net: Introduce for_each_netdev_rcu() iterator

Adds RCU management to the list of netdevices.

Convert some for_each_netdev() users to RCU version, if
it can avoid read_lock-ing dev_base_lock

Ie:
	read_lock(&dev_base_loack);
	for_each_netdev(net, dev)
		some_action();
	read_unlock(&dev_base_lock);

becomes :

	rcu_read_lock();
	for_each_netdev_rcu(net, dev)
		some_action();
	rcu_read_unlock();


Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/linux/netdevice.h

@@ -1081,6 +1081,8 @@ extern rwlock_t				dev_base_lock;		/* Device list lock */
 
 #define for_each_netdev(net, d)		\
 		list_for_each_entry(d, &(net)->dev_base_head, dev_list)
+#define for_each_netdev_rcu(net, d)		\
+		list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
 #define for_each_netdev_safe(net, d, n)	\
 		list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
 #define for_each_netdev_continue(net, d)		\

net/core/dev.c

@@ -175,7 +175,7 @@ static struct list_head ptype_all __read_mostly;	/* Taps */
  * The @dev_base_head list is protected by @dev_base_lock and the rtnl
  * semaphore.
  *
- * Pure readers hold dev_base_lock for reading.
+ * Pure readers hold dev_base_lock for reading, or rcu_read_lock()
  *
  * Writers must hold the rtnl semaphore while they loop through the
  * dev_base_head list, and hold dev_base_lock for writing when they do the
@@ -212,7 +212,7 @@ static int list_netdevice(struct net_device *dev)
 	ASSERT_RTNL();
 
 	write_lock_bh(&dev_base_lock);
-	list_add_tail(&dev->dev_list, &net->dev_base_head);
+	list_add_tail_rcu(&dev->dev_list, &net->dev_base_head);
 	hlist_add_head_rcu(&dev->name_hlist, dev_name_hash(net, dev->name));
 	hlist_add_head_rcu(&dev->index_hlist,
 			   dev_index_hash(net, dev->ifindex));
@@ -229,7 +229,7 @@ static void unlist_netdevice(struct net_device *dev)
 
 	/* Unlink dev from the device chain */
 	write_lock_bh(&dev_base_lock);
-	list_del(&dev->dev_list);
+	list_del_rcu(&dev->dev_list);
 	hlist_del_rcu(&dev->name_hlist);
 	hlist_del_rcu(&dev->index_hlist);
 	write_unlock_bh(&dev_base_lock);
@@ -799,15 +799,15 @@ struct net_device *dev_get_by_flags(struct net *net, unsigned short if_flags,
 	struct net_device *dev, *ret;
 
 	ret = NULL;
-	read_lock(&dev_base_lock);
-	for_each_netdev(net, dev) {
+	rcu_read_lock();
+	for_each_netdev_rcu(net, dev) {
 		if (((dev->flags ^ if_flags) & mask) == 0) {
 			dev_hold(dev);
 			ret = dev;
 			break;
 		}
 	}
-	read_unlock(&dev_base_lock);
+	rcu_read_unlock();
 	return ret;
 }
 EXPORT_SYMBOL(dev_get_by_flags);
@@ -3077,18 +3077,18 @@ static int dev_ifconf(struct net *net, char __user *arg)
  *	in detail.
  */
 void *dev_seq_start(struct seq_file *seq, loff_t *pos)
-	__acquires(dev_base_lock)
+	__acquires(RCU)
 {
 	struct net *net = seq_file_net(seq);
 	loff_t off;
 	struct net_device *dev;
 
-	read_lock(&dev_base_lock);
+	rcu_read_lock();
 	if (!*pos)
 		return SEQ_START_TOKEN;
 
 	off = 1;
-	for_each_netdev(net, dev)
+	for_each_netdev_rcu(net, dev)
 		if (off++ == *pos)
 			return dev;
 
@@ -3097,16 +3097,18 @@ void *dev_seq_start(struct seq_file *seq, loff_t *pos)
 
 void *dev_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 {
-	struct net *net = seq_file_net(seq);
+	struct net_device *dev = (v == SEQ_START_TOKEN) ?
+				  first_net_device(seq_file_net(seq)) :
+				  next_net_device((struct net_device *)v);
+
 	++*pos;
-	return v == SEQ_START_TOKEN ?
-		first_net_device(net) : next_net_device((struct net_device *)v);
+	return rcu_dereference(dev);
 }
 
 void dev_seq_stop(struct seq_file *seq, void *v)
-	__releases(dev_base_lock)
+	__releases(RCU)
 {
-	read_unlock(&dev_base_lock);
+	rcu_read_unlock();
 }
 
 static void dev_seq_printf_stats(struct seq_file *seq, struct net_device *dev)

net/decnet/af_decnet.c

@@ -749,17 +749,17 @@ static int dn_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 
 	if (!(saddr->sdn_flags & SDF_WILD)) {
 		if (le16_to_cpu(saddr->sdn_nodeaddrl)) {
-			read_lock(&dev_base_lock);
+			rcu_read_lock();
 			ldev = NULL;
-			for_each_netdev(&init_net, dev) {
+			for_each_netdev_rcu(&init_net, dev) {
 				if (!dev->dn_ptr)
 					continue;
 				if (dn_dev_islocal(dev, dn_saddr2dn(saddr))) {
 					ldev = dev;
 					break;
 				}
 			}
-			read_unlock(&dev_base_lock);
+			rcu_read_unlock();
 			if (ldev == NULL)
 				return -EADDRNOTAVAIL;
 		}

net/decnet/dn_fib.c

@@ -607,8 +607,8 @@ static void dn_fib_del_ifaddr(struct dn_ifaddr *ifa)
 	ASSERT_RTNL();
 
 	/* Scan device list */
-	read_lock(&dev_base_lock);
-	for_each_netdev(&init_net, dev) {
+	rcu_read_lock();
+	for_each_netdev_rcu(&init_net, dev) {
 		dn_db = dev->dn_ptr;
 		if (dn_db == NULL)
 			continue;
@@ -619,7 +619,7 @@ static void dn_fib_del_ifaddr(struct dn_ifaddr *ifa)
 			}
 		}
 	}
-	read_unlock(&dev_base_lock);
+	rcu_read_unlock();
 
 	if (found_it == 0) {
 		fib_magic(RTM_DELROUTE, RTN_LOCAL, ifa->ifa_local, 16, ifa);

net/decnet/dn_route.c

@@ -908,8 +908,8 @@ static int dn_route_output_slow(struct dst_entry **pprt, const struct flowi *old
 			dev_put(dev_out);
 			goto out;
 		}
-		read_lock(&dev_base_lock);
-		for_each_netdev(&init_net, dev) {
+		rcu_read_lock();
+		for_each_netdev_rcu(&init_net, dev) {
 			if (!dev->dn_ptr)
 				continue;
 			if (!dn_dev_islocal(dev, oldflp->fld_src))
@@ -922,7 +922,7 @@ static int dn_route_output_slow(struct dst_entry **pprt, const struct flowi *old
 			dev_out = dev;
 			break;
 		}
-		read_unlock(&dev_base_lock);
+		rcu_read_unlock();
 		if (dev_out == NULL)
 			goto out;
 		dev_hold(dev_out);

net/ipv4/devinet.c

@@ -876,32 +876,28 @@ __be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope)
 		if (!addr)
 			addr = ifa->ifa_local;
 	} endfor_ifa(in_dev);
-no_in_dev:
-	rcu_read_unlock();
 
+no_in_dev:
 	if (addr)
-		goto out;
+		goto out_unlock;
 
 	/* Not loopback addresses on loopback should be preferred
 	   in this case. It is importnat that lo is the first interface
 	   in dev_base list.
 	 */
-	read_lock(&dev_base_lock);
-	rcu_read_lock();
-	for_each_netdev(net, dev) {
+	for_each_netdev_rcu(net, dev) {
 		if ((in_dev = __in_dev_get_rcu(dev)) == NULL)
 			continue;
 
 		for_primary_ifa(in_dev) {
 			if (ifa->ifa_scope != RT_SCOPE_LINK &&
 			    ifa->ifa_scope <= scope) {
 				addr = ifa->ifa_local;
-				goto out_unlock_both;
+				goto out_unlock;
 			}
 		} endfor_ifa(in_dev);
 	}
-out_unlock_both:
-	read_unlock(&dev_base_lock);
+out_unlock:
 	rcu_read_unlock();
 out:
 	return addr;
@@ -962,17 +958,15 @@ __be32 inet_confirm_addr(struct in_device *in_dev,
 		return confirm_addr_indev(in_dev, dst, local, scope);
 
 	net = dev_net(in_dev->dev);
-	read_lock(&dev_base_lock);
 	rcu_read_lock();
-	for_each_netdev(net, dev) {
+	for_each_netdev_rcu(net, dev) {
 		if ((in_dev = __in_dev_get_rcu(dev))) {
 			addr = confirm_addr_indev(in_dev, dst, local, scope);
 			if (addr)
 				break;
 		}
 	}
 	rcu_read_unlock();
-	read_unlock(&dev_base_lock);
 
 	return addr;
 }
@@ -1240,18 +1234,18 @@ static void devinet_copy_dflt_conf(struct net *net, int i)
 {
 	struct net_device *dev;
 
-	read_lock(&dev_base_lock);
-	for_each_netdev(net, dev) {
+	rcu_read_lock();
+	for_each_netdev_rcu(net, dev) {
 		struct in_device *in_dev;
-		rcu_read_lock();
+
 		in_dev = __in_dev_get_rcu(dev);
 		if (in_dev && !test_bit(i, in_dev->cnf.state))
 			in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i];
-		rcu_read_unlock();
 	}
-	read_unlock(&dev_base_lock);
+	rcu_read_unlock();
 }
 
+/* called with RTNL locked */
 static void inet_forward_change(struct net *net)
 {
 	struct net_device *dev;
@@ -1260,7 +1254,6 @@ static void inet_forward_change(struct net *net)
 	IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on;
 	IPV4_DEVCONF_DFLT(net, FORWARDING) = on;
 
-	read_lock(&dev_base_lock);
 	for_each_netdev(net, dev) {
 		struct in_device *in_dev;
 		if (on)
@@ -1271,7 +1264,6 @@ static void inet_forward_change(struct net *net)
 			IN_DEV_CONF_SET(in_dev, FORWARDING, on);
 		rcu_read_unlock();
 	}
-	read_unlock(&dev_base_lock);
 }
 
 static int devinet_conf_proc(ctl_table *ctl, int write,

net/ipv6/addrconf.c

@@ -481,19 +481,17 @@ static void addrconf_forward_change(struct net *net, __s32 newf)
 	struct net_device *dev;
 	struct inet6_dev *idev;
 
-	read_lock(&dev_base_lock);
-	for_each_netdev(net, dev) {
-		rcu_read_lock();
+	rcu_read_lock();
+	for_each_netdev_rcu(net, dev) {
 		idev = __in6_dev_get(dev);
 		if (idev) {
 			int changed = (!idev->cnf.forwarding) ^ (!newf);
 			idev->cnf.forwarding = newf;
 			if (changed)
 				dev_forward_change(idev);
 		}
-		rcu_read_unlock();
 	}
-	read_unlock(&dev_base_lock);
+	rcu_read_unlock();
 }
 
 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
@@ -1137,10 +1135,9 @@ int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
 	hiscore->rule = -1;
 	hiscore->ifa = NULL;
 
-	read_lock(&dev_base_lock);
 	rcu_read_lock();
 
-	for_each_netdev(net, dev) {
+	for_each_netdev_rcu(net, dev) {
 		struct inet6_dev *idev;
 
 		/* Candidate Source Address (section 4)
@@ -1235,7 +1232,6 @@ int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
 		read_unlock_bh(&idev->lock);
 	}
 	rcu_read_unlock();
-	read_unlock(&dev_base_lock);
 
 	if (!hiscore->ifa)
 		return -EADDRNOTAVAIL;
@@ -4052,19 +4048,17 @@ static void addrconf_disable_change(struct net *net, __s32 newf)
 	struct net_device *dev;
 	struct inet6_dev *idev;
 
-	read_lock(&dev_base_lock);
-	for_each_netdev(net, dev) {
-		rcu_read_lock();
+	rcu_read_lock();
+	for_each_netdev_rcu(net, dev) {
 		idev = __in6_dev_get(dev);
 		if (idev) {
 			int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
 			idev->cnf.disable_ipv6 = newf;
 			if (changed)
 				dev_disable_change(idev);
 		}
-		rcu_read_unlock();
 	}
-	read_unlock(&dev_base_lock);
+	rcu_read_unlock();
 }
 
 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)

net/ipv6/anycast.c

@@ -404,13 +404,13 @@ int ipv6_chk_acast_addr(struct net *net, struct net_device *dev,
 
 	if (dev)
 		return ipv6_chk_acast_dev(dev, addr);
-	read_lock(&dev_base_lock);
-	for_each_netdev(net, dev)
+	rcu_read_lock();
+	for_each_netdev_rcu(net, dev)
 		if (ipv6_chk_acast_dev(dev, addr)) {
 			found = 1;
 			break;
 		}
-	read_unlock(&dev_base_lock);
+	rcu_read_unlock();
 	return found;
 }
 

net/netrom/nr_route.c

@@ -597,15 +597,15 @@ struct net_device *nr_dev_first(void)
 {
 	struct net_device *dev, *first = NULL;
 
-	read_lock(&dev_base_lock);
-	for_each_netdev(&init_net, dev) {
+	rcu_read_lock();
+	for_each_netdev_rcu(&init_net, dev) {
 		if ((dev->flags & IFF_UP) && dev->type == ARPHRD_NETROM)
 			if (first == NULL || strncmp(dev->name, first->name, 3) < 0)
 				first = dev;
 	}
 	if (first)
 		dev_hold(first);
-	read_unlock(&dev_base_lock);
+	rcu_read_unlock();
 
 	return first;
 }
@@ -617,16 +617,17 @@ struct net_device *nr_dev_get(ax25_address *addr)
 {
 	struct net_device *dev;
 
-	read_lock(&dev_base_lock);
-	for_each_netdev(&init_net, dev) {
-		if ((dev->flags & IFF_UP) && dev->type == ARPHRD_NETROM && ax25cmp(addr, (ax25_address *)dev->dev_addr) == 0) {
+	rcu_read_lock();
+	for_each_netdev_rcu(&init_net, dev) {
+		if ((dev->flags & IFF_UP) && dev->type == ARPHRD_NETROM &&
+		    ax25cmp(addr, (ax25_address *)dev->dev_addr) == 0) {
 			dev_hold(dev);
 			goto out;
 		}
 	}
 	dev = NULL;
 out:
-	read_unlock(&dev_base_lock);
+	rcu_read_unlock();
 	return dev;
 }