net: dont hold rtnl mutex during netlink dump callbacks

Four years ago, Patrick made a change to hold rtnl mutex during netlink
dump callbacks.

I believe it was a wrong move. This slows down concurrent dumps, making
good old /proc/net/ files faster than rtnetlink in some situations.

This occurred to me because one "ip link show dev ..." was _very_ slow
on a workload adding/removing network devices in background.

All dump callbacks are able to use RCU locking now, so this patch does
roughly a revert of commits :

1c2d670f366 : [RTNETLINK]: Hold rtnl_mutex during netlink dump callbacks
6313c1e0992 : [RTNETLINK]: Remove unnecessary locking in dump callbacks

This let writers fight for rtnl mutex and readers going full speed.

It also takes care of phonet : phonet_route_get() is now called from rcu
read section. I renamed it to phonet_route_get_rcu()

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Patrick McHardy <kaber@trash.net>
Cc: Remi Denis-Courmont <remi.denis-courmont@nokia.com>
Acked-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

2 files changed, 7 insertions, 8 deletions

diff --git a/net/core/fib_rules.c b/net/core/fib_rules.c
index 8248ebb5891d..3911586e12e4 100644
--- a/net/core/fib_rules.c
+++ b/net/core/fib_rules.c
@@ -590,7 +590,8 @@ static int dump_rules(struct sk_buff *skb, struct netlink_callback *cb,
 	int idx = 0;
 	struct fib_rule *rule;
 
-	list_for_each_entry(rule, &ops->rules_list, list) {
+	rcu_read_lock();
+	list_for_each_entry_rcu(rule, &ops->rules_list, list) {
 		if (idx < cb->args[1])
 			goto skip;
 
diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index d7c4bb4b1820..296331257195 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -1007,10 +1007,11 @@ static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
 	s_h = cb->args[0];
 	s_idx = cb->args[1];
 
+	rcu_read_lock();
 	for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
 		idx = 0;
 		head = &net->dev_index_head[h];
-		hlist_for_each_entry(dev, node, head, index_hlist) {
+		hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
 			if (idx < s_idx)
 				goto cont;
 			if (rtnl_fill_ifinfo(skb, dev, RTM_NEWLINK,
@@ -1023,6 +1024,7 @@ cont:
 		}
 	}
 out:
+	rcu_read_unlock();
 	cb->args[1] = idx;
 	cb->args[0] = h;
 
@@ -1879,7 +1881,6 @@ static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
 	int min_len;
 	int family;
 	int type;
-	int err;
 
 	type = nlh->nlmsg_type;
 	if (type > RTM_MAX)
@@ -1906,11 +1907,8 @@ static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
 		if (dumpit == NULL)
 			return -EOPNOTSUPP;
 
-		__rtnl_unlock();
 		rtnl = net->rtnl;
-		err = netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
-		rtnl_lock();
-		return err;
+		return netlink_dump_start(rtnl, skb, nlh, dumpit, NULL);
 	}
 
 	memset(rta_buf, 0, (rtattr_max * sizeof(struct rtattr *)));
@@ -1980,7 +1978,7 @@ static int __net_init rtnetlink_net_init(struct net *net)
 {
 	struct sock *sk;
 	sk = netlink_kernel_create(net, NETLINK_ROUTE, RTNLGRP_MAX,
-				   rtnetlink_rcv, &rtnl_mutex, THIS_MODULE);
+				   rtnetlink_rcv, NULL, THIS_MODULE);
 	if (!sk)
 		return -ENOMEM;
 	net->rtnl = sk;