On Mon, Sep 21, 2020 at 02:50:56PM +0300, Andy Shevchenko wrote: > On Sun, Sep 06, 2020 at 03:43:47PM -0600, Simon Glass wrote: > > Generate ACPI information for this device so that Linux can use it > > correctly. > > > + ret = acpi_device_write_interrupt_or_gpio(ctx, (struct udevice *)dev, > > + "ready-gpios"); > > + if (ret < 0) > > + return log_msg_ret("irq_gpio", ret); > > I looked a bit closer at the acpi_table.c and would like to emphasize on > lessons learn from BIOS mistakes found in the wild with GPIOs. > > Because GPIO resources are quite badly described in ACPI (it seems MS failed > to > deliver GPIO abstraction to ACPI and to Windows API), there are some corner > cases, in order to mitigate which we need to consider the following to avoid > potential glitches and misconfiguration: > > - GpioIo() doesn't have any means of Active Low / High setting, the _DSD must > be provided to mitigate this. > > - GpioIo() doesn't properly communicate the initial state of the output pin, > thus Linux assumes the simple rule: > > Pull Bias Polarity Requested... > > Implicit x AS IS (assumed firmware configured for us) > Explicit x (no _DSD) as Pull Bias (Up == High, Down == Low), > assuming non-active (Polarity = !Pull Bias) > > Down Low as low, assuming active > Down High as high, assuming non-active > Up Low as low, assuming non-active > Up High as high, assuming active > > Hopefully this helps (and maybe can be added to some documentation). > > P.S. Why I2cSerialBus() and not I2cSerialBusV2() ?
Forgot one important aspect. GpioIo() can be used as interrupt and in this case the IoRestriction mustn't be OutputOnly. It also requires active_low flag from _DSD in cases where it's needed (better to always provide than rely on above assumption made on OS level). -- With Best Regards, Andy Shevchenko