Date:	Wed, 28 Aug 2013 14:02:21 +0200
From:	Veaceslav Falico <vfalico@...hat.com>
To:	netdev@...r.kernel.org
Cc:	Veaceslav Falico <vfalico@...hat.com>,
	"David S. Miller" <davem@...emloft.net>,
	Eric Dumazet <edumazet@...gle.com>,
	Jiri Pirko <jiri@...nulli.us>,
	Alexander Duyck <alexander.h.duyck@...el.com>,
	Cong Wang <amwang@...hat.com>
Subject: [PATCH net-next v2 02/13] net: add lower_dev_list to net_device and make a full mesh

This patch adds lower_dev_list list_head to net_device, which is the same
as upper_dev_list, only for lower devices, and begins to use it in the same
way as the upper list.

It also changes the way the whole adjacent device lists work - now they
contain *all* of upper/lower devices, not only the first level. The first
level devices are distinguished by the bool neighbour field in
netdev_adjacent, also added by this patch.

There are cases when a device can be added several times to the adjacent
list, the simplest would be:

     /---- eth0.10 ---\
eth0-		       --- bond0
     \---- eth0.20 ---/

where both bond0 and eth0 'see' each other in the adjacent lists two times.
To avoid duplication of netdev_adjacent structures ref_nr is being kept as
the number of times the device was added to the list.

The 'full view' is achieved by adding, on link creation, all of the
upper_dev's upper_dev_list devices as upper devices to all of the
lower_dev's lower_dev_list devices (and to the lower_dev itself), and vice
versa. On unlink they are removed using the same logic.

I've tested it with thousands vlans/bonds/bridges, everything works ok and
no observable lags even on a huge number of interfaces.

Memory footprint for 128 devices interconnected with each other via both
upper and lower (which is impossible, but for the comparison) lists would be:

128*128*2*sizeof(netdev_adjacent) = 1.5MB

but in the real world we usualy have at most several devices with slaves
and a lot of vlans, so the footprint will be much lower.

v2: new patch

CC: "David S. Miller" <davem@...emloft.net>
CC: Eric Dumazet <edumazet@...gle.com>
CC: Jiri Pirko <jiri@...nulli.us>
CC: Alexander Duyck <alexander.h.duyck@...el.com>
CC: Cong Wang <amwang@...hat.com>
Signed-off-by: Veaceslav Falico <vfalico@...hat.com>
---
 include/linux/netdevice.h |    1 +
 net/core/dev.c            |  256 +++++++++++++++++++++++++++++++++++++++-----
 2 files changed, 228 insertions(+), 29 deletions(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index 077363d..5ccf5b7 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -1125,6 +1125,7 @@ struct net_device {
 	struct list_head	napi_list;
 	struct list_head	unreg_list;
 	struct list_head	upper_dev_list; /* List of upper devices */
+	struct list_head	lower_dev_list;
 
 
 	/* currently active device features */
diff --git a/net/core/dev.c b/net/core/dev.c
index c2ac3b8..5d73f9f 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -4369,7 +4369,15 @@ softnet_break:
 
 struct netdev_adjacent {
 	struct net_device *dev;
+	/* master device, we can only have one */
 	bool master;
+	/* is it first-level lower/upper or not */
+	bool neighbour;
+	/* how many times we've tried added this link. if it becomes 0 -
+	 * we can remove the link and free the structure. For neighbour
+	 * links it should always be 1.
+	 */
+	u16 ref_nr;
 	struct list_head list;
 	struct rcu_head rcu;
 	struct list_head search_list;
@@ -4408,18 +4416,25 @@ static bool __netdev_search_upper_dev(struct net_device *dev,
 	return ret;
 }
 
-static struct netdev_adjacent *__netdev_find_upper(struct net_device *dev,
-						struct net_device *upper_dev)
+static struct netdev_adjacent *__netdev_find_adj(struct net_device *dev,
+						 struct net_device *adj_dev,
+						 bool upper)
 {
-	struct netdev_adjacent *upper;
+	struct netdev_adjacent *adj;
+	struct list_head *dev_list;
 
-	list_for_each_entry(upper, &dev->upper_dev_list, list) {
-		if (upper->dev == upper_dev)
-			return upper;
+	dev_list = upper ? &dev->upper_dev_list : &dev->lower_dev_list;
+
+	list_for_each_entry(adj, dev_list, list) {
+		if (adj->dev == adj_dev)
+			return adj;
 	}
 	return NULL;
 }
 
+#define __netdev_find_upper(d, ud) __netdev_find_adj(d, ud, true)
+#define __netdev_find_lower(d, ld) __netdev_find_adj(d, ld, false)
+
 /**
  * netdev_has_upper_dev - Check if device is linked to an upper device
  * @dev: device
@@ -4470,7 +4485,7 @@ struct net_device *netdev_master_upper_dev_get(struct net_device *dev)
 		return NULL;
 
 	upper = list_first_entry(&dev->upper_dev_list,
-				 struct netdev_upper, list);
+				 struct netdev_adjacent, list);
 	if (likely(upper->master))
 		return upper->dev;
 	return NULL;
@@ -4489,17 +4504,133 @@ struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev)
 	struct netdev_adjacent *upper;
 
 	upper = list_first_or_null_rcu(&dev->upper_dev_list,
-				       struct netdev_upper, list);
+				       struct netdev_adjacent, list);
 	if (upper && likely(upper->master))
 		return upper->dev;
 	return NULL;
 }
 EXPORT_SYMBOL(netdev_master_upper_dev_get_rcu);
 
+static int __netdev_adjacent_dev_insert(struct net_device *dev,
+					struct net_device *adj_dev,
+					bool neighbour, bool master,
+					bool upper)
+{
+	struct netdev_adjacent *adj;
+
+	adj = __netdev_find_adj(dev, adj_dev, upper);
+
+	if (adj) {
+		BUG_ON(neighbour);
+		adj->ref_nr++;
+		return 0;
+	}
+
+	adj = kmalloc(sizeof(*adj), GFP_KERNEL);
+	if (!adj)
+		return -ENOMEM;
+
+	adj->dev = adj_dev;
+	adj->master = master;
+	adj->neighbour = neighbour;
+	adj->ref_nr = 1;
+	INIT_LIST_HEAD(&adj->search_list);
+
+	dev_hold(adj_dev);
+	pr_debug("dev_hold for %s, because of %s link added from %s to %s\n",
+		 adj_dev->name, upper ? "upper" : "lower", dev->name,
+		 adj_dev->name);
+
+	if (!upper) {
+		list_add_tail_rcu(&adj->list, &dev->lower_dev_list);
+		return 0;
+	}
+
+	/* Ensure that master upper link is always the first item in list. */
+	if (master)
+		list_add_rcu(&adj->list, &dev->upper_dev_list);
+	else
+		list_add_tail_rcu(&adj->list, &dev->upper_dev_list);
+
+	return 0;
+}
+
+#define __netdev_upper_dev_insert(dev, adev, m, d) \
+		__netdev_adjacent_dev_insert(dev, adev, d, m, true)
+#define __netdev_lower_dev_insert(dev, adev, d) \
+		__netdev_adjacent_dev_insert(dev, adev, d, false, false)
+
+void __netdev_adjacent_dev_remove(struct net_device *dev,
+				 struct net_device *adj_dev, bool upper)
+{
+	struct netdev_adjacent *adj;
+
+	if (upper)
+		adj = __netdev_find_upper(dev, adj_dev);
+	else
+		adj = __netdev_find_lower(dev, adj_dev);
+
+	/* try to maintain the mesh, sorry people, if you've paniced here -
+	 * you've tried to remove an unexisting link, and that is bad.
+	 */
+	if (!adj)
+		BUG();
+
+	if (adj->ref_nr > 1) {
+		adj->ref_nr--;
+		return;
+	}
+
+	list_del_rcu(&adj->list);
+	pr_debug("dev_put for %s, because of %s link removed from %s to %s\n",
+		 adj_dev->name, upper ? "upper" : "lower", dev->name,
+		 adj_dev->name);
+	dev_put(adj_dev);
+	kfree_rcu(adj, rcu);
+}
+
+#define __netdev_upper_dev_remove(d, ud) \
+		__netdev_adjacent_dev_remove(d, ud, true)
+#define __netdev_lower_dev_remove(d, ld) \
+		__netdev_adjacent_dev_remove(d, ld, false)
+
+int __netdev_adjacent_dev_insert_link(struct net_device *dev,
+				      struct net_device *upper_dev,
+				      bool master, bool neighbour)
+{
+	int ret;
+
+	ret = __netdev_upper_dev_insert(dev, upper_dev, master, neighbour);
+	if (ret)
+		return ret;
+
+	ret = __netdev_lower_dev_insert(upper_dev, dev, neighbour);
+	if (ret) {
+		__netdev_upper_dev_remove(dev, upper_dev);
+		return ret;
+	}
+
+	return 0;
+}
+
+#define __netdev_adjacent_dev_link(d, ud) \
+		__netdev_adjacent_dev_insert_link(d, ud, false, false)
+#define __netdev_adjacent_dev_link_neighbour(d, ud, m) \
+		__netdev_adjacent_dev_insert_link(d, ud, m, true)
+
+void __netdev_adjacent_dev_unlink(struct net_device *dev,
+				 struct net_device *upper_dev)
+{
+	__netdev_upper_dev_remove(dev, upper_dev);
+	__netdev_lower_dev_remove(upper_dev, dev);
+}
+
+
 static int __netdev_upper_dev_link(struct net_device *dev,
 				   struct net_device *upper_dev, bool master)
 {
-	struct netdev_adjacent *upper;
+	struct netdev_adjacent *i, *j, *to_i, *to_j;
+	int ret = 0;
 
 	ASSERT_RTNL();
 
@@ -4516,22 +4647,76 @@ static int __netdev_upper_dev_link(struct net_device *dev,
 	if (master && netdev_master_upper_dev_get(dev))
 		return -EBUSY;
 
-	upper = kmalloc(sizeof(*upper), GFP_KERNEL);
-	if (!upper)
-		return -ENOMEM;
+	ret = __netdev_adjacent_dev_link_neighbour(dev, upper_dev, master);
+	if (ret)
+		return ret;
 
-	upper->dev = upper_dev;
-	upper->master = master;
-	INIT_LIST_HEAD(&upper->search_list);
+	/* Now that we interconnected these devs, make all the upper_dev's
+	 * upper_dev_list visible to every dev's lower_dev_list and vice
+	 * versa, and don't forget the devices itself. All of these
+	 * connections are non-neighbours.
+	 */
+	list_for_each_entry(i, &upper_dev->upper_dev_list, list) {
+		list_for_each_entry(j, &dev->lower_dev_list, list) {
+			ret = __netdev_adjacent_dev_link(i->dev, j->dev);
+			if (ret)
+				goto rollback_mesh;
+		}
+	}
+
+	/* add dev to every upper_dev's upper device */
+	list_for_each_entry(i, &upper_dev->upper_dev_list, list) {
+		ret = __netdev_adjacent_dev_link(dev, i->dev);
+		if (ret)
+			goto rollback_upper_mesh;
+	}
+
+	/* add upper_dev to every dev's lower device */
+	list_for_each_entry(i, &dev->lower_dev_list, list) {
+		ret = __netdev_adjacent_dev_link(i->dev, upper_dev);
+		if (ret)
+			goto rollback_lower_mesh;
+	}
 
-	/* Ensure that master upper link is always the first item in list. */
-	if (master)
-		list_add_rcu(&upper->list, &dev->upper_dev_list);
-	else
-		list_add_tail_rcu(&upper->list, &dev->upper_dev_list);
-	dev_hold(upper_dev);
 	call_netdevice_notifiers(NETDEV_CHANGEUPPER, dev);
 	return 0;
+
+rollback_lower_mesh:
+	to_i = i;
+	list_for_each_entry(i, &dev->lower_dev_list, list) {
+		if (i == to_i)
+			break;
+		__netdev_adjacent_dev_unlink(i->dev, upper_dev);
+	}
+
+	i = NULL;
+
+rollback_upper_mesh:
+	to_i = i;
+	list_for_each_entry(i, &upper_dev->upper_dev_list, list) {
+		if (i == to_i)
+			break;
+		__netdev_adjacent_dev_unlink(dev, i->dev);
+	}
+
+	i = j = NULL;
+
+rollback_mesh:
+	to_i = i;
+	to_j = j;
+	list_for_each_entry(i, &dev->lower_dev_list, list) {
+		list_for_each_entry(j, &upper_dev->upper_dev_list, list) {
+			if (i == to_i && j == to_j)
+				break;
+			__netdev_adjacent_dev_unlink(i->dev, j->dev);
+		}
+		if (i == to_i)
+			break;
+	}
+
+	__netdev_adjacent_dev_unlink(dev, upper_dev);
+
+	return ret;
 }
 
 /**
@@ -4580,16 +4765,28 @@ EXPORT_SYMBOL(netdev_master_upper_dev_link);
 void netdev_upper_dev_unlink(struct net_device *dev,
 			     struct net_device *upper_dev)
 {
-	struct netdev_adjacent *upper;
-
+	struct netdev_adjacent *i, *j;
 	ASSERT_RTNL();
 
-	upper = __netdev_find_upper(dev, upper_dev);
-	if (!upper)
-		return;
-	list_del_rcu(&upper->list);
-	dev_put(upper_dev);
-	kfree_rcu(upper, rcu);
+	__netdev_adjacent_dev_unlink(dev, upper_dev);
+
+	/* Here is the tricky part. We must remove all dev's lower
+	 * devices from all upper_dev's upper devices and vice
+	 * versa, to maintain the graph relationship.
+	 */
+	list_for_each_entry(i, &dev->lower_dev_list, list)
+		list_for_each_entry(j, &upper_dev->upper_dev_list, list)
+			__netdev_adjacent_dev_unlink(i->dev, j->dev);
+
+	/* remove also the devices itself from lower/upper device
+	 * list
+	 */
+	list_for_each_entry(i, &dev->lower_dev_list, list)
+		__netdev_adjacent_dev_unlink(i->dev, upper_dev);
+
+	list_for_each_entry(i, &upper_dev->upper_dev_list, list)
+		__netdev_adjacent_dev_unlink(dev, i->dev);
+
 	call_netdevice_notifiers(NETDEV_CHANGEUPPER, dev);
 }
 EXPORT_SYMBOL(netdev_upper_dev_unlink);
@@ -5850,6 +6047,7 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
 	INIT_LIST_HEAD(&dev->unreg_list);
 	INIT_LIST_HEAD(&dev->link_watch_list);
 	INIT_LIST_HEAD(&dev->upper_dev_list);
+	INIT_LIST_HEAD(&dev->lower_dev_list);
 	dev->priv_flags = IFF_XMIT_DST_RELEASE;
 	setup(dev);
 
-- 
1.7.1

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Powered by blists - more mailing lists