Re: [PATCH net] net: phy: micrel: populate .soft_reset for KSZ9131

[Maxime Chevallier]

On Fri, 5 Jan 2024 15:36:29 +0100
Andrew Lunn <andrew@xxxxxxx> wrote:

> On Fri, Jan 05, 2024 at 09:43:22AM +0000, Russell King (Oracle) wrote:
> > On Fri, Jan 05, 2024 at 10:52:42AM +0200, Claudiu wrote:  
> > > The order of PHY-related operations in ravb_open() is as follows:
> > > ravb_open() ->
> > >   ravb_phy_start() ->
> > >     ravb_phy_init() ->
> > >       of_phy_connect() ->
> > >         phy_connect_direct() ->
> > > 	  phy_attach_direct() ->
> > > 	    phy_init_hw() ->
> > > 	      phydev->drv->soft_reset()
> > > 	      phydev->drv->config_init()
> > > 	      phydev->drv->config_intr()
> > > 	    phy_resume()
> > > 	      kszphy_resume()
> > > 
> > > The order of PHY-related operations in ravb_close is as follows:
> > > ravb_close() ->
> > >   phy_stop() ->
> > >     phy_suspend() ->
> > >       kszphy_suspend() ->
> > >         genphy_suspend()
> > > 	  // set BMCR_PDOWN bit in MII_BMCR  
> > 
> > Andrew,
> > 
> > This looks wrong to me - shouldn't we be resuming the PHY before
> > attempting to configure it?  
> 
> Hummm. The opposite of phy_stop() is phy_start(). So it would be the
> logical order to perform the resume as the first action of
> phy_start(), not phy_attach_direct().
> 
> In phy_connect_direct(), we don't need the PHY to be operational
> yet. That happens with phy_start().
> 
> The standard says:
> 
>   22.2.4.1.5 Power down
> 
>   The PHY may be placed in a low-power consumption state by setting
>   bit 0.11 to a logic one. Clearing bit 0.11 to zero allows normal
>   operation. The specific behavior of a PHY in the power-down state is
>   implementation specific. While in the power-down state, the PHY
>   shall respond to management transactions.
> 
> So i would say this PHY is broken, its not responding to all
> management transactions. So in that respect, Claudiu fix is correct.
> 
> But i also somewhat agree with you, this looks wrong, but in a
> different way to how you see it. However, moving the phy_resume() to
> phy_start() seems a bit risky. So i'm not sure we should actually do
> that.

Looking at other PHYs similar to it like the 9031, the .soft_reset()
was added to fix some similar issues :

Issue :
https://lore.kernel.org/netdev/a63ca542-db96-40ed-201d-59c609f565ce@xxxxxxxxx/

Fix :
https://lore.kernel.org/netdev/6d3b1dce-7633-51a1-0556-97cd03304c2c@xxxxxxxxx/

We couldn't get a proper explanation back then. Could it be that they
suffer from the same problem, but that it was more clearly documented
for the 9131 ?

Maxime