Re: [PATCH v2 03/12] PCI: aardvark: Add PHY support

[Miquel Raynal]

Hi Marek,

Marek BehÃn <marek.behun@xxxxxx> wrote on Wed, 19 Dec 2018 16:28:35
+0100:

> Hi,
> 
> One thing for which I would like to be able to disable comphy is that
> each consumes about 100mW of power. On Turris Mox we configure the
> comphys to SGMII1, PCIe and USB3 modes. If there is no USB device
> plugged, the USB3 phy can be disabled, and save 100mW of power. If the
> PCIe extension module is not present, the PCIe can too be disabled, and
> if there is no switch nor SFP module present, so can SGMII1.

Indeed not all PHY types implement ->power_off() (see in ATF code).
Better ask Marvell directly for that.

> 
> The other reason is this: if the SGMII phy is set to 1G mode, and then
> powered on second time in 2.5G mode, will it work? I would like to

This should work, yes.

> patch mvneta driver to power on/off the comphy, if the device node is
> present in device tree. But then the system can request such a change
> (SGMII to 2500BASE-X or back).
> 
> Marek
> 
> On Tue, 18 Dec 2018 14:41:30 +0100
> Miquel Raynal <miquel.raynal@xxxxxxxxxxx> wrote:
> 
> > Hi Marek,
> > 
> > Marek Behun <marek.behun@xxxxxx> wrote on Tue, 18 Dec 2018 14:09:20
> > +0100:
> >   
> > > > [2]
> > > > https://github.com/ARM-software/arm-trusted-firmware/blob/master/drivers/marvell/comphy/phy-comphy-3700.c      
> > > 
> > > Yes, I used mainline atf (it did not work out of the box with 18.09
> > > atf-marvell of course). But there is no _power_off function for
> > > SGMII, nor a digital_reset function like in cp110 implementation.    
> > 
> > Indeed, but why would you need one? Just use the helpers from the core
> > and if there is no implementation, nothing should happen and the
> > helper should exit without error. Just call
> > phy_set_mode()/phy_power_on() an you should be good.
> > 
> > 
> > Thanks,
> > MiquÃl  
> 

For the record, I found out what was wrong in my code, toggling the
reset lines do not produce random effects anymore.


Thanks,
MiquÃl