scratch, leaving unconfigured GPIOs to be left at the default direction, which is typically input not output. So set the direction and default value explicitly. Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> --- drivers/gpio/gpiolib-acpi.c | 6 ++++++ 1 file changed, 6 insertions(+)