From: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Date: Thu, 4 Jan 2018 13:15:22 +0000 (+0100)
Subject: Missing nfnl_lock()/nfnl_unlock() is added to ip_set_net_exit()
X-Git-Tag: v6.35~4
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=7527bd335b6384ae59a86c11f643ec5ff488831c;p=thirdparty%2Fipset.git

Missing nfnl_lock()/nfnl_unlock() is added to ip_set_net_exit()

Patch "netfilter: ipset: use nfnl_mutex_is_locked" is added the real
mutex locking check, which revealed the missing locking in ip_set_net_exit().
---

diff --git a/kernel/net/netfilter/ipset/ip_set_core.c b/kernel/net/netfilter/ipset/ip_set_core.c
index 6fbbaeaa..31704e83 100644
--- a/kernel/net/netfilter/ipset/ip_set_core.c
+++ b/kernel/net/netfilter/ipset/ip_set_core.c
@@ -2130,6 +2130,7 @@ ip_set_net_exit(struct net *net)
 
 	inst->is_deleted = true; /* flag for ip_set_nfnl_put */
 
+	nfnl_lock(NFNL_SUBSYS_IPSET);
 	for (i = 0; i < inst->ip_set_max; i++) {
 		set = ip_set(inst, i);
 		if (set) {
@@ -2137,6 +2138,7 @@ ip_set_net_exit(struct net *net)
 			ip_set_destroy_set(set);
 		}
 	}
+	nfnl_unlock(NFNL_SUBSYS_IPSET);
 	kfree(rcu_dereference_protected(inst->ip_set_list, 1));
 #ifndef HAVE_NET_OPS_ID
 	kfree(inst);