Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:7 as permitted sender) client-ip=2620:137:e000::3:7;
Date:   Tue, 3 Oct 2023 20:26:01 +0200
From:   Andrew Lunn <andrew@lunn.ch>
To:     Jakub Kicinski <kuba@kernel.org>
Cc:     Maxime Chevallier <maxime.chevallier@bootlin.com>,
        davem@davemloft.net, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org, Eric Dumazet <edumazet@google.com>,
        Paolo Abeni <pabeni@redhat.com>,
        Florian Fainelli <f.fainelli@gmail.com>,
        Heiner Kallweit <hkallweit1@gmail.com>,
        Russell King <linux@armlinux.org.uk>,
        Vladimir Oltean <vladimir.oltean@nxp.com>,
        Oleksij Rempel <linux@rempel-privat.de>,
        =?iso-8859-1?Q?Nicol=F2?= Veronese <nicveronese@gmail.com>,
        thomas.petazzoni@bootlin.com,
        Christophe Leroy <christophe.leroy@csgroup.eu>
Subject: Re: [RFC PATCH net-next 6/7] net: ethtool: add a netlink command to
 get PHY information
Message-ID: <ffc6ff4a-d1af-4643-a538-fd13e6be9e06@lunn.ch>
References: <20230907092407.647139-1-maxime.chevallier@bootlin.com>
 <20230907092407.647139-7-maxime.chevallier@bootlin.com>
 <20230908084606.5707e1b1@kernel.org>
 <20230914113613.54fe125c@fedora>
 <20231003065535.34a3a4e0@kernel.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20231003065535.34a3a4e0@kernel.org>
Precedence: bulk

On Tue, Oct 03, 2023 at 06:55:35AM -0700, Jakub Kicinski wrote:
> On Thu, 14 Sep 2023 11:36:13 +0200 Maxime Chevallier wrote:
> > I'm currently implementing this, and I was wondering if it could be
> > worth it to include a pointer to struct phy_device directly in
> > ethnl_req_info.
> > 
> > This would share the logic for all netlink commands that target a
> > phy_device :
> > 
> >  - plca
> >  - pse-pd
> >  - cabletest
> >  - other future commands
> > 
> > Do you see this as acceptable ? we would grab the phy_device that
> > matches the passed phy_index in the request, and if none is specified,
> > we default to dev->phydev.
> 
> You may need to be careful with that. It could work in practice but 
> the req_info is parsed without holding any locks, IIRC. And there
> may also be some interplay between PHY state and ethnl_ops_begin().

We also need to ensure it is totally optional. There are MAC drivers
which reinvent the wheel in firmware. They can have multiple PHYs, or
PHY and SFP in parallel etc. All the typologies which you are
considering for phylink. Ideally we want the uAPI to work for
everybody, not just phylink. Its not our problem how said firmware
actually works, and what additional wheels they need to re-implement,
but we should try not to block them.

    Andrew