Skip to contentSkip to conversation

bonding: try to enable SG features when adding a new, slave

PATCH NAK Ding Tianhong 2 participants 3 messages Jan 14, 2014 – Jan 17, 2014

Subsystems (3)
Networking Network drivers Bonding driver
lore
Share & tools
lore

Ding Tianhong

author
#0 lore raw

PATCH v2 2/2 bonding: try to enable SG features when adding a new, slave

The commit b0ce3508(bonding: allow TSO being set on bonding master) has make the TSO being set for bond dev, but in some situation, if the slave did not set the NETIF_F_SG features yet, the bond master will miss the TSO features in netdev_fix_features because the TSO is depended on SG.

If the slave hw support SG features, but not set yet, I will try to open it when enslave the dev, better for performance.

Signed-off-by Ding Tianhong <[email protected]>
---
 drivers/net/bonding/bond_main.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
drivers/net/bonding/bond_main.c +15 -0
index 4cfe14e..e1044ac 100644
@@ -1083,6 +1083,19 @@ static netdev_features_t bond_fix_features(struct net_device *dev,
 	return features;
 }
 
+/* If the slave support SG, but not being used yet,
+ * enable the SG features for this slave.
+ */
+static void bond_update_slave_features(struct net_device *dev)
+{
+	if (dev->hw_features & ~dev->features & NETIF_F_SG) {
+		pr_debug("The new dev %s support SG, but not being used yet, enable it\n",
+			 dev->name);
+		dev->wanted_features |= NETIF_F_SG;
+		netdev_update_features(dev);
+	}
+}
+
 #define BOND_VLAN_FEATURES	(NETIF_F_ALL_CSUM | NETIF_F_SG | \
 				 NETIF_F_FRAGLIST | NETIF_F_ALL_TSO | \
 				 NETIF_F_HIGHDMA | NETIF_F_LRO)

@@ -1587,6 +1600,8 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)
 	}
 
 	bond->slave_cnt++;
+	/* Try to enable the SG features for slave dev. */
+	bond_update_slave_features(slave_dev);
 	bond_compute_features(bond);
 	bond_set_carrier(bond);
Signature · 1 line
1.8.0
maintainer
#1 lore raw

From: Ding Tianhong <[email protected]>
Date: Tue, 14 Jan 2014 17:00:35 +0800

Quoted 10 lines

The commit b0ce3508(bonding: allow TSO being set on bonding master) has make the TSO being set for bond dev, but in some situation, if the slave did not set the NETIF_F_SG features yet, the bond master will miss the TSO features in netdev_fix_features because the TSO is depended on SG.

If the slave hw support SG features, but not set yet, I will try to open it when enslave the dev, better for performance.

Signed-off-by Ding Tianhong <[email protected]>

I really don't think we should force enable device features in slaves that perhaps the user intentionally disabled, or perhaps the driver has a reason to disable by default (lower performance, etc.)

I'm not applying this series, sorry.

author
#2 lore raw
Quoted 12 lines · David Miller

From: Ding Tianhong <[email protected]>
Date: Tue, 14 Jan 2014 17:00:35 +0800

Quoted 9 lines

The commit b0ce3508(bonding: allow TSO being set on bonding master) has make the TSO being set for bond dev, but in some situation, if the slave did not set the NETIF_F_SG features yet, the bond master will miss the TSO features in netdev_fix_features because the TSO is depended on SG.

If the slave hw support SG features, but not set yet, I will try to open it when enslave the dev, better for performance.

Quoted 9 lines
Quoted 1 line
Signed-off-by Ding Tianhong <[email protected]>

I really don't think we should force enable device features in slaves that perhaps the user intentionally disabled, or perhaps the driver has a reason to disable by default (lower performance, etc.)

I'm not applying this series, sorry.

Yes, pls miss this one, after the discussion with Veaceslav, I think it is not reasonable to do this. thanks.

Ding