[NETFILTER] early_drop() imrovement (v3)

[Vasily Averin <vvs@sw.ru>]

When the number of conntracks is reached nf_conntrack_max limit, early_drop()
tries to free one of already used conntracks. If it does not find any conntracks
that may be freed, it leads to transmission errors.
In current implementation the conntracks are searched in one hash bucket only.
It have some drawbacks: if used hash bucket is empty we have not any chances to
find something. On the other hand the hash bucket can contain a huge number of
conntracks and its check can last a long time.
The proposed patch limits the number of checked conntracks by default number of
conntracks in one hash bucket (NF_CT_PER_BUCKET) and allows to search conntracks
in other hash buckets. As result in any case the search will have the same
chances to free one of the conntracks and the check will not lead to long delays.

Signed-off-by:	Vasily Averin <vvs@sw.ru>

diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_conntrack_core.c
index e132c8a..d984bce 100644
--- a/net/netfilter/nf_conntrack_core.c
+++ b/net/netfilter/nf_conntrack_core.c
@@ -76,6 +76,8 @@ static unsigned int nf_conntrack_next_id;
 DEFINE_PER_CPU(struct ip_conntrack_stat, nf_conntrack_stat);
 EXPORT_PER_CPU_SYMBOL(nf_conntrack_stat);

+#define NF_CT_PER_BUCKET	8U
+
 /*
  * This scheme offers various size of "struct nf_conn" dependent on
  * features(helper, nat, ...)
@@ -525,7 +527,7 @@ EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken);

 /* There's a small race here where we may free a just-assured
    connection.  Too bad: we're in trouble anyway. */
-static int early_drop(struct list_head *chain)
+static int __early_drop(struct list_head *chain, unsigned int *cnt)
 {
 	/* Traverse backwards: gives us oldest, which is roughly LRU */
 	struct nf_conntrack_tuple_hash *h;
@@ -540,6 +542,8 @@ static int early_drop(struct list_head *chain)
 			atomic_inc(&ct->ct_general.use);
 			break;
 		}
+		if (!--(*cnt))
+			break;
 	}
 	read_unlock_bh(&nf_conntrack_lock);

@@ -555,6 +559,21 @@ static int early_drop(struct list_head *chain)
 	return dropped;
 }

+static int early_drop(const struct nf_conntrack_tuple *orig)
+{
+	unsigned int i, hash, cnt;
+	int ret = 0;
+
+	hash = hash_conntrack(orig);
+	cnt = NF_CT_PER_BUCKET;
+
+	for (i = 0;
+		!ret && cnt && i < nf_conntrack_htable_size;
+			++i, hash = ++hash % nf_conntrack_htable_size)
+		ret = __early_drop(&nf_conntrack_hash[hash], &cnt);
+	return ret;
+}
+
 static struct nf_conn *
 __nf_conntrack_alloc(const struct nf_conntrack_tuple *orig,
 		     const struct nf_conntrack_tuple *repl,
@@ -574,9 +593,7 @@ __nf_conntrack_alloc(const struct nf_conntrack_tuple *orig,

 	if (nf_conntrack_max
 	    && atomic_read(&nf_conntrack_count) > nf_conntrack_max) {
-		unsigned int hash = hash_conntrack(orig);
-		/* Try dropping from this hash chain. */
-		if (!early_drop(&nf_conntrack_hash[hash])) {
+		if (!early_drop(orig)) {
 			atomic_dec(&nf_conntrack_count);
 			if (net_ratelimit())
 				printk(KERN_WARNING
@@ -1226,7 +1243,7 @@ int __init nf_conntrack_init(void)
 		if (nf_conntrack_htable_size < 16)
 			nf_conntrack_htable_size = 16;
 	}
-	nf_conntrack_max = 8 * nf_conntrack_htable_size;
+	nf_conntrack_max = NF_CT_PER_BUCKET * nf_conntrack_htable_size;

 	printk("nf_conntrack version %s (%u buckets, %d max)\n",
 	       NF_CONNTRACK_VERSION, nf_conntrack_htable_size,


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/