[IPV4]: Replace __in_dev_get with __in_dev_get_rcu/rtnl

The following patch renames __in_dev_get() to __in_dev_get_rtnl() and
introduces __in_dev_get_rcu() to cover the second case.

1) RCU with refcnt should use in_dev_get().
2) RCU without refcnt should use __in_dev_get_rcu().
3) All others must hold RTNL and use __in_dev_get_rtnl().

There is one exception in net/ipv4/route.c which is in fact a pre-existing
race condition.  I've marked it as such so that we remember to fix it.

This patch is based on suggestions and prior work by Suzanne Wood and
Paul McKenney.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

10 files changed, 32 insertions, 30 deletions

diff --git a/net/ipv4/arp.c b/net/ipv4/arp.c
index ec0e368..b425748 100644
--- a/net/ipv4/arp.c
+++ b/net/ipv4/arp.c
@@ -241,7 +241,7 @@ static int arp_constructor(struct neighbour *neigh)
 	neigh->type = inet_addr_type(addr);
 
 	rcu_read_lock();
-	in_dev = rcu_dereference(__in_dev_get(dev));
+	in_dev = __in_dev_get_rcu(dev);
 	if (in_dev == NULL) {
 		rcu_read_unlock();
 		return -EINVAL;
@@ -989,8 +989,8 @@ static int arp_req_set(struct arpreq *r, struct net_device * dev)
 			ipv4_devconf.proxy_arp = 1;
 			return 0;
 		}
-		if (__in_dev_get(dev)) {
-			__in_dev_get(dev)->cnf.proxy_arp = 1;
+		if (__in_dev_get_rtnl(dev)) {
+			__in_dev_get_rtnl(dev)->cnf.proxy_arp = 1;
 			return 0;
 		}
 		return -ENXIO;
@@ -1095,8 +1095,8 @@ static int arp_req_delete(struct arpreq *r, struct net_device * dev)
 				ipv4_devconf.proxy_arp = 0;
 				return 0;
 			}
-			if (__in_dev_get(dev)) {
-				__in_dev_get(dev)->cnf.proxy_arp = 0;
+			if (__in_dev_get_rtnl(dev)) {
+				__in_dev_get_rtnl(dev)->cnf.proxy_arp = 0;
 				return 0;
 			}
 			return -ENXIO;
diff --git a/net/ipv4/devinet.c b/net/ipv4/devinet.c
index ba2895a..74f2207 100644
--- a/net/ipv4/devinet.c
+++ b/net/ipv4/devinet.c
@@ -351,7 +351,7 @@ static int inet_insert_ifa(struct in_ifaddr *ifa)
 
 static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa)
 {
-	struct in_device *in_dev = __in_dev_get(dev);
+	struct in_device *in_dev = __in_dev_get_rtnl(dev);
 
 	ASSERT_RTNL();
 
@@ -449,7 +449,7 @@ static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg
 		goto out;
 
 	rc = -ENOBUFS;
-	if ((in_dev = __in_dev_get(dev)) == NULL) {
+	if ((in_dev = __in_dev_get_rtnl(dev)) == NULL) {
 		in_dev = inetdev_init(dev);
 		if (!in_dev)
 			goto out;
@@ -584,7 +584,7 @@ int devinet_ioctl(unsigned int cmd, void __user *arg)
 	if (colon)
 		*colon = ':';
 
-	if ((in_dev = __in_dev_get(dev)) != NULL) {
+	if ((in_dev = __in_dev_get_rtnl(dev)) != NULL) {
 		if (tryaddrmatch) {
 			/* Matthias Andree */
 			/* compare label and address (4.4BSD style) */
@@ -748,7 +748,7 @@ rarok:
 
 static int inet_gifconf(struct net_device *dev, char __user *buf, int len)
 {
-	struct in_device *in_dev = __in_dev_get(dev);
+	struct in_device *in_dev = __in_dev_get_rtnl(dev);
 	struct in_ifaddr *ifa;
 	struct ifreq ifr;
 	int done = 0;
@@ -791,7 +791,7 @@ u32 inet_select_addr(const struct net_device *dev, u32 dst, int scope)
 	struct in_device *in_dev;
 
 	rcu_read_lock();
-	in_dev = __in_dev_get(dev);
+	in_dev = __in_dev_get_rcu(dev);
 	if (!in_dev)
 		goto no_in_dev;
 
@@ -818,7 +818,7 @@ no_in_dev:
 	read_lock(&dev_base_lock);
 	rcu_read_lock();
 	for (dev = dev_base; dev; dev = dev->next) {
-		if ((in_dev = __in_dev_get(dev)) == NULL)
+		if ((in_dev = __in_dev_get_rcu(dev)) == NULL)
 			continue;
 
 		for_primary_ifa(in_dev) {
@@ -887,7 +887,7 @@ u32 inet_confirm_addr(const struct net_device *dev, u32 dst, u32 local, int scop
 
 	if (dev) {
 		rcu_read_lock();
-		if ((in_dev = __in_dev_get(dev)))
+		if ((in_dev = __in_dev_get_rcu(dev)))
 			addr = confirm_addr_indev(in_dev, dst, local, scope);
 		rcu_read_unlock();
 
@@ -897,7 +897,7 @@ u32 inet_confirm_addr(const struct net_device *dev, u32 dst, u32 local, int scop
 	read_lock(&dev_base_lock);
 	rcu_read_lock();
 	for (dev = dev_base; dev; dev = dev->next) {
-		if ((in_dev = __in_dev_get(dev))) {
+		if ((in_dev = __in_dev_get_rcu(dev))) {
 			addr = confirm_addr_indev(in_dev, dst, local, scope);
 			if (addr)
 				break;
@@ -957,7 +957,7 @@ static int inetdev_event(struct notifier_block *this, unsigned long event,
 			 void *ptr)
 {
 	struct net_device *dev = ptr;
-	struct in_device *in_dev = __in_dev_get(dev);
+	struct in_device *in_dev = __in_dev_get_rtnl(dev);
 
 	ASSERT_RTNL();
 
@@ -1078,7 +1078,7 @@ static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
 		if (idx > s_idx)
 			s_ip_idx = 0;
 		rcu_read_lock();
-		if ((in_dev = __in_dev_get(dev)) == NULL) {
+		if ((in_dev = __in_dev_get_rcu(dev)) == NULL) {
 			rcu_read_unlock();
 			continue;
 		}
@@ -1149,7 +1149,7 @@ void inet_forward_change(void)
 	for (dev = dev_base; dev; dev = dev->next) {
 		struct in_device *in_dev;
 		rcu_read_lock();
-		in_dev = __in_dev_get(dev);
+		in_dev = __in_dev_get_rcu(dev);
 		if (in_dev)
 			in_dev->cnf.forwarding = on;
 		rcu_read_unlock();
diff --git a/net/ipv4/fib_frontend.c b/net/ipv4/fib_frontend.c
index 4e1379f..e61bc71 100644
--- a/net/ipv4/fib_frontend.c
+++ b/net/ipv4/fib_frontend.c
@@ -173,7 +173,7 @@ int fib_validate_source(u32 src, u32 dst, u8 tos, int oif,
 
 	no_addr = rpf = 0;
 	rcu_read_lock();
-	in_dev = __in_dev_get(dev);
+	in_dev = __in_dev_get_rcu(dev);
 	if (in_dev) {
 		no_addr = in_dev->ifa_list == NULL;
 		rpf = IN_DEV_RPFILTER(in_dev);
@@ -607,7 +607,7 @@ static int fib_inetaddr_event(struct notifier_block *this, unsigned long event,
 static int fib_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
 {
 	struct net_device *dev = ptr;
-	struct in_device *in_dev = __in_dev_get(dev);
+	struct in_device *in_dev = __in_dev_get_rtnl(dev);
 
 	if (event == NETDEV_UNREGISTER) {
 		fib_disable_ip(dev, 2);
diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c
index d41219e..186f20c 100644
--- a/net/ipv4/fib_semantics.c
+++ b/net/ipv4/fib_semantics.c
@@ -1087,7 +1087,7 @@ fib_convert_rtentry(int cmd, struct nlmsghdr *nl, struct rtmsg *rtm,
 		rta->rta_oif = &dev->ifindex;
 		if (colon) {
 			struct in_ifaddr *ifa;
-			struct in_device *in_dev = __in_dev_get(dev);
+			struct in_device *in_dev = __in_dev_get_rtnl(dev);
 			if (!in_dev)
 				return -ENODEV;
 			*colon = ':';
@@ -1268,7 +1268,7 @@ int fib_sync_up(struct net_device *dev)
 			}
 			if (nh->nh_dev == NULL || !(nh->nh_dev->flags&IFF_UP))
 				continue;
-			if (nh->nh_dev != dev || __in_dev_get(dev) == NULL)
+			if (nh->nh_dev != dev || !__in_dev_get_rtnl(dev))
 				continue;
 			alive++;
 			spin_lock_bh(&fib_multipath_lock);
diff --git a/net/ipv4/igmp.c b/net/ipv4/igmp.c
index 70c44e4..8b6d393 100644
--- a/net/ipv4/igmp.c
+++ b/net/ipv4/igmp.c
@@ -1323,7 +1323,7 @@ static struct in_device * ip_mc_find_dev(struct ip_mreqn *imr)
 	}
 	if (dev) {
 		imr->imr_ifindex = dev->ifindex;
-		idev = __in_dev_get(dev);
+		idev = __in_dev_get_rtnl(dev);
 	}
 	return idev;
 }
diff --git a/net/ipv4/ip_gre.c b/net/ipv4/ip_gre.c
index f0d5740..896ce3f 100644
--- a/net/ipv4/ip_gre.c
+++ b/net/ipv4/ip_gre.c
@@ -1104,10 +1104,10 @@ static int ipgre_open(struct net_device *dev)
 			return -EADDRNOTAVAIL;
 		dev = rt->u.dst.dev;
 		ip_rt_put(rt);
-		if (__in_dev_get(dev) == NULL)
+		if (__in_dev_get_rtnl(dev) == NULL)
 			return -EADDRNOTAVAIL;
 		t->mlink = dev->ifindex;
-		ip_mc_inc_group(__in_dev_get(dev), t->parms.iph.daddr);
+		ip_mc_inc_group(__in_dev_get_rtnl(dev), t->parms.iph.daddr);
 	}
 	return 0;
 }
diff --git a/net/ipv4/ipmr.c b/net/ipv4/ipmr.c
index 9dbf590..302b7eb 100644
--- a/net/ipv4/ipmr.c
+++ b/net/ipv4/ipmr.c
@@ -149,7 +149,7 @@ struct net_device *ipmr_new_tunnel(struct vifctl *v)
 		if (err == 0 && (dev = __dev_get_by_name(p.name)) != NULL) {
 			dev->flags |= IFF_MULTICAST;
 
-			in_dev = __in_dev_get(dev);
+			in_dev = __in_dev_get_rtnl(dev);
 			if (in_dev == NULL && (in_dev = inetdev_init(dev)) == NULL)
 				goto failure;
 			in_dev->cnf.rp_filter = 0;
@@ -278,7 +278,7 @@ static int vif_delete(int vifi)
 
 	dev_set_allmulti(dev, -1);
 
-	if ((in_dev = __in_dev_get(dev)) != NULL) {
+	if ((in_dev = __in_dev_get_rtnl(dev)) != NULL) {
 		in_dev->cnf.mc_forwarding--;
 		ip_rt_multicast_event(in_dev);
 	}
@@ -421,7 +421,7 @@ static int vif_add(struct vifctl *vifc, int mrtsock)
 		return -EINVAL;
 	}
 
-	if ((in_dev = __in_dev_get(dev)) == NULL)
+	if ((in_dev = __in_dev_get_rtnl(dev)) == NULL)
 		return -EADDRNOTAVAIL;
 	in_dev->cnf.mc_forwarding++;
 	dev_set_allmulti(dev, +1);
diff --git a/net/ipv4/netfilter/ip_conntrack_netbios_ns.c b/net/ipv4/netfilter/ip_conntrack_netbios_ns.c
index 577bac2..186646e 100644
--- a/net/ipv4/netfilter/ip_conntrack_netbios_ns.c
+++ b/net/ipv4/netfilter/ip_conntrack_netbios_ns.c
@@ -58,7 +58,7 @@ static int help(struct sk_buff **pskb,
 		goto out;
 
 	rcu_read_lock();
-	in_dev = __in_dev_get(rt->u.dst.dev);
+	in_dev = __in_dev_get_rcu(rt->u.dst.dev);
 	if (in_dev != NULL) {
 		for_primary_ifa(in_dev) {
 			if (ifa->ifa_broadcast == iph->daddr) {
diff --git a/net/ipv4/netfilter/ipt_REDIRECT.c b/net/ipv4/netfilter/ipt_REDIRECT.c
index 715cb61..5245bfd 100644
--- a/net/ipv4/netfilter/ipt_REDIRECT.c
+++ b/net/ipv4/netfilter/ipt_REDIRECT.c
@@ -93,7 +93,7 @@ redirect_target(struct sk_buff **pskb,
 		newdst = 0;
 		
 		rcu_read_lock();
-		indev = __in_dev_get((*pskb)->dev);
+		indev = __in_dev_get_rcu((*pskb)->dev);
 		if (indev && (ifa = indev->ifa_list))
 			newdst = ifa->ifa_local;
 		rcu_read_unlock();
diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index 8549f26..381dd6a 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c
@@ -2128,7 +2128,7 @@ int ip_route_input(struct sk_buff *skb, u32 daddr, u32 saddr,
 		struct in_device *in_dev;
 
 		rcu_read_lock();
-		if ((in_dev = __in_dev_get(dev)) != NULL) {
+		if ((in_dev = __in_dev_get_rcu(dev)) != NULL) {
 			int our = ip_check_mc(in_dev, daddr, saddr,
 				skb->nh.iph->protocol);
 			if (our
@@ -2443,7 +2443,9 @@ static int ip_route_output_slow(struct rtable **rp, const struct flowi *oldflp)
 		err = -ENODEV;
 		if (dev_out == NULL)
 			goto out;
-		if (__in_dev_get(dev_out) == NULL) {
+
+		/* RACE: Check return value of inet_select_addr instead. */
+		if (__in_dev_get_rtnl(dev_out) == NULL) {
 			dev_put(dev_out);
 			goto out;	/* Wrong error code */
 		}