Re: [PATCH] net: phy: mscc: enable MAC SerDes autonegotiation

["Russell King (Oracle)" <linux@armlinux.org.uk>]

On Sat, Feb 26, 2022 at 12:53:27PM +0530, Raag Jadav wrote:
> On Thu, Feb 24, 2022 at 10:48:57AM +0000, Russell King (Oracle) wrote:
> > Sorry for the late comment on this patch.
> > 
> > On Sat, Feb 05, 2022 at 12:14:52PM +0530, Raag Jadav wrote:
> > > +static int vsc85xx_config_inband_aneg(struct phy_device *phydev, bool enabled)
> > > +{
> > > +	int rc;
> > > +	u16 reg_val = 0;
> > > +
> > > +	if (enabled)
> > > +		reg_val = MSCC_PHY_SERDES_ANEG;
> > > +
> > > +	mutex_lock(&phydev->lock);
> > > +
> > > +	rc = phy_modify_paged(phydev, MSCC_PHY_PAGE_EXTENDED_3,
> > > +			      MSCC_PHY_SERDES_PCS_CTRL, MSCC_PHY_SERDES_ANEG,
> > > +			      reg_val);
> > > +
> > > +	mutex_unlock(&phydev->lock);
> > 
> > What is the reason for the locking here?
> > 
> > phy_modify_paged() itself is safe due to the MDIO bus lock, so you
> > shouldn't need locking around it.
> > 
> 
> True.
> 
> My initial thought was to have serialized access at PHY level,
> as we have multiple ports to work with.
> But I guess MDIO bus lock could do the job as well.

The MDIO bus lock is the only lock that will guarantee that no other
users can nip onto the bus and possibly access your PHY in the middle
of an operation that requires more than one access to complete. Adding
local locking at PHY driver level does not give you those guarantees.
This is exactly why phy_modify() etc was added - because phy_read()..
phy_write() does not give that guarantee.

As an example of something that could interfere - the userspace MII
ioctls.

> I've gone through Vladimir's patches and they look more promising
> than this approach.
> Let me know if I could be of any help.

I haven't seen them - so up to you.

Thanks.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!