Skip to contentSkip to conversation

netfilter: conntrack: allow increasing bucket size via sysctl too

PATCH Applied Florian Westphal 2 participants 2 messages Jun 22, 2016 – Jun 23, 2016

Subsystems (3)
lore
Share & tools
lore

Florian Westphal

author maintainer
#0 lore raw

PATCH v3 1/1 netfilter: conntrack: allow increasing bucket size via sysctl too

No need to restrict this to module parameter.

We export a copy of the real hash size -- when user alters the value we allocate the new table, copy entries etc before we update the real size to the requested one.

This is also needed because the real size is used by concurrent readers and cannot be changed without synchronizing the conntrack generation seqcnt.

We only allow changing this value from the initial net namespace.

Tested using http-client-benchmark vs. httpterm with concurrent

while true;do
echo $RANDOM > /proc/sys/net/netfilter/nf_conntrack_buckets
done

---
Changes since v2:
add back nf_conntrack-sysctl.txt change
don't call synchronize_net twice

Changes since v1:
place assignment inside CONFIG_SYSCL so we don't get build breakage.

 Documentation/networking/nf_conntrack-sysctl.txt |  3 +-
 include/net/netfilter/nf_conntrack.h             |  1 +
 net/netfilter/nf_conntrack_core.c                | 41 ++++++++++++++++--------
 net/netfilter/nf_conntrack_standalone.c          | 36 ++++++++++++++++++---
 4 files changed, 62 insertions(+), 19 deletions(-)
Documentation/networking/nf_conntrack-sysctl.txt +2 -1
index f55599c..4fb51d3 100644
@@ -7,12 +7,13 @@ nf_conntrack_acct - BOOLEAN
 	Enable connection tracking flow accounting. 64-bit byte and packet
 	counters per flow are added.
 
-nf_conntrack_buckets - INTEGER (read-only)
+nf_conntrack_buckets - INTEGER
 	Size of hash table. If not specified as parameter during module
 	loading, the default size is calculated by dividing total memory
 	by 16384 to determine the number of buckets but the hash table will
 	never have fewer than 32 and limited to 16384 buckets. For systems
 	with more than 4GB of memory it will be 65536 buckets.
+	This sysctl is only writeable in the initial net namespace.
 
 nf_conntrack_checksum - BOOLEAN
 	0 - disabled
include/net/netfilter/nf_conntrack.h +1 -0
index dd78bea..e7f6290 100644
@@ -287,6 +287,7 @@ static inline bool nf_is_loopback_packet(const struct sk_buff *skb)
 struct kernel_param;
 
 int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp);
+int nf_conntrack_hash_resize(unsigned int hashsize);
 extern unsigned int nf_conntrack_htable_size;
 extern unsigned int nf_conntrack_max;
 
net/netfilter/nf_conntrack_core.c +28 -13

64 lines not on this page · raw message

net/netfilter/nf_conntrack_standalone.c +31 -5
index f87e84e..a0cc191 100644
@@ -434,8 +434,29 @@ static void nf_conntrack_standalone_fini_proc(struct net *net)
 
 #ifdef CONFIG_SYSCTL
 /* Log invalid packets of a given protocol */
-static int log_invalid_proto_min = 0;
-static int log_invalid_proto_max = 255;
+static int log_invalid_proto_min __read_mostly;
+static int log_invalid_proto_max __read_mostly = 255;
+
+/* size the user *wants to set */
+static unsigned int nf_conntrack_htable_size_user __read_mostly;
+
+static int
+nf_conntrack_hash_sysctl(struct ctl_table *table, int write,
+			 void __user *buffer, size_t *lenp, loff_t *ppos)
+{
+	int ret;
+
+	ret = proc_dointvec(table, write, buffer, lenp, ppos);
+	if (ret < 0 || !write)
+		return ret;
+
+	/* update ret, we might not be able to satisfy request */
+	ret = nf_conntrack_hash_resize(nf_conntrack_htable_size_user);
+
+	/* update it to the actual value used by conntrack */
+	nf_conntrack_htable_size_user = nf_conntrack_htable_size;
+	return ret;
+}
 
 static struct ctl_table_header *nf_ct_netfilter_header;
 
@@ -456,10 +477,10 @@ static struct ctl_table nf_ct_sysctl_table[] = {
 	},
 	{
 		.procname       = "nf_conntrack_buckets",
-		.data           = &nf_conntrack_htable_size,
+		.data           = &nf_conntrack_htable_size_user,
 		.maxlen         = sizeof(unsigned int),
-		.mode           = 0444,
-		.proc_handler   = proc_dointvec,
+		.mode           = 0644,
+		.proc_handler   = nf_conntrack_hash_sysctl,
 	},
 	{
 		.procname	= "nf_conntrack_checksum",
@@ -517,6 +538,9 @@ static int nf_conntrack_standalone_init_sysctl(struct net *net)
 	if (net->user_ns != &init_user_ns)
 		table[0].procname = NULL;
 
+	if (!net_eq(&init_net, net))
+		table[2].mode = 0444;
+
 	net->ct.sysctl_header = register_net_sysctl(net, "net/netfilter", table);
 	if (!net->ct.sysctl_header)
 		goto out_unregister_netfilter;
@@ -606,6 +630,8 @@ static int __init nf_conntrack_standalone_init(void)
 		ret = -ENOMEM;
 		goto out_sysctl;
 	}
+
+	nf_conntrack_htable_size_user = nf_conntrack_htable_size;
 #endif
 
 	ret = register_pernet_subsys(&nf_conntrack_net_ops);
Signature · 1 line
2.7.3
maintainer Applied
#1 lore raw
Quoted 10 lines · Florian Westphal

No need to restrict this to module parameter.

We export a copy of the real hash size -- when user alters the value we allocate the new table, copy entries etc before we update the real size to the requested one.

This is also needed because the real size is used by concurrent readers and cannot be changed without synchronizing the conntrack generation seqcnt.

Quoted 7 lines

We only allow changing this value from the initial net namespace.

Tested using http-client-benchmark vs. httpterm with concurrent

while true;do
echo $RANDOM > /proc/sys/net/netfilter/nf_conntrack_buckets
done

Applied, thanks Florian.