[net-next PATCH v2 0/2] bridge: more efficient fdb dumping

From: Jamal Hadi Salim <jhs@mojatatu.com>

Changes:
Suggestions from Vlad
 Get rid of rcu read lock since rtnl_lock is being held
 simplify for readability

Jamal Hadi Salim (2):
  bridge: fdb dumping takes a filter device
  bridge: netlink dump interface at par with brctl

 drivers/net/ethernet/intel/i40e/i40e_main.c      |    3 +-
 drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c |    5 +-
 include/linux/netdevice.h                        |    1 +
 include/linux/rtnetlink.h                        |    1 +
 net/bridge/br_fdb.c                              |   16 +++++
 net/bridge/br_private.h                          |    2 +-
 net/core/rtnetlink.c                             |   75 ++++++++++++++++++----
 7 files changed, 85 insertions(+), 18 deletions(-)

-- 
1.7.9.5

From: Jamal Hadi Salim <jhs@mojatatu.com>

Dumping a bridge fdb dumps every fdb entry
held. With this change we are going to filter
on selected bridge port.

Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
---
 drivers/net/ethernet/intel/i40e/i40e_main.c      |    3 ++-
 drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c |    5 +++--
 include/linux/netdevice.h                        |    1 +
 include/linux/rtnetlink.h                        |    1 +
 net/bridge/br_fdb.c                              |    5 +++++
 net/bridge/br_private.h                          |    2 +-
 net/core/rtnetlink.c                             |    7 ++++---
 7 files changed, 17 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index b167fc2..6d2df6d 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c

@@ -7029,13 +7029,14 @@ static int i40e_ndo_fdb_del(struct ndmsg *ndm,
 static int i40e_ndo_fdb_dump(struct sk_buff *skb,
 			     struct netlink_callback *cb,
 			     struct net_device *dev,
+			     struct net_device *filter_dev,
 			     int idx)
 {
 	struct i40e_netdev_priv *np = netdev_priv(dev);
 	struct i40e_pf *pf = np->vsi->back;
 
 	if (pf->flags & I40E_FLAG_SRIOV_ENABLED)
-		idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
+		idx = ndo_dflt_fdb_dump(skb, cb, dev, filter_dev, idx);
 
 	return idx;
 }

diff --git a/drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c b/drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c
index 4fc1867..b82179d 100644
--- a/drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c
+++ b/drivers/net/ethernet/qlogic/qlcnic/qlcnic_main.c

@@ -427,12 +427,13 @@ static int qlcnic_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
 }
 
 static int qlcnic_fdb_dump(struct sk_buff *skb, struct netlink_callback *ncb,
-			struct net_device *netdev, int idx)
+			struct net_device *netdev, struct net_device *filter_dev,
+			int idx)
 {
 	struct qlcnic_adapter *adapter = netdev_priv(netdev);
 
 	if (!adapter->fdb_mac_learn)
-		return ndo_dflt_fdb_dump(skb, ncb, netdev, idx);
+		return ndo_dflt_fdb_dump(skb, ncb, netdev, filter_dev, idx);
 
 	if ((adapter->flags & QLCNIC_ESWITCH_ENABLED) ||
 	    qlcnic_sriov_check(adapter))

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 66f9a04..8d46984 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h

@@ -1114,6 +1114,7 @@ struct net_device_ops {
 	int			(*ndo_fdb_dump)(struct sk_buff *skb,
 						struct netlink_callback *cb,
 						struct net_device *dev,
+						struct net_device *filter_dev,
 						int idx);
 
 	int			(*ndo_bridge_setlink)(struct net_device *dev,

diff --git a/include/linux/rtnetlink.h b/include/linux/rtnetlink.h
index 953937e..167bae7 100644
--- a/include/linux/rtnetlink.h
+++ b/include/linux/rtnetlink.h

@@ -78,6 +78,7 @@ extern void __rtnl_unlock(void);
 extern int ndo_dflt_fdb_dump(struct sk_buff *skb,
 			     struct netlink_callback *cb,
 			     struct net_device *dev,
+			     struct net_device *filter_dev,
 			     int idx);
 extern int ndo_dflt_fdb_add(struct ndmsg *ndm,
 			    struct nlattr *tb[],

diff --git a/net/bridge/br_fdb.c b/net/bridge/br_fdb.c
index 7be3366..6edecd1 100644
--- a/net/bridge/br_fdb.c
+++ b/net/bridge/br_fdb.c

@@ -676,6 +676,7 @@ errout:
 int br_fdb_dump(struct sk_buff *skb,
 		struct netlink_callback *cb,
 		struct net_device *dev,
+		struct net_device *filter_dev,
 		int idx)
 {
 	struct net_bridge *br = netdev_priv(dev);

@@ -691,6 +692,10 @@ int br_fdb_dump(struct sk_buff *skb,
 			if (idx < cb->args[0])
 				goto skip;
 
+			if (filter_dev && (!f->dst || !f->dst->dev ||
+					   f->dst->dev != filter_dev))
+				goto skip;
+
 			if (fdb_fill_info(skb, br, f,
 					  NETLINK_CB(cb->skb).portid,
 					  cb->nlh->nlmsg_seq,

diff --git a/net/bridge/br_private.h b/net/bridge/br_private.h
index 23caf5b..62a7fa2 100644
--- a/net/bridge/br_private.h
+++ b/net/bridge/br_private.h

@@ -399,7 +399,7 @@ int br_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
 int br_fdb_add(struct ndmsg *nlh, struct nlattr *tb[], struct net_device *dev,
 	       const unsigned char *addr, u16 nlh_flags);
 int br_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
-		struct net_device *dev, int idx);
+		struct net_device *dev, struct net_device *fdev, int idx);
 int br_fdb_sync_static(struct net_bridge *br, struct net_bridge_port *p);
 void br_fdb_unsync_static(struct net_bridge *br, struct net_bridge_port *p);
 

diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index 1063996..7471060 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c

@@ -2509,6 +2509,7 @@ skip:
 int ndo_dflt_fdb_dump(struct sk_buff *skb,
 		      struct netlink_callback *cb,
 		      struct net_device *dev,
+		      struct net_device *filter_dev,
 		      int idx)
 {
 	int err;

@@ -2539,13 +2540,13 @@ static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
 			br_dev = netdev_master_upper_dev_get(dev);
 			ops = br_dev->netdev_ops;
 			if (ops->ndo_fdb_dump)
-				idx = ops->ndo_fdb_dump(skb, cb, dev, idx);
+				idx = ops->ndo_fdb_dump(skb, cb, dev, NULL, idx);
 		}
 
 		if (dev->netdev_ops->ndo_fdb_dump)
-			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, idx);
+			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL, idx);
 		else
-			idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
+			idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
 	}
 	rcu_read_unlock();
 

-- 
1.7.9.5

From: Jamal Hadi Salim <jhs@mojatatu.com>

Actually better than brctl showmacs because we can filter by bridge
port in the kernel.
The current bridge netlink interface doesnt scale when you have many
bridges each with large fdbs or even bridges with many bridge ports

And now for the science non-fiction novel you have all been
waiting for..

//lets see what bridge ports we have
root@moja-1:/configs/may30-iprt/bridge# ./bridge link show
8: eth1 state DOWN : <BROADCAST,MULTICAST> mtu 1500 master br0 state
disabled priority 32 cost 19
17: sw1-p1 state DOWN : <BROADCAST,NOARP> mtu 1500 master br0 state
disabled priority 32 cost 100

// show all..
root@moja-1:/configs/may30-iprt/bridge# ./bridge fdb show
33:33:00:00:00:01 dev bond0 self permanent
33:33:00:00:00:01 dev dummy0 self permanent
33:33:00:00:00:01 dev ifb0 self permanent
33:33:00:00:00:01 dev ifb1 self permanent
33:33:00:00:00:01 dev eth0 self permanent
01:00:5e:00:00:01 dev eth0 self permanent
33:33:ff:22:01:01 dev eth0 self permanent
02:00:00:12:01:02 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:05 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:07 dev eth1 self permanent
33:33:00:00:00:01 dev eth1 self permanent
33:33:00:00:00:01 dev gretap0 self permanent
da:ac:46:27:d9:53 dev sw1-p1 vlan 0 master br0 permanent
33:33:00:00:00:01 dev sw1-p1 self permanent

//filter by bridge
root@moja-1:/configs/may30-iprt/bridge# ./bridge fdb show br br0
02:00:00:12:01:02 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:05 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:07 dev eth1 self permanent
33:33:00:00:00:01 dev eth1 self permanent
da:ac:46:27:d9:53 dev sw1-p1 vlan 0 master br0 permanent
33:33:00:00:00:01 dev sw1-p1 self permanent

// bridge sw1 has no ports attached..
root@moja-1:/configs/may30-iprt/bridge# ./bridge fdb show br sw1

//filter by port
root@moja-1:/configs/may30-iprt/bridge# ./bridge fdb show brport eth1
02:00:00:12:01:02 vlan 0 master br0 permanent
00:17:42:8a:b4:05 vlan 0 master br0 permanent
00:17:42:8a:b4:07 self permanent
33:33:00:00:00:01 self permanent

// filter by port + bridge
root@moja-1:/configs/may30-iprt/bridge# ./bridge fdb show br br0 brport
sw1-p1
da:ac:46:27:d9:53 vlan 0 master br0 permanent
33:33:00:00:00:01 self permanent

// for shits and giggles, lets change the mac that br0 uses
// Note: a magical fdb entry with no brport is added ...
root@moja-1:/configs/may30-iprt/bridge# ip link set dev br0 address
02:00:00:12:01:04

// can we see the unicorn?
root@moja-1:/configs/may30-iprt/bridge# ./bridge fdb show
33:33:00:00:00:01 dev bond0 self permanent
33:33:00:00:00:01 dev dummy0 self permanent
33:33:00:00:00:01 dev ifb0 self permanent
33:33:00:00:00:01 dev ifb1 self permanent
33:33:00:00:00:01 dev eth0 self permanent
01:00:5e:00:00:01 dev eth0 self permanent
33:33:ff:22:01:01 dev eth0 self permanent
02:00:00:12:01:02 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:05 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:07 dev eth1 self permanent
33:33:00:00:00:01 dev eth1 self permanent
33:33:00:00:00:01 dev gretap0 self permanent
02:00:00:12:01:04 dev br0 vlan 0 master br0 permanent <=== there it is
da:ac:46:27:d9:53 dev sw1-p1 vlan 0 master br0 permanent
33:33:00:00:00:01 dev sw1-p1 self permanent

//can we see it if we filter by bridge?
root@moja-1:/configs/may30-iprt/bridge# ./bridge fdb show br br0
02:00:00:12:01:02 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:05 dev eth1 vlan 0 master br0 permanent
00:17:42:8a:b4:07 dev eth1 self permanent
33:33:00:00:00:01 dev eth1 self permanent
02:00:00:12:01:04 dev br0 vlan 0 master br0 permanent <=== there it is
da:ac:46:27:d9:53 dev sw1-p1 vlan 0 master br0 permanent
33:33:00:00:00:01 dev sw1-p1 self permanent

Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
---
 net/bridge/br_fdb.c  |   17 ++++++++++--
 net/core/rtnetlink.c |   74 ++++++++++++++++++++++++++++++++++++++++----------
 2 files changed, 74 insertions(+), 17 deletions(-)

diff --git a/net/bridge/br_fdb.c b/net/bridge/br_fdb.c
index 6edecd1..0776ee6 100644
--- a/net/bridge/br_fdb.c
+++ b/net/bridge/br_fdb.c

@@ -692,9 +692,20 @@ int br_fdb_dump(struct sk_buff *skb,
 			if (idx < cb->args[0])
 				goto skip;
 
-			if (filter_dev && (!f->dst || !f->dst->dev ||
-					   f->dst->dev != filter_dev))
-				goto skip;
+			if (filter_dev && (!f->dst || f->dst->dev != filter_dev)) {
+				if (filter_dev != dev)
+					goto skip;
+				else {
+					/* 
+					 * !f->dst is a speacial case for bridge
+					 * It means the MAC belongs to the bridge
+					 * Therefore need a little more filtering
+					 * we only want to dump the !f->dst case
+					 */
+					if (f->dst)
+						goto skip;
+				}
+			}
 
 			if (fdb_fill_info(skb, br, f,
 					  NETLINK_CB(cb->skb).portid,

diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index 7471060..a8ba071 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c

@@ -2527,28 +2527,74 @@ EXPORT_SYMBOL(ndo_dflt_fdb_dump);
 
 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
 {
-	int idx = 0;
-	struct net *net = sock_net(skb->sk);
 	struct net_device *dev;
+	struct nlattr *tb[IFLA_MAX+1];
+	struct net_device *bdev = NULL;
+	struct net_device *br_dev = NULL;
+	const struct net_device_ops *ops = NULL;
+	const struct net_device_ops *cops = NULL;
+	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
+	struct net *net = sock_net(skb->sk);
+	int brport_idx = 0;
+	int br_idx = 0;
+	int idx = 0;
+
+	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
+			ifla_policy) == 0) {
+		if (tb[IFLA_MASTER])
+			br_idx = nla_get_u32(tb[IFLA_MASTER]);
+	}
+	
+	brport_idx = ifm->ifi_index;
+
+	if (br_idx) {
+		br_dev = __dev_get_by_index(net, br_idx);
+		if (!br_dev) {
+			return -ENODEV;
+		}
+		ops = br_dev->netdev_ops;
+		bdev = br_dev;
+	}
 
-	rcu_read_lock();
 	for_each_netdev_rcu(net, dev) {
+
+		if (brport_idx && (dev->ifindex != brport_idx))
+			continue;
+
+		if (!br_idx) { /* user did not specify a specific bridge */
+			if (dev->priv_flags & IFF_BRIDGE_PORT) {
+				br_dev = netdev_master_upper_dev_get(dev);
+				cops = br_dev->netdev_ops;
+			}
+
+			bdev = dev;
+		} else {
+			if (dev != br_dev &&
+			    !(dev->priv_flags & IFF_BRIDGE_PORT))
+				continue;
+
+			if (br_dev != netdev_master_upper_dev_get(dev) &&
+			    !(dev->priv_flags & IFF_EBRIDGE))
+				continue;
+
+			bdev = br_dev;
+			cops = ops;
+		}
+
 		if (dev->priv_flags & IFF_BRIDGE_PORT) {
-			struct net_device *br_dev;
-			const struct net_device_ops *ops;
+			if (cops && cops->ndo_fdb_dump) {
+				idx = cops->ndo_fdb_dump(skb, cb, br_dev, dev, idx);
+			}
+		}
 
-			br_dev = netdev_master_upper_dev_get(dev);
-			ops = br_dev->netdev_ops;
-			if (ops->ndo_fdb_dump)
-				idx = ops->ndo_fdb_dump(skb, cb, dev, NULL, idx);
+		idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
+		if (dev->netdev_ops->ndo_fdb_dump) {
+			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, bdev, dev,
+							    idx);
 		}
 
-		if (dev->netdev_ops->ndo_fdb_dump)
-			idx = dev->netdev_ops->ndo_fdb_dump(skb, cb, dev, NULL, idx);
-		else
-			idx = ndo_dflt_fdb_dump(skb, cb, dev, NULL, idx);
+		cops = NULL;
 	}
-	rcu_read_unlock();
 
 	cb->args[0] = idx;
 	return skb->len;

-- 
1.7.9.5

On Tue, 2014-06-24 at 10:15 -0400, Jamal Hadi Salim wrote:

From: Jamal Hadi Salim <jhs@mojatatu.com>

...

 
 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
 {
-	int idx = 0;
-	struct net *net = sock_net(skb->sk);
 	struct net_device *dev;
+	struct nlattr *tb[IFLA_MAX+1];
+	struct net_device *bdev = NULL;
+	struct net_device *br_dev = NULL;
+	const struct net_device_ops *ops = NULL;
+	const struct net_device_ops *cops = NULL;
+	struct ifinfomsg *ifm = nlmsg_data(cb->nlh);
+	struct net *net = sock_net(skb->sk);
+	int brport_idx = 0;
+	int br_idx = 0;
+	int idx = 0;
+
+	if (nlmsg_parse(cb->nlh, sizeof(struct ifinfomsg), tb, IFLA_MAX,
+			ifla_policy) == 0) {
+		if (tb[IFLA_MASTER])
+			br_idx = nla_get_u32(tb[IFLA_MASTER]);
+	}
+	
+	brport_idx = ifm->ifi_index;
+
+	if (br_idx) {
+		br_dev = __dev_get_by_index(net, br_idx);
+		if (!br_dev) {
+			return -ENODEV;
+		}
+		ops = br_dev->netdev_ops;
+		bdev = br_dev;
+	}
 
-	rcu_read_lock();
 	for_each_netdev_rcu(net, dev) {

for_each_netdev() 

On Tue, 24 Jun 2014 10:15:19 -0400
Jamal Hadi Salim [off-list ref] wrote:

+			if (filter_dev && (!f->dst || f->dst->dev != filter_dev)) {
+				if (filter_dev != dev)
+					goto skip;
+				else {
+					/* 
+					 * !f->dst is a speacial case for bridge
+					 * It means the MAC belongs to the bridge
+					 * Therefore need a little more filtering
+					 * we only want to dump the !f->dst case
+					 */
+					if (f->dst)
+						goto skip;
+				}

After goto, the else is not needed, and it removes one level of already
deep indentation.

		

			if (filter_dev && (!f->dst || f->dst->dev != filter_dev)) {
				if (filter_dev != dev)
					goto skip;
				/* 
				 * !f->dst is a speacial case for bridge
				 * It means the MAC belongs to the bridge
				 * Therefore need a little more filtering
				 * we only want to dump the !f->dst case
				 */
				if (f->dst)
					goto skip;