Re: [net-next RFC PATCH v2 3/3] net: phy: Add Airoha AN8855 Internal Switch Gigabit PHY

[Christian Marangi <ansuelsmth@gmail.com>]

On Wed, Oct 23, 2024 at 07:00:22PM +0200, Andrew Lunn wrote:
> > +static int an8855_config_init(struct phy_device *phydev)
> > +{
> > +	struct air_an8855_priv *priv = phydev->priv;
> > +	int ret;
> > +
> > +	/* Enable HW auto downshift */
> > +	ret = phy_write(phydev, AN8855_PHY_PAGE_CTRL, AN8855_PHY_EXT_PAGE);
> > +	if (ret)
> > +		return ret;
> > +	ret = phy_set_bits(phydev, AN8855_PHY_EXT_REG_14,
> > +			   AN8855_PHY_EN_DOWN_SHFIT);
> > +	if (ret)
> > +		return ret;
> > +	ret = phy_write(phydev, AN8855_PHY_PAGE_CTRL, AN8855_PHY_NORMAL_PAGE);
> > +	if (ret)
> > +		return ret;
> 
> There are locking issues here, which is why we have the helpers
> phy_select_page() and phy_restore_page(). The air_en8811h.c gets this
> right.

Ugh didn't think about it... The switch address is shared with the PHY
so yes this is a problem.

Consider that this page thing comes from my speculation... Not really
use if 1f select page... 
From what I observed
0x0 PHY page
0x1 this strange EXT
0x4 acess switch register (every PHY can access the switch)

> 
> Is there anything in common with the en8811h? Does it also support
> downshift? Can its LED code be used here?
> 

For some reason part of the LED are controlled by the switch and some
are by the PHY. I still have to investigate that (not giving priority to
it... just on my todo)

For downshift as you notice it's a single bit with no count...
From their comments in the original driver it's said "Enable HW
autodownshift"

Trying to reach them but currently it's all very obscure.

-- 
	Ansuel