Re: [net-next PATCH v4 01/15] Documentation: ACPI: DSD: Document MDIO PHY

["Rafael J. Wysocki" <rafael@kernel.org>]

On Fri, Jan 29, 2021 at 7:48 AM Calvin Johnson
<calvin.johnson@oss.nxp.com> wrote:
>
> On Thu, Jan 28, 2021 at 02:27:00PM +0100, Rafael J. Wysocki wrote:
> > On Thu, Jan 28, 2021 at 2:12 PM Calvin Johnson
> > <calvin.johnson@oss.nxp.com> wrote:
> > >
> > > On Thu, Jan 28, 2021 at 01:00:40PM +0100, Rafael J. Wysocki wrote:
> > > > On Thu, Jan 28, 2021 at 12:27 PM Calvin Johnson
> > > > <calvin.johnson@oss.nxp.com> wrote:
> > > > >
> > > > > Hi Rafael,
> > > > >
> > > > > Thanks for the review. I'll work on all the comments.
> > > > >
> > > > > On Fri, Jan 22, 2021 at 08:22:21PM +0100, Rafael J. Wysocki wrote:
> > > > > > On Fri, Jan 22, 2021 at 4:43 PM Calvin Johnson
> > > > > > <calvin.johnson@oss.nxp.com> wrote:
> > > > > > >
> > > > > > > Introduce ACPI mechanism to get PHYs registered on a MDIO bus and
> > > > > > > provide them to be connected to MAC.
> > > > > > >
> > > > > > > Describe properties "phy-handle" and "phy-mode".
> > > > > > >
> > > > > > > Signed-off-by: Calvin Johnson <calvin.johnson@oss.nxp.com>
> > > > > > > ---
> > > > > > >
> > > > > > > Changes in v4:
> > > > > > > - More cleanup
> > > > > >
> > > > > > This looks much better that the previous versions IMV, some nits below.
> > > > > >
> > > > > > > Changes in v3: None
> > > > > > > Changes in v2:
> > > > > > > - Updated with more description in document
> > > > > > >
> > > > > > >  Documentation/firmware-guide/acpi/dsd/phy.rst | 129 ++++++++++++++++++
> > > > > > >  1 file changed, 129 insertions(+)
> > > > > > >  create mode 100644 Documentation/firmware-guide/acpi/dsd/phy.rst
> > > > > > >
> > > > > > > diff --git a/Documentation/firmware-guide/acpi/dsd/phy.rst b/Documentation/firmware-guide/acpi/dsd/phy.rst
> > > > > > > new file mode 100644
> > > > > > > index 000000000000..76fca994bc99
> > > > > > > --- /dev/null
> > > > > > > +++ b/Documentation/firmware-guide/acpi/dsd/phy.rst
> > > > > > > @@ -0,0 +1,129 @@
> > > > > > > +.. SPDX-License-Identifier: GPL-2.0
> > > > > > > +
> > > > > > > +=========================
> > > > > > > +MDIO bus and PHYs in ACPI
> > > > > > > +=========================
> > > > > > > +
> > > > > > > +The PHYs on an MDIO bus [1] are probed and registered using
> > > > > > > +fwnode_mdiobus_register_phy().
> > > > > >
> > > > > > Empty line here, please.
> > > > > >
> > > > > > > +Later, for connecting these PHYs to MAC, the PHYs registered on the
> > > > > > > +MDIO bus have to be referenced.
> > > > > > > +
> > > > > > > +The UUID given below should be used as mentioned in the "Device Properties
> > > > > > > +UUID For _DSD" [2] document.
> > > > > > > +   - UUID: daffd814-6eba-4d8c-8a91-bc9bbf4aa301
> > > > > >
> > > > > > I would drop the above paragraph.
> > > > > >
> > > > > > > +
> > > > > > > +This document introduces two _DSD properties that are to be used
> > > > > > > +for PHYs on the MDIO bus.[3]
> > > > > >
> > > > > > I'd say "for connecting PHYs on the MDIO bus [3] to the MAC layer."
> > > > > > above and add the following here:
> > > > > >
> > > > > > "These properties are defined in accordance with the "Device
> > > > > > Properties UUID For _DSD" [2] document and the
> > > > > > daffd814-6eba-4d8c-8a91-bc9bbf4aa301 UUID must be used in the Device
> > > > > > Data Descriptors containing them."
> > > > > >
> > > > > > > +
> > > > > > > +phy-handle
> > > > > > > +----------
> > > > > > > +For each MAC node, a device property "phy-handle" is used to reference
> > > > > > > +the PHY that is registered on an MDIO bus. This is mandatory for
> > > > > > > +network interfaces that have PHYs connected to MAC via MDIO bus.
> > > > > > > +
> > > > > > > +During the MDIO bus driver initialization, PHYs on this bus are probed
> > > > > > > +using the _ADR object as shown below and are registered on the MDIO bus.
> > > > > >
> > > > > > Do you want to mention the "reg" property here?  I think it would be
> > > > > > useful to do that.
> > > > >
> > > > > No. I think we should adhere to _ADR in MDIO case. The "reg" property for ACPI
> > > > > may be useful for other use cases that Andy is aware of.
> > > >
> > > > The code should reflect this, then.  I mean it sounds like you want to
> > > > check the "reg" property only if this is a non-ACPI node.
> > >
> > > Right. For MDIO case, that is what is required.
> > > "reg" for DT and "_ADR" for ACPI.
> > >
> > > However, Andy pointed out [1] that ACPI nodes can also hold reg property and
> > > therefore, fwnode_get_id() need to be capable to handling that situation as
> > > well.
> >
> > No, please don't confuse those two things.
> >
> > Yes, ACPI nodes can also hold a "reg" property, but the meaning of it
> > depends on the binding which is exactly my point: _ADR is not a
> > fallback replacement for "reg" in general and it is not so for MDIO
> > too.  The new function as proposed doesn't match the MDIO requirements
> > and so it should not be used for MDIO.
> >
> > For MDIO, the exact flow mentioned above needs to be implemented (and
> > if someone wants to use it for their use case too, fine).
> >
> > Otherwise the code wouldn't match the documentation.
>
> In that case, is this good?

It would work, but I would introduce a wrapper around the _ADR
evaluation, something like:

int acpi_get_local_address(acpi_handle handle, u32 *addr)
{
      unsigned long long adr;
      acpi_status status;

      status = acpi_evaluate_integer(handle, METHOD_NAME__ADR, NULL, &adr);
      if (ACPI_FAILURE(status))
                return -ENODATA;

      *addr = (u32)adr;
      return 0;
}

in drivers/acpi/utils.c and add a static inline stub always returning
-ENODEV for it for !CONFIG_ACPI.

> /**
>  * fwnode_get_local_addr - Get the local address of fwnode.
>  * @fwnode: firmware node
>  * @addr: addr value contained in the fwnode
>  *
>  * For DT, retrieve the value of the "reg" property for @fwnode.
>  *
>  * In the ACPI case, evaluate the _ADR object located under the
>  * given node, if present, and provide its return value to the
>  * caller.
>  *
>  * Return 0 on success or a negative error code.
>  */
> int fwnode_get_local_addr(struct fwnode_handle *fwnode, u32 *addr)
> {
>         int ret;
>
>         if (is_of_node(fwnode))
>                 return of_property_read_u32(to_of_node(fwnode), "reg", addr);

So you can write the below as

if (is_acpi_device_node(fwnode))
    return acpi_get_local_address(ACPI_HANDLE_FWNODE(fwnode), addr);

return -EINVAL;

and this should compile just fine if CONFIG_ACPI is unset, so you can
avoid the whole #ifdeffery in this function.

>
> #ifdef CONFIG_ACPI
>         if (is_acpi_node(fwnode)) {
>                 unsigned long long adr;
>                 acpi_status status;
>
>                 status = acpi_evaluate_integer(ACPI_HANDLE_FWNODE(fwnode),
>                                                METHOD_NAME__ADR, NULL, &adr);
>                 if (ACPI_FAILURE(status))
>                         return -ENODATA;
>                 *addr = (u32)adr;
>                 return 0;
>         }
> #endif
>         return -EINVAL;
> }