Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 16 Nov 2022 12:24:06 +0200
From:   Vladimir Oltean <olteanv@gmail.com>
To:     netdev@kapio-technology.com
Cc:     Ido Schimmel <idosch@idosch.org>, davem@davemloft.net,
        kuba@kernel.org, netdev@vger.kernel.org,
        Andrew Lunn <andrew@lunn.ch>,
        Florian Fainelli <f.fainelli@gmail.com>,
        Eric Dumazet <edumazet@google.com>,
        Paolo Abeni <pabeni@redhat.com>,
        open list <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v8 net-next 0/2] mv88e6xxx: Add MAB offload support
Message-ID: <20221116102406.gg6h7gvkx55f2ojj@skbuf>
References: <20221115102833.ahwnahrqstcs2eug@skbuf>
 <7c02d4f14e59a6e26431c086a9bb9643@kapio-technology.com>
 <20221115111034.z5bggxqhdf7kbw64@skbuf>
 <0cd30d4517d548f35042a535fd994831@kapio-technology.com>
 <20221115122237.jfa5aqv6hauqid6l@skbuf>
 <fb1707b55bd8629770e77969affaa2f9@kapio-technology.com>
 <20221115145650.gs7crhkidbq5ko6v@skbuf>
 <f229503b98d772c936f1fc8ca826a14f@kapio-technology.com>
 <20221115161846.2st2kjxylfvlncib@skbuf>
 <e05f69915a2522fc1e9854194afcc87b@kapio-technology.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <e05f69915a2522fc1e9854194afcc87b@kapio-technology.com>
Precedence: bulk

On Tue, Nov 15, 2022 at 07:40:02PM +0100, netdev@kapio-technology.com wrote:
> So, I will not present you with a graph as it is a tedious process (probably
> it is some descending gaussian curve wrt timeout occurring).
> 
> But 100ms fails, 125 I had 1 port fail, at 140, 150  and 180 I saw timeouts
> resulting in fdb add fails, like (and occasional port fail):
> 
> mv88e6085 1002b000.ethernet-1:04: Timeout while waiting for switch
> mv88e6085 1002b000.ethernet-1:04: port 0 failed to add be:7c:96:06:9f:09 vid
> 1 to fdb: -110
> 
> At around 200 ms it looks like it is getting stable (like 5 runs, no
> problems).
> 
> So with the gaussian curve tail whipping ones behind (risque of failure) it
> might need to be like 300 ms in my case... :-)

Pick a value that is high enough to be reliable and submit a patch to
"net" where you present the evidence for it (top-level MDIO controller,
SoC, switch, kernel). I don't believe there's much to read into. A large
timeout shouldn't have a negative effect on the MDIO performance,
because it just determines how long it takes until the kernel declares
it dead, rather than how long it takes for transactions to actually take
place.