Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.34 as permitted sender) client-ip=23.128.96.34;
Date:   Mon, 23 Oct 2023 15:19:24 +0300
From:   Vladimir Oltean <olteanv@gmail.com>
To:     Ante Knezic <ante.knezic@helmholz.de>
Cc:     UNGLinuxDriver@microchip.com, andrew@lunn.ch, conor+dt@kernel.org,
        davem@davemloft.net, devicetree@vger.kernel.org,
        edumazet@google.com, f.fainelli@gmail.com,
        krzysztof.kozlowski+dt@linaro.org, kuba@kernel.org,
        linux-kernel@vger.kernel.org, marex@denx.de,
        netdev@vger.kernel.org, o.rempel@pengutronix.de, pabeni@redhat.com,
        robh+dt@kernel.org, woojung.huh@microchip.com
Subject: Re: [PATCH net-next v4 2/2] net:dsa:microchip: add property to select
Message-ID: <20231023121924.udseyuy7t77dscwl@skbuf>
References: <20231020143759.eknrcfbztrc543mm@skbuf>
 <20231023072700.17060-1-ante.knezic@helmholz.de>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20231023072700.17060-1-ante.knezic@helmholz.de>
Precedence: bulk

On Mon, Oct 23, 2023 at 09:27:00AM +0200, Ante Knezic wrote:
> As far as I am aware only the KSZ8863 and KSZ8873 have this property available,
> but the biggger issue might be in scaling this to port property as the register
> "Forward Invalid VID Frame and Host Mode" where the setting is applied is
> located under "Advanced Control Registers" section which is actually global at
> least looking from the switch point of view. Usually port properties are more
> applicable when registers in question are located under "Port Registers" section.
> This is somewhat similar to for example enabling the tail tag mode which is 
> again used only by the port 3 interface and is control from "Global Control 1"
> register.
> With this in mind - if you still believe we should move this to port dt 
> property, then should we forbid setting the property for any other port other 
> than port 3, and can/should this be enforced by the dt schema?

I have no doubt that RMII settings are port settings. Scaling up the implementation
to multiple ports on other switches doesn't mean that the DT binding shouldn't be
per port.

Anyway, the per-port access to a global switch setting is indeed a common theme
with the old Micrel switches. I once tried to introduce the concept of "wacky"
regmap regfields for that:
https://patchwork.kernel.org/project/netdevbpf/patch/20230316161250.3286055-3-vladimir.oltean@nxp.com/

but I don't have hardware to test and nobody who does picked up upon the regfield
idea, it seems.