On Mon, Jan 14, 2019 at 01:30:03PM +0100, Karoly Pados wrote:
> There is a bug in the current GPIO code for ftdi_sio: it failed to take USB
> autosuspend into account. If the device is in autosuspend, calls to
> usb_control_msg() fail with -EHOSTUNREACH. Because the standard value for
> autosuspend timeout is usually 2-5 seconds, this made it almost impossible
> to use the GPIOs on machines that have USB autosuspend enabled. This patch
> fixes the issue by acquiring a PM lock on the device for the duration of
> the USB transfers. Tested on an FT231X device.
>
> Signed-off-by: Karoly Pados <pados@xxxxxxxx>
> ---
> Please consider backporting to 4.20.x, otherwise the GPIO driver is not
> really usable for anybody with USB autosuspend enabled (eg. many laptops),
> at least not without manual configuration.
>
> drivers/usb/serial/ftdi_sio.c | 21 ++++++++++++++++++++-
> 1 file changed, 20 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/usb/serial/ftdi_sio.c b/drivers/usb/serial/ftdi_sio.c
> index 1ab2a6191013..01813dce37f2 100644
> --- a/drivers/usb/serial/ftdi_sio.c
> +++ b/drivers/usb/serial/ftdi_sio.c
> @@ -1783,6 +1783,13 @@ static int ftdi_set_bitmode(struct usb_serial_port *port, u8 mode)
> int result;
> u16 val;
>
> + result = usb_autopm_get_interface(port->serial->interface);
Both of the functions where you add autopm calls already have a local
variable for port->serial.
> + if (result) {
> + dev_err(&port->serial->interface->dev,
> + "Failed to wake device from autosuspend.\n");
And we tend not to log errors for these calls.
> + return result;
> + }
I replaced and port->serial with serial and dropped the dev_err:s before
applying for -rc3 with a Fixes and CC-stable tag.
Thanks,
Johan
