On Mon, Nov 20, 2023 at 10:31 AM Sakari Ailus <sakari.ailus@xxxxxxxxxxxxxxx> wrote: > > Hi Laurent, > > On Sat, Nov 18, 2023 at 08:50:49PM +0200, Laurent Pinchart wrote: > > Hi Sakari, > > > > Thank you for the patch. > > > > On Fri, Nov 17, 2023 at 01:14:29PM +0200, Sakari Ailus wrote: > > > Document that acpi_dev_state_d0() can be used to tell if the device was > > > powered on for probe. > > > > > > Signed-off-by: Sakari Ailus <sakari.ailus@xxxxxxxxxxxxxxx> > > > --- > > > Documentation/firmware-guide/acpi/non-d0-probe.rst | 8 ++++++++ > > > 1 file changed, 8 insertions(+) > > > > > > diff --git a/Documentation/firmware-guide/acpi/non-d0-probe.rst b/Documentation/firmware-guide/acpi/non-d0-probe.rst > > > index 7afd16701a02..815bcc8db69f 100644 > > > --- a/Documentation/firmware-guide/acpi/non-d0-probe.rst > > > +++ b/Documentation/firmware-guide/acpi/non-d0-probe.rst > > > @@ -24,6 +24,14 @@ there's a problem with the device, the driver likely probes just fine but the > > > first user will find out the device doesn't work, instead of a failure at probe > > > time. This feature should thus be used sparingly. > > > > > > +ACPI framework > > > +-------------- > > > + > > > +Use the Linux ACPI framework function :c:func:`acpi_dev_state_d0()` to tell > > > +whether the device was powered on for probe. :c:func:`acpi_dev_state_d0()` > > > +returns true if the device is powered on, false otherwise. For non-ACPI backed > > > +devices it returns true always. > > > + > > > > While this is true, I don't want to see drivers having to call > > ACPI-specific functions, the same way you dislike OF-specific functions > > in drivers. Please find a better way to handle this. > > The functionality is only available on ACPI and the function does the right > thing on non-ACPI platforms. I don't see an issue here. The issue would be calling an ACPI-specific function from code that's otherwise firmware-agnostic, AFAICS. It would be good to have a more generic way of checking whether or not a device is operational. > Feel free to post DT binding patches on suggested device power state during > probe. :-) I think DT would benefit from this as well: the at24 driver is > widely used and suddenly making probe() not talk to the chip (or even power > it up) at all would probably be seen as a regression. In the DT case it is more complicated, though, at least in general, because there may be multiple clocks and regulators the device depends on and you may need to toggle a GPIO line too.
