Date: Fri, 23 Jun 2023 14:55:25 +0100
From: "Russell King (Oracle)" <linux@...linux.org.uk>
To: Andrew Lunn <andrew@...n.ch>, Revanth Kumar Uppala <ruppala@...dia.com>,
	Narayan Reddy <narayanr@...dia.com>
Cc: "hkallweit1@...il.com" <hkallweit1@...il.com>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
	"linux-tegra@...r.kernel.org" <linux-tegra@...r.kernel.org>
Subject: Re: [PATCH] net: phy: Enhance fixed PHY to support 10G and 5G

On Fri, Jun 23, 2023 at 03:29:13PM +0200, Andrew Lunn wrote:
> On Fri, Jun 23, 2023 at 12:28:49PM +0000, Revanth Kumar Uppala wrote:
> > 
> > 
> > > -----Original Message-----
> > > From: Andrew Lunn <andrew@...n.ch>
> > > Sent: Wednesday, June 21, 2023 11:00 PM
> > > To: Revanth Kumar Uppala <ruppala@...dia.com>
> > > Cc: hkallweit1@...il.com; netdev@...r.kernel.org; linux-
> > > tegra@...r.kernel.org; Narayan Reddy <narayanr@...dia.com>
> > > Subject: Re: [PATCH] net: phy: Enhance fixed PHY to support 10G and 5G
> > > 
> > > External email: Use caution opening links or attachments
> > > 
> > > 
> > > On Wed, Jun 21, 2023 at 10:28:53PM +0530, Revanth Kumar Uppala wrote:
> > > > Add 10G and 5G speed entries for fixed PHY framework.These are needed
> > > > for the platforms which doesn't have a PHY driver.
> > > >
> > > > Signed-off-by: Revanth Kumar Uppala <ruppala@...dia.com>
> > > > Signed-off-by: Narayan Reddy <narayanr@...dia.com>
> > > 
> > > This is the second time you have sent me this patch. You have failed to answer
> > > the questions i asked you the last time.....
> > Apologies for sending twice.
> > C45 registers are not defined in the kernel as of now. But, we need to display the speed as 5G/10G when the same is configured as fixed link in DT node.
> > It will be great if you can share any data for handling this.
> > As of now, with this change we have taken care of providing proper speed log in kernel when 5G/10G is added as fixed links in DT node.
> 
> This is architecturally wrong. As i said, swphy emulates a C22 PHY,
> and a C22 PHY does not support speeds greater than 1G. To make swphy
> really support 5G and 10G, you would need to add C45 support, and then
> extend the default genphy driver to look at the C45 registers as well.
> 
> However, that is all pointless. As i said, phylink fixed-link is not
> limited to 1G speeds. Given what i see in Cc: i assume this is for a
> tegre SoC? And that uses a Synopsys MAC? So you probably want to
> modify the dwc driver to use phylink.

Absolutely correct. I seem to recall having had this come up before,
and I think it was explained at the time, but I don't seem to find
anything in my "sent" mailboxes for the start of 2022 to present.

(To nvidia)

The classical swphy/fixed-phy offers a software emulated clause 22 PHY
so that phylib can be re-used to make a fixed link work without needing
special code paths in phylib nor in MAC drivers.

However, clause 22 PHYs do not support speeds in excess of 1G, so this
places a hard ceiling on the speed that can be supported with this
method. PHYLIB's clause 45 support is specific to vendor PHYs, and
the "generic" implementation only supports 10G speed. Emulating a
specific vendor PHY to achieve this old way of supporting fixed links
when we have a better way is really a waste of effort.

The "better way" is phylink, which makes fixed links work without
needing to resort to PHY emulation, and thus it can support any speed
that a MAC happens to support. This is the modern way.

We (the phylib and phylink maintainers) will not entertain extending
the old now legacy method using swphy/fixed-phy for fixed links to
include any faster speeds.

Thanks.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!

Powered by blists - more mailing lists