Re: [PATCH net-next 07/10] net: ethernet: freescale: xgmac: Separate C22 and C45 transactions for xgmac

[Vladimir Oltean <olteanv@gmail.com>]

On Tue, May 10, 2022 at 09:01:58PM +0200, Andrew Lunn wrote:
> On Tue, May 10, 2022 at 09:28:18PM +0300, Vladimir Oltean wrote:
> > On Sun, May 08, 2022 at 05:30:46PM +0200, Andrew Lunn wrote:
> > > The xgmac MDIO bus driver can perform both C22 and C45 transfers.
> > > Create separate functions for each and register the C45 versions using
> > > the new API calls where appropriate.
> > > 
> > > Signed-off-by: Andrew Lunn <andrew@lunn.ch>
> > > ---
> > >  drivers/net/ethernet/freescale/xgmac_mdio.c | 154 +++++++++++++++-----
> > >  1 file changed, 117 insertions(+), 37 deletions(-)
> > > 
> > > diff --git a/drivers/net/ethernet/freescale/xgmac_mdio.c b/drivers/net/ethernet/freescale/xgmac_mdio.c
> > > index ec90da1de030..ddfe6bf1f231 100644
> > > --- a/drivers/net/ethernet/freescale/xgmac_mdio.c
> > > +++ b/drivers/net/ethernet/freescale/xgmac_mdio.c
> > > @@ -128,30 +128,59 @@ static int xgmac_wait_until_done(struct device *dev,
> > >  	return 0;
> > >  }
> > >  
> > > -/*
> > > - * Write value to the PHY for this device to the register at regnum,waiting
> > > +/* Write value to the PHY for this device to the register at regnum,waiting
> > >   * until the write is done before it returns.  All PHY configuration has to be
> > >   * done through the TSEC1 MIIM regs.
> > >   */
> > > -static int xgmac_mdio_write(struct mii_bus *bus, int phy_id, int regnum, u16 value)
> > > +static int xgmac_mdio_write_c22(struct mii_bus *bus, int phy_id, int regnum,
> > > +				u16 value)
> > >  {
> > >  	struct mdio_fsl_priv *priv = (struct mdio_fsl_priv *)bus->priv;
> > >  	struct tgec_mdio_controller __iomem *regs = priv->mdio_base;
> > > -	uint16_t dev_addr;
> > > +	bool endian = priv->is_little_endian;
> > >  	u32 mdio_ctl, mdio_stat;
> > > +	u16 dev_addr;
> > >  	int ret;
> > > +
> > > +	mdio_stat = xgmac_read32(&regs->mdio_stat, endian);
> > > +	dev_addr = regnum & 0x1f;
> > 
> > Please move this either to the variable declaration, or near the mdio_ctl write,
> > or just integrate it into the macro argument.
> 
> There are masks like this in some drivers, others don't. Since for the
> majority of the MDIO bus drivers i don't have the hardware i was
> trying to keep my changes to a minimum, so i'm less likely to break
> it.
> 
> Once we have all the bus drivers converted, we can validate all the
> requests in the core to guarantee no users are passing invalid values
> to the drivers. And then all these masks can be removed.

Sure, I was going to revisit this comment, keep the masking with 0x1f,
I remembered in the meanwhile that it's supposed to represent
MII_MMD_CTRL_DEVAD_MASK, and that there is other stuff potentially
encoded in the devad, like post-increment stuff.

> > 
> > > +	mdio_stat &= ~MDIO_STAT_ENC;
> > > +
> > 
> > You can remove this empty line during read-modify-write patterns.
> 
> Sure, but just an FYI: the old code probably did it that way. My aim
> is to split C22 from C45, not re-write/clean up every driver. I have
> over 40 patches in total, without doing cleanups.

You are modifying this part of the code anyway.