Re: [PATCH net-next v1] net: dsa: fix bridge support for drivers without port_bridge_flags callback

From: Vladimir Oltean <olteanv@gmail.com>
To: Oleksij Rempel <o.rempel@pengutronix.de>
Cc: Andrew Lunn <andrew@lunn.ch>,
	Florian Fainelli <f.fainelli@gmail.com>,
	Vivien Didelot <vivien.didelot@gmail.com>,
	"David S. Miller" <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>,
	kernel@pengutronix.de, netdev@vger.kernel.org,
	linux-kernel@vger.kernel.org,
	Russell King <linux@armlinux.org.uk>
Subject: Re: [PATCH net-next v1] net: dsa: fix bridge support for drivers without port_bridge_flags callback
Date: Wed, 21 Apr 2021 19:40:09 +0300	[thread overview]
Message-ID: <20210421164009.jylcxlab2evi2gcr@skbuf> (raw)
In-Reply-To: <20210421130540.12522-1-o.rempel@pengutronix.de>

On Wed, Apr 21, 2021 at 03:05:40PM +0200, Oleksij Rempel wrote:
> Starting with patch:
> a8b659e7ff75 ("net: dsa: act as passthrough for bridge port flags")
> 
> drivers without "port_bridge_flags" callback will fail to join the bridge.
> Looking at the code, -EOPNOTSUPP seems to be the proper return value,
> which makes at least microchip and atheros switches work again.
> 
> Fixes: a8b659e7ff75 ("net: dsa: act as passthrough for bridge port flags")
> Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
> ---
>  net/dsa/port.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/net/dsa/port.c b/net/dsa/port.c
> index 01e30264b25b..6379d66a6bb3 100644
> --- a/net/dsa/port.c
> +++ b/net/dsa/port.c
> @@ -550,7 +550,7 @@ int dsa_port_bridge_flags(const struct dsa_port *dp,
>  	struct dsa_switch *ds = dp->ds;
>  
>  	if (!ds->ops->port_bridge_flags)
> -		return -EINVAL;
> +		return -EOPNOTSUPP;
>  
>  	return ds->ops->port_bridge_flags(ds, dp->index, flags, extack);
>  }
> -- 
> 2.29.2
> 

The Fixes: tag should be:

Fixes: 5961d6a12c13 ("net: dsa: inherit the actual bridge port flags at join time")

What we return to the bridge is -EINVAL, via dsa_port_pre_bridge_flags()
(it is a two-step calling convention thing). But dsa_port_bridge_flags()
should never return that -EINVAL to the bridge, because the bridge
should just stop if the "pre" call returned an error.

So the -EINVAL return value from dsa_port_bridge_flags() is just for
callers who don't bother to call "pre".

To be honest I don't know why I wrote dsa_port_inherit_brport_flags this
way. It might be better to just do:

-----------------------------[cut here]-----------------------------

diff --git a/net/dsa/port.c b/net/dsa/port.c
index 01e30264b25b..d5e227a77fbc 100644
--- a/net/dsa/port.c
+++ b/net/dsa/port.c
@@ -138,8 +138,12 @@ static int dsa_port_inherit_brport_flags(struct dsa_port *dp,
 		if (br_port_flag_is_set(brport_dev, BIT(flag)))
 			flags.val = BIT(flag);
 
+		err = dsa_port_pre_bridge_flags(dp, flags, extack);
+		if (err == -EINVAL)
+			continue;
+
 		err = dsa_port_bridge_flags(dp, flags, extack);
-		if (err && err != -EOPNOTSUPP)
+		if (err)
 			return err;
 	}
 
@@ -159,8 +163,12 @@ static void dsa_port_clear_brport_flags(struct dsa_port *dp)
 		flags.mask = BIT(flag);
 		flags.val = val & BIT(flag);
 
+		err = dsa_port_pre_bridge_flags(dp, flags, NULL);
+		if (err == -EINVAL)
+			continue;
+
 		err = dsa_port_bridge_flags(dp, flags, NULL);
-		if (err && err != -EOPNOTSUPP)
+		if (err)
 			dev_err(dp->ds->dev,
 				"failed to clear bridge port flag %lu: %pe\n",
 				flags.val, ERR_PTR(err));
-----------------------------[cut here]-----------------------------
