net: dynamically allocate fqdir structures

Following patch will add rcu grace period before fqdir
rhashtable destruction, so we need to dynamically allocate
fqdir structures to not force expensive synchronize_rcu() calls
in netns dismantle path.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

3 files changed, 19 insertions, 18 deletions

diff --git a/net/ipv4/inet_fragment.c b/net/ipv4/inet_fragment.c
index a5ec5d956793..b4432f209c71 100644
--- a/net/ipv4/inet_fragment.c
+++ b/net/ipv4/inet_fragment.c
@@ -150,6 +150,7 @@ void fqdir_exit(struct fqdir *fqdir)
 	fqdir->high_thresh = 0; /* prevent creation of new frags */
 
 	rhashtable_free_and_destroy(&fqdir->rhashtable, inet_frags_free_cb, NULL);
+	kfree(fqdir);
 }
 EXPORT_SYMBOL(fqdir_exit);
 
diff --git a/net/ipv4/ip_fragment.c b/net/ipv4/ip_fragment.c
index d59269bbe1b6..1ffaec056821 100644
--- a/net/ipv4/ip_fragment.c
+++ b/net/ipv4/ip_fragment.c
@@ -209,7 +209,7 @@ static struct ipq *ip_find(struct net *net, struct iphdr *iph,
 	};
 	struct inet_frag_queue *q;
 
-	q = inet_frag_find(&net->ipv4.fqdir, &key);
+	q = inet_frag_find(net->ipv4.fqdir, &key);
 	if (!q)
 		return NULL;
 
@@ -589,12 +589,12 @@ static int __net_init ip4_frags_ns_ctl_register(struct net *net)
 			goto err_alloc;
 
 	}
-	table[0].data	= &net->ipv4.fqdir.high_thresh;
-	table[0].extra1	= &net->ipv4.fqdir.low_thresh;
-	table[1].data	= &net->ipv4.fqdir.low_thresh;
-	table[1].extra2	= &net->ipv4.fqdir.high_thresh;
-	table[2].data	= &net->ipv4.fqdir.timeout;
-	table[3].data	= &net->ipv4.fqdir.max_dist;
+	table[0].data	= &net->ipv4.fqdir->high_thresh;
+	table[0].extra1	= &net->ipv4.fqdir->low_thresh;
+	table[1].data	= &net->ipv4.fqdir->low_thresh;
+	table[1].extra2	= &net->ipv4.fqdir->high_thresh;
+	table[2].data	= &net->ipv4.fqdir->timeout;
+	table[3].data	= &net->ipv4.fqdir->max_dist;
 
 	hdr = register_net_sysctl(net, "net/ipv4", table);
 	if (!hdr)
@@ -642,6 +642,9 @@ static int __net_init ipv4_frags_init_net(struct net *net)
 {
 	int res;
 
+	res = fqdir_init(&net->ipv4.fqdir, &ip4_frags, net);
+	if (res < 0)
+		return res;
 	/* Fragment cache limits.
 	 *
 	 * The fragment memory accounting code, (tries to) account for
@@ -656,30 +659,27 @@ static int __net_init ipv4_frags_init_net(struct net *net)
 	 * we will prune down to 3MB, making room for approx 8 big 64K
 	 * fragments 8x128k.
 	 */
-	net->ipv4.fqdir.high_thresh = 4 * 1024 * 1024;
-	net->ipv4.fqdir.low_thresh  = 3 * 1024 * 1024;
+	net->ipv4.fqdir->high_thresh = 4 * 1024 * 1024;
+	net->ipv4.fqdir->low_thresh  = 3 * 1024 * 1024;
 	/*
 	 * Important NOTE! Fragment queue must be destroyed before MSL expires.
 	 * RFC791 is wrong proposing to prolongate timer each fragment arrival
 	 * by TTL.
 	 */
-	net->ipv4.fqdir.timeout = IP_FRAG_TIME;
+	net->ipv4.fqdir->timeout = IP_FRAG_TIME;
 
-	net->ipv4.fqdir.max_dist = 64;
+	net->ipv4.fqdir->max_dist = 64;
 
-	res = fqdir_init(&net->ipv4.fqdir, &ip4_frags, net);
-	if (res < 0)
-		return res;
 	res = ip4_frags_ns_ctl_register(net);
 	if (res < 0)
-		fqdir_exit(&net->ipv4.fqdir);
+		fqdir_exit(net->ipv4.fqdir);
 	return res;
 }
 
 static void __net_exit ipv4_frags_exit_net(struct net *net)
 {
 	ip4_frags_ns_ctl_unregister(net);
-	fqdir_exit(&net->ipv4.fqdir);
+	fqdir_exit(net->ipv4.fqdir);
 }
 
 static struct pernet_operations ip4_frags_ops = {
diff --git a/net/ipv4/proc.c b/net/ipv4/proc.c
index 3927e00084e8..b613572c6616 100644
--- a/net/ipv4/proc.c
+++ b/net/ipv4/proc.c
@@ -72,8 +72,8 @@ static int sockstat_seq_show(struct seq_file *seq, void *v)
 	seq_printf(seq, "RAW: inuse %d\n",
 		   sock_prot_inuse_get(net, &raw_prot));
 	seq_printf(seq,  "FRAG: inuse %u memory %lu\n",
-		   atomic_read(&net->ipv4.fqdir.rhashtable.nelems),
-		   frag_mem_limit(&net->ipv4.fqdir));
+		   atomic_read(&net->ipv4.fqdir->rhashtable.nelems),
+		   frag_mem_limit(net->ipv4.fqdir));
 	return 0;
 }