Message-ID: <rbavthifczzgwxmgtb4hbv6hnqb57timfzvbizscdtxz62ookg@bgrwergjyulb>
Date: Tue, 26 Sep 2023 18:27:45 +0300
From: Serge Semin <fancer.lancer@...il.com>
To: "Russell King (Oracle)" <linux@...linux.org.uk>
Cc: Raju Lakkaraju <Raju.Lakkaraju@...rochip.com>, netdev@...r.kernel.org, 
	davem@...emloft.net, Jose.Abreu@...opsys.com, hkallweit1@...il.com, 
	UNGLinuxDriver@...rochip.com
Subject: Re: [PATCH net-next] net: pcs: xpcs: Add 2500BASE-X case in get
 state for XPCS drivers

On Tue, Sep 26, 2023 at 03:46:23PM +0100, Russell King (Oracle) wrote:
> On Tue, Sep 26, 2023 at 03:09:47PM +0300, Serge Semin wrote:
> > Hi Russell
> > 
> > On Tue, Sep 26, 2023 at 12:45:47PM +0100, Russell King (Oracle) wrote:
> > > On Tue, Sep 26, 2023 at 02:39:21PM +0300, Serge Semin wrote:
> > > > Hi Raju
> > > > 
> > > > On Mon, Sep 25, 2023 at 01:21:42PM +0530, Raju Lakkaraju wrote:
> > > > > Add DW_2500BASEX case in xpcs_get_state( ) to update speed, duplex and pause
> > > > > Update the port mode and autonegotiation
> > > > > 
> > > > > Signed-off-by: Raju Lakkaraju <Raju.Lakkaraju@...rochip.com>
> > > > > ---
> > > > >  drivers/net/pcs/pcs-xpcs.c | 31 +++++++++++++++++++++++++++++++
> > > > >  drivers/net/pcs/pcs-xpcs.h |  4 ++++
> > > > >  2 files changed, 35 insertions(+)
> > > > > 
> > > > > diff --git a/drivers/net/pcs/pcs-xpcs.c b/drivers/net/pcs/pcs-xpcs.c
> > > > > index 4dbc21f604f2..4f89dcedf0fc 100644
> > > > > --- a/drivers/net/pcs/pcs-xpcs.c
> > > > > +++ b/drivers/net/pcs/pcs-xpcs.c
> > > > > @@ -1090,6 +1090,30 @@ static int xpcs_get_state_c37_1000basex(struct dw_xpcs *xpcs,
> > > > >  	return 0;
> > > > >  }
> > > > >  
> > > > > +static int xpcs_get_state_2500basex(struct dw_xpcs *xpcs,
> > > > > +				    struct phylink_link_state *state)
> > > > > +{
> > > > > +	int sts, lpa;
> > > > > +
> > > > > +	sts = xpcs_read(xpcs, MDIO_MMD_VEND2, DW_VR_MII_MMD_STS);
> > > > 
> > > > > +	lpa = xpcs_read(xpcs, MDIO_MMD_VEND2, DW_VR_MII_MMD_LP_BABL);
> > > > > +	if (sts < 0 || lpa < 0) {
> > > > > +		state->link = false;
> > > > > +		return sts;
> > > > > +	}
> > > > 
> > > > The HW manual says: "The host uses this page to know the link
> > > > partner's ability when the base page is received through Clause 37
> > > > auto-negotiation." Seeing xpcs_config_2500basex() disables
> > > > auto-negotiation and lpa value is unused anyway why do you even need
> > > > to read the LP_BABL register?
> > > 
> > 
> > > Since you have access to the hardware manual, what does it say about
> > > clause 37 auto-negotiation when operating in 2500base-X mode?
> > 
> > Here are the parts which mention 37 & SGMII AN in the 2.5G context:
> > 
> > 1. "Clause 37 (& SGMII) auto-negotiation is supported in 2.5G mode if
> >     the link partner is also operating in the equivalent 2.5G mode."
> > 
> > 2. "During the Clause 37/SGMII as the auto-negotiation link timer
> >     operates with a faster clock in the 2.5G mode, the timeout duration
> >     reduces by a factor of 2.5. To restore the standard specified timeout
> >     period, the respective registers must be re-programmed."
> > 
> > I guess the entire 2.5G-thing understanding could be generalized by
> > the next sentence from the HW manual: "The 2.5G mode of operation is
> > functionally the same as 1000BASE-X/KX mode, except that the
> > clock-rate is 2.5 times the original rate. In this mode, the
> > Serdes/PHY operates at a serial baud-rate of 3.125 Gbps and DWC_xpcs
> > data-path and the GMII interface to MAC operates at 312.5 MHz (instead
> > of 125 MHz)." Thus here is another info regarding AN in that context:
> > 
> > 3. "The DWC_xpcs operates either in 10/100/1000Mbps rate or
> > 25/250/2500Mbps rates respectively with SGMII auto-negotiation. The
> > DWC_xpcs cannot support switching or negotiation between 1G and 2.5G
> > rates using auto-negotiation."
> 
> Thanks for the clarification.
> 
> So this hardware, just like Marvell hardware, operates 2500BASE-X merely
> by up-clocking, and all the features that were available at 1000BASE-X
> are also available at 2500BASE-X, including the in-band signalling.
> 
> Therefore, I think rather than disabling AN outright, the
> PHYLINK_PCS_NEG_* mode passed in should be checked to determine whether
> inband should be used or not.

Just an additional note which might be relevant in the context of the
DW XPCS 1G/2.5G C37 AN. The C37 auto-negotiation feature will be
unavailable for 1000BASE-X and thus for 2500BASE-X if the IP-core is
synthesized with no support of one. It is determined by the CL37_AN
IP-core synthesize parameter state:

Enable SGMII Clause 37 | Description: Configures DWC_xpcs to support the
Auto-Negotiation       |              Clause 37 auto-negotiation
                       |
                       | Dependencies: This option is available in the
                       |   following configurations:
                       |   - SUPPORT_1G = Enabled
                       |   - MAIN_MODE = Backplane Ethernet PCS and
                       |     BACKPLANE_ETH_CONFIG = KX_Only or KX4_KX
                       |     or KR_KX or KR_KX4_KX mode
                       |
                       | Default Value: Enabled for configurations with
                       |   MAIN_MODE = 1000BASEX-Only PCS and Disabled
                       |   for all other configurations
                       |
                       | HDL Parameter Name: CL37_AN

So depending on the particular (vendor-specific) device configuration
the C37 AN might still unavailable even though the device supports the
1000BASE-X and 2500BASE-X interfaces.

-Serge(y)

> 
> -- 
> 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