[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20220412062957.353260366@linuxfoundation.org>
Date: Tue, 12 Apr 2022 08:30:26 +0200
From: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
To: linux-kernel@...r.kernel.org
Cc: Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
stable@...r.kernel.org,
Luiz Angelo Daros de Luca <luizluca@...il.com>,
Vladimir Oltean <vladimir.oltean@....com>,
"David S. Miller" <davem@...emloft.net>,
Sasha Levin <sashal@...nel.org>
Subject: [PATCH 5.17 208/343] Revert "net: dsa: stop updating master MTU from master.c"
From: Vladimir Oltean <vladimir.oltean@....com>
[ Upstream commit 066dfc4290406b1b0b014ae3267d4266a344efd1 ]
This reverts commit a1ff94c2973c43bc1e2677ac63ebb15b1d1ff846.
Switch drivers that don't implement ->port_change_mtu() will cause the
DSA master to remain with an MTU of 1500, since we've deleted the other
code path. In turn, this causes a regression for those systems, where
MTU-sized traffic can no longer be terminated.
Revert the change taking into account the fact that rtnl_lock() is now
taken top-level from the callers of dsa_master_setup() and
dsa_master_teardown(). Also add a comment in order for it to be
absolutely clear why it is still needed.
Fixes: a1ff94c2973c ("net: dsa: stop updating master MTU from master.c")
Reported-by: Luiz Angelo Daros de Luca <luizluca@...il.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@....com>
Tested-by: Luiz Angelo Daros de Luca <luizluca@...il.com>
Signed-off-by: David S. Miller <davem@...emloft.net>
Signed-off-by: Sasha Levin <sashal@...nel.org>
---
net/dsa/master.c | 25 ++++++++++++++++++++++++-
1 file changed, 24 insertions(+), 1 deletion(-)
diff --git a/net/dsa/master.c b/net/dsa/master.c
index 880f910b23a9..10b51ffbb6f4 100644
--- a/net/dsa/master.c
+++ b/net/dsa/master.c
@@ -337,11 +337,24 @@ static const struct attribute_group dsa_group = {
static struct lock_class_key dsa_master_addr_list_lock_key;
+static void dsa_master_reset_mtu(struct net_device *dev)
+{
+ int err;
+
+ err = dev_set_mtu(dev, ETH_DATA_LEN);
+ if (err)
+ netdev_dbg(dev,
+ "Unable to reset MTU to exclude DSA overheads\n");
+}
+
int dsa_master_setup(struct net_device *dev, struct dsa_port *cpu_dp)
{
+ const struct dsa_device_ops *tag_ops = cpu_dp->tag_ops;
struct dsa_switch *ds = cpu_dp->ds;
struct device_link *consumer_link;
- int ret;
+ int mtu, ret;
+
+ mtu = ETH_DATA_LEN + dsa_tag_protocol_overhead(tag_ops);
/* The DSA master must use SET_NETDEV_DEV for this to work. */
consumer_link = device_link_add(ds->dev, dev->dev.parent,
@@ -351,6 +364,15 @@ int dsa_master_setup(struct net_device *dev, struct dsa_port *cpu_dp)
"Failed to create a device link to DSA switch %s\n",
dev_name(ds->dev));
+ /* The switch driver may not implement ->port_change_mtu(), case in
+ * which dsa_slave_change_mtu() will not update the master MTU either,
+ * so we need to do that here.
+ */
+ ret = dev_set_mtu(dev, mtu);
+ if (ret)
+ netdev_warn(dev, "error %d setting MTU to %d to include DSA overhead\n",
+ ret, mtu);
+
/* If we use a tagging format that doesn't have an ethertype
* field, make sure that all packets from this point on get
* sent to the tag format's receive function.
@@ -388,6 +410,7 @@ void dsa_master_teardown(struct net_device *dev)
sysfs_remove_group(&dev->dev.kobj, &dsa_group);
dsa_netdev_ops_set(dev, NULL);
dsa_master_ethtool_teardown(dev);
+ dsa_master_reset_mtu(dev);
dsa_master_set_promiscuity(dev, -1);
dev->dsa_ptr = NULL;
--
2.35.1
Powered by blists - more mailing lists