Re: [PATCH RFC russell-king 0/4] Support for RollBall 10G copper SFP modules

["Marek Behún" <marek.behun@nic.cz>]

On Tue, 18 Aug 2020 17:34:15 +0100
Russell King - ARM Linux admin <linux@armlinux.org.uk> wrote:

> On Tue, Aug 18, 2020 at 05:47:24PM +0200, Marek Behún wrote:
> > On Tue, 18 Aug 2020 16:36:49 +0100
> > Russell King - ARM Linux admin <linux@armlinux.org.uk> wrote:
> >   
> > > On Tue, Aug 18, 2020 at 05:30:55PM +0200, Marek Behún wrote:  
> > > > On Tue, 18 Aug 2020 16:08:35 +0100
> > > > Russell King - ARM Linux admin <linux@armlinux.org.uk> wrote:
> > > >     
> > > > > > Otherwise it looks nice. I will test this. On what branch
> > > > > > does this apply?      
> > > > > 
> > > > > My unpublished tip of the universe development.  Here's a
> > > > > version on the clearfog branch:    
> > > > 
> > > > Russell, it seems you do not have commit
> > > > 
> > > > e11703330a5d ("net: phy: marvell10g: support XFI rate matching
> > > > mode")
> > > > 
> > > > in that branch.    
> > > 
> > > I don't.  That commit is in 5.9-rc1, I'm based on 5.8 at the
> > > moment. 
> > 
> > I am reworking it so that it applies on top of master together with
> > your other commits from clearfog branch.  
> 
> If it works for you, I'll rework my branches to adopt this approach.
> 

Russell,

if you have time please rebase your work on top of Linus' master or
net-next.

Btw one of your patches in clearfog branch breaks mvneta
(although there are other patches which fix this break).
Check this out:
  patch
      net: mvneta: move 1ms clock control into mac_prepare/mac_finish
  removes definition of variables new_clk and gmac_clk in
  mvneta_mac_config, but does not remove one instance of usage, which
  can be seen in subsequent patch
      net: mvneta: convert to phylink pcs operations
  this is fixed by patch
      net: mvneta: split out GMAC
  where all of this is moved to mvgmac.c file.
I found out because I did not apply the last patch for some reason
(maybe it didn't apply on Linus' master or something, I don't remember),
and it failed to compile.

http://git.arm.linux.org.uk/cgit/linux-arm.git/commit/?h=clearfog&id=ca096c11e6798b1f4da8466ab0c3bf42b6e9fb81
http://git.arm.linux.org.uk/cgit/linux-arm.git/commit/?h=clearfog&id=1e345c538cc73d0b0a63c01a13e69a865905b7ae
http://git.arm.linux.org.uk/cgit/linux-arm.git/commit/?h=clearfog&id=dc33cc833537d06b198c9226d4a230cfc33569a6

I have just sent patches that add 88E6393X switch to mv88e6xxx driver,
since I am testing these SFPs on a Marvell Customer Reference Board
containing this switch.

The board also contains the other PHY supported by the marvell10g PHY
driver, 88E2110, so I will test my changes also on this PHY.

I am waiting for newer documentation for 88E3110, since the one I have
is outdated and does not contain descriptions on how to resolve USXGMII
auto-negotiation, which I would also like to add.

Marek