linux-kernel - Re: [PATCH] net: phy: mscc: enable MAC SerDes autonegotiation

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 7 Feb 2022 23:19:48 +0530
From:   Raag Jadav <raagjadav@...il.com>
To:     Andrew Lunn <andrew@...n.ch>
Cc:     Heiner Kallweit <hkallweit1@...il.com>,
        Russell King <linux@...linux.org.uk>,
        "David S. Miller" <davem@...emloft.net>,
        Jakub Kicinski <kuba@...nel.org>,
        Steen Hegelund <steen.hegelund@...rochip.com>,
        Bjarni Jonasson <bjarni.jonasson@...rochip.com>,
        netdev@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] net: phy: mscc: enable MAC SerDes autonegotiation

On Sun, Feb 06, 2022 at 07:01:41PM +0100, Andrew Lunn wrote:
> On Sun, Feb 06, 2022 at 10:42:34PM +0530, Raag Jadav wrote:
> > On Sat, Feb 05, 2022 at 03:57:49PM +0100, Andrew Lunn wrote:
> > > On Sat, Feb 05, 2022 at 12:14:52PM +0530, Raag Jadav wrote:
> > > > Enable MAC SerDes autonegotiation to distinguish between
> > > > 1000BASE-X, SGMII and QSGMII MAC.
> > > 
> > > How does autoneg help you here? It just tells you about duplex, pause
> > > etc. It does not indicate 1000BaseX, SGMII etc. The PHY should be
> > > using whatever mode it was passed in phydev->interface, which the MAC
> > > sets when it calls the connection function. If the PHY dynamically
> > > changes its host side mode as a result of what that line side is
> > > doing, it should also change phydev->interface. However, as far as i
> > > can see, the mscc does not do this.
> > >
> > 
> > Once the PHY auto-negotiates parameters such as speed and duplex mode
> > with its link partner over the copper link as per IEEE 802.3 Clause 27,
> > the link partner’s capabilities are then transferred by PHY to MAC
> > over 1000BASE-X or SGMII link using the auto-negotiation functionality
> > defined in IEEE 802.3z Clause 37.
> 
> None of this allows you to distinguish between 1000BASE-X, SGMII and
> QSGMII, which is what the commit message says.
> 

I agree, the current commit message is misleading.

> It does allow you to get duplex, pause, and maybe speed via in band
> signalling. But you should also be getting the same information out of
> band, via the phylib callback.
> 
> There are some MACs which don't seem to work correctly without the in
> band signalling, so maybe that is your problem? Please could you give
> more background about your problem, what MAC and PHY combination are
> you using, what problem you are seeing, etc.
> 

MAC implementation[1] in a lot of NXP SoCs comes with in-band aneg enabled
by default, and it does expect Clause 37 auto-negotiation to complete
between MAC and PHY before the actual data transfer happens.

[1] https://community.nxp.com/pwmxy87654/attachments/pwmxy87654/t-series/3241/1/AN3869(1).pdf

I faced such issue while integrating VSC85xx PHY
with one of the recent NXP SoC having similar MAC implementation.
Not sure if this is a problem on MAC side or PHY side,
But having Clause 37 support should help in most cases I believe.

Cheers,
Raag

>     Andrew
>