[Patch net v2] rtnetlink: avoid RCU read lock when holding RTNL

[Cong Wang <xiyou.wangcong@gmail.com>]

From: Cong Wang <cong.wang@bytedance.com>

When we call af_ops->set_link_af() we hold a RCU read lock
as we retrieve af_ops from the RCU protected list, but this
is unnecessary because we already hold RTNL lock, which is
the writer lock for protecting rtnl_af_ops, so it is safer
than RCU read lock. Similar for af_ops->validate_link_af().

This was not a problem until we begin to take mutex lock
down the path of ->set_link_af() in __ipv6_dev_mc_dec()
recently. We can just drop the RCU read lock there and
assert RTNL lock.

Reported-and-tested-by: syzbot+7d941e89dd48bcf42573@syzkaller.appspotmail.com
Fixes: 63ed8de4be81 ("mld: add mc_lock for protecting per-interface mld data")
Tested-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: Cong Wang <cong.wang@bytedance.com>
---
 net/core/rtnetlink.c | 26 +++++++-------------------
 net/ipv4/devinet.c   |  4 ++--
 2 files changed, 9 insertions(+), 21 deletions(-)

diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index 714d5fa38546..04b4f0f2a3d2 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -543,7 +543,9 @@ static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 {
 	const struct rtnl_af_ops *ops;
 
-	list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
+	ASSERT_RTNL();
+
+	list_for_each_entry(ops, &rtnl_af_ops, list) {
 		if (ops->family == family)
 			return ops;
 	}
@@ -2274,27 +2276,18 @@ static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
 			const struct rtnl_af_ops *af_ops;
 
-			rcu_read_lock();
 			af_ops = rtnl_af_lookup(nla_type(af));
-			if (!af_ops) {
-				rcu_read_unlock();
+			if (!af_ops)
 				return -EAFNOSUPPORT;
-			}
 
-			if (!af_ops->set_link_af) {
-				rcu_read_unlock();
+			if (!af_ops->set_link_af)
 				return -EOPNOTSUPP;
-			}
 
 			if (af_ops->validate_link_af) {
 				err = af_ops->validate_link_af(dev, af);
-				if (err < 0) {
-					rcu_read_unlock();
+				if (err < 0)
 					return err;
-				}
 			}
-
-			rcu_read_unlock();
 		}
 	}
 
@@ -2868,17 +2861,12 @@ static int do_setlink(const struct sk_buff *skb,
 		nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
 			const struct rtnl_af_ops *af_ops;
 
-			rcu_read_lock();
-
 			BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
 
 			err = af_ops->set_link_af(dev, af, extack);
-			if (err < 0) {
-				rcu_read_unlock();
+			if (err < 0)
 				goto errout;
-			}
 
-			rcu_read_unlock();
 			status |= DO_SETLINK_NOTIFY;
 		}
 	}
diff --git a/net/ipv4/devinet.c b/net/ipv4/devinet.c
index 2e35f68da40a..50deeff48c8b 100644
--- a/net/ipv4/devinet.c
+++ b/net/ipv4/devinet.c
@@ -1955,7 +1955,7 @@ static int inet_validate_link_af(const struct net_device *dev,
 	struct nlattr *a, *tb[IFLA_INET_MAX+1];
 	int err, rem;
 
-	if (dev && !__in_dev_get_rcu(dev))
+	if (dev && !__in_dev_get_rtnl(dev))
 		return -EAFNOSUPPORT;
 
 	err = nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla,
@@ -1981,7 +1981,7 @@ static int inet_validate_link_af(const struct net_device *dev,
 static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla,
 			    struct netlink_ext_ack *extack)
 {
-	struct in_device *in_dev = __in_dev_get_rcu(dev);
+	struct in_device *in_dev = __in_dev_get_rtnl(dev);
 	struct nlattr *a, *tb[IFLA_INET_MAX+1];
 	int rem;
 
-- 
2.25.1