Re: [PATCH net-next v6 1/4] net: phy: add USXGMII link partner ability constants

[Michael Walle]

Am 2020-07-16 00:44, schrieb Vladimir Oltean:
> On Wed, Jul 15, 2020 at 10:33:23PM +0200, Michael Walle wrote:
> > Am 2020-07-13 20:37, schrieb Michael Walle:
> > > Am 2020-07-13 20:23, schrieb Russell King - ARM Linux admin:
> > > > On Thu, Jul 09, 2020 at 11:35:23PM +0200, Michael Walle wrote:
> > > > > The constants are taken from the USXGMII Singleport Copper Interface
> > > > > specification. The naming are based on the SGMII ones, but with
> > > > > an MDIO_
> > > > > prefix.
> > > > >
> > > > > Signed-off-by: Michael Walle <michael@xxxxxxxx>
> > > > > ---
> > > > >  include/uapi/linux/mdio.h | 26 ++++++++++++++++++++++++++
> > > > >  1 file changed, 26 insertions(+)
> > > > >
> > > > > diff --git a/include/uapi/linux/mdio.h b/include/uapi/linux/mdio.h
> > > > > index 4bcb41c71b8c..784723072578 100644
> > > > > --- a/include/uapi/linux/mdio.h
> > > > > +++ b/include/uapi/linux/mdio.h
> > > > > @@ -324,4 +324,30 @@ static inline __u16 mdio_phy_id_c45(int
> > > > > prtad, int devad)
> > > > >  	return MDIO_PHY_ID_C45 | (prtad << 5) | devad;
> > > > >  }
> > > > >
> > > > > +/* UsxgmiiChannelInfo[15:0] for USXGMII in-band auto-negotiation.*/
> > > > > +#define MDIO_LPA_USXGMII_EEE_CLK_STP	0x0080	/* EEE clock stop
> > > > > supported */
> > > > > +#define MDIO_LPA_USXGMII_EEE		0x0100	/* EEE supported */
> > > > > +#define MDIO_LPA_USXGMII_SPD_MASK	0x0e00	/* USXGMII speed mask */
> > > > > +#define MDIO_LPA_USXGMII_FULL_DUPLEX	0x1000	/* USXGMII full
> > > > > duplex */
> > > > > +#define MDIO_LPA_USXGMII_DPX_SPD_MASK	0x1e00	/* USXGMII duplex
> > > > > and speed bits */
> > > > > +#define MDIO_LPA_USXGMII_10		0x0000	/* 10Mbps */
> > > > > +#define MDIO_LPA_USXGMII_10HALF		0x0000	/* 10Mbps half-duplex */
> > > > > +#define MDIO_LPA_USXGMII_10FULL		0x1000	/* 10Mbps full-duplex */
> > > > > +#define MDIO_LPA_USXGMII_100		0x0200	/* 100Mbps */
> > > > > +#define MDIO_LPA_USXGMII_100HALF	0x0200	/* 100Mbps half-duplex */
> > > > > +#define MDIO_LPA_USXGMII_100FULL	0x1200	/* 100Mbps full-duplex */
> > > > > +#define MDIO_LPA_USXGMII_1000		0x0400	/* 1000Mbps */
> > > > > +#define MDIO_LPA_USXGMII_1000HALF	0x0400	/* 1000Mbps half-duplex */
> > > > > +#define MDIO_LPA_USXGMII_1000FULL	0x1400	/* 1000Mbps full-duplex */
> > > > > +#define MDIO_LPA_USXGMII_10G		0x0600	/* 10Gbps */
> > > > > +#define MDIO_LPA_USXGMII_10GHALF	0x0600	/* 10Gbps half-duplex */
> > > > > +#define MDIO_LPA_USXGMII_10GFULL	0x1600	/* 10Gbps full-duplex */
> > > > > +#define MDIO_LPA_USXGMII_2500		0x0800	/* 2500Mbps */
> > > > > +#define MDIO_LPA_USXGMII_2500HALF	0x0800	/* 2500Mbps half-duplex */
> > > > > +#define MDIO_LPA_USXGMII_2500FULL	0x1800	/* 2500Mbps full-duplex */
> > > > > +#define MDIO_LPA_USXGMII_5000		0x0a00	/* 5000Mbps */
> > > > > +#define MDIO_LPA_USXGMII_5000HALF	0x0a00	/* 5000Mbps half-duplex */
> > > > > +#define MDIO_LPA_USXGMII_5000FULL	0x1a00	/* 5000Mbps full-duplex */
> > > > > +#define MDIO_LPA_USXGMII_LINK		0x8000	/* PHY link with
> > > > > copper-side partner */
> > > >
> > > > btw, the only thing which is missing from this is bit 0.
> > >
> > > TBH, I didn't know how to name it. Any suggestions?
> >
> > NXP calls it ABIL0, in xilinx docs its called USXGMII [1]. In the 
> > USXGMII
> > spec, its "set to 1 (0 is SGMII)" which I don't understand because its
> > also 1 for SGMII, right? At least as described in the 
> > tx_configReg[15:0] in
> > the SGMII spec.
> >
> > #define MDIO_USXGMII_USXGMII 0x0001 ?
> >
> > -michael
> >
> > [1] 
> > https://www.xilinx.com/support/documentation/ip_documentation/usxgmii/v1_0/pg251-usxgmii.pdf
>
> The explanation in the spec is quite cryptic, I've taken that to mean
> "corresponds to bit 0 in SGMII". Hence the reason why, in the code I've
> introduced in Felix, this is simply used as ADVERTISE_SGMII. I have no
> problem in creating an alias to ADVERTISE_SGMII named
> MDIO_LPA_USXGMII_SGMII. That being said, I don't see, right now, a
> practical situation where you might want to parse bit 0 from LPA, it's
> just like in Forrest Gump: "life is like a box of chocolates, you never
> know what you're gonna get". Adding it now to the UAPI might very well
> be a non-issue.

what about
  #define MDIO_USXGMII_ADVERTISE 0x0001 /* must always be set */

Russell, do you agree? Then I'd send a new version with your 
"Reviewed-by:".

-michael