On Mon, Jul 22, 2019 at 02:26:45PM +0000, Jose Abreu wrote:
> From: Andrew Lunn <andrew@lunn.ch>
> Date: Jul/22/2019, 15:19:43 (UTC+00:00)
> 
> > On Mon, Jul 22, 2019 at 01:58:20PM +0000, Jose Abreu wrote:
> > > From: Andrew Lunn <andrew@lunn.ch>
> > > Date: Jul/22/2019, 14:40:23 (UTC+00:00)
> > > 
> > > > Does this mean that all stmmac variants support 1G? There are none
> > > > which just support Fast Ethernet?
> > > 
> > > This glue logic drivers sometimes reflect a custom IP that's Synopsys 
> > > based but modified by customer, so I can't know before-hand what's the 
> > > supported max speed. There are some old versions that don't support 1G 
> > > but I expect that PHY driver limits this ...
> > 
> > If a Fast PHY is used, then yes, it would be limited. But sometimes a
> > 1G PHY is used because they are cheaper than a Fast PHY.
> >  
> > > > I'm also not sure the change fits the problem. Why did it not
> > > > negotiate 100FULL rather than 10Half? You are only moving the 1G
> > > > speeds around, so 100 speeds should of been advertised and selected.
> > > 
> > > Hmm, now that I'm looking at it closer I agree with you. Maybe link 
> > > partner or PHY doesn't support 100M ?
> > 
> > In the working case, ethtool shows the link partner supports 10, 100,
> > and 1G. So something odd is going on here.
> > 
> > You fix does seems reasonable, and it has been reported to fix the
> > issue, but it would be good to understand what is going on here.
> 
> Agreed!
> 
> Ondrej, can you please share dmesg log and ethtool output with the fixed 
> patch ?

See the attachment, or this link:

  https://megous.com/dl/tmp/dmesg-5.3-working

regards,
	Ondrej

> ---
> Thanks,
> Jose Miguel Abreu