1 files changed, 104 insertions, 88 deletions

diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_conntrack_core.c
index 96ba19fc8155..94e18fb9690d 100644
--- a/net/netfilter/nf_conntrack_core.c
+++ b/net/netfilter/nf_conntrack_core.c
@@ -21,7 +21,6 @@
 #include <linux/stddef.h>
 #include <linux/slab.h>
 #include <linux/random.h>
-#include <linux/jhash.h>
 #include <linux/siphash.h>
 #include <linux/err.h>
 #include <linux/percpu.h>
@@ -66,22 +65,19 @@ EXPORT_SYMBOL_GPL(nf_conntrack_hash);
 
 struct conntrack_gc_work {
 	struct delayed_work	dwork;
-	u32			last_bucket;
+	u32			next_bucket;
 	bool			exiting;
 	bool			early_drop;
-	long			next_gc_run;
 };
 
 static __read_mostly struct kmem_cache *nf_conntrack_cachep;
 static DEFINE_SPINLOCK(nf_conntrack_locks_all_lock);
 static __read_mostly bool nf_conntrack_locks_all;
 
-/* every gc cycle scans at most 1/GC_MAX_BUCKETS_DIV part of table */
-#define GC_MAX_BUCKETS_DIV	128u
-/* upper bound of full table scan */
-#define GC_MAX_SCAN_JIFFIES	(16u * HZ)
-/* desired ratio of entries found to be expired */
-#define GC_EVICT_RATIO	50u
+#define GC_SCAN_INTERVAL	(120u * HZ)
+#define GC_SCAN_MAX_DURATION	msecs_to_jiffies(10)
+
+#define MAX_CHAINLEN	64u
 
 static struct conntrack_gc_work conntrack_gc_work;
 
@@ -149,7 +145,15 @@ static void nf_conntrack_all_lock(void)
 
 	spin_lock(&nf_conntrack_locks_all_lock);
 
-	nf_conntrack_locks_all = true;
+	/* For nf_contrack_locks_all, only the latest time when another
+	 * CPU will see an update is controlled, by the "release" of the
+	 * spin_lock below.
+	 * The earliest time is not controlled, an thus KCSAN could detect
+	 * a race when nf_conntract_lock() reads the variable.
+	 * WRITE_ONCE() is used to ensure the compiler will not
+	 * optimize the write.
+	 */
+	WRITE_ONCE(nf_conntrack_locks_all, true);
 
 	for (i = 0; i < CONNTRACK_LOCKS; i++) {
 		spin_lock(&nf_conntrack_locks[i]);
@@ -181,25 +185,31 @@ EXPORT_SYMBOL_GPL(nf_conntrack_htable_size);
 unsigned int nf_conntrack_max __read_mostly;
 EXPORT_SYMBOL_GPL(nf_conntrack_max);
 seqcount_spinlock_t nf_conntrack_generation __read_mostly;
-static unsigned int nf_conntrack_hash_rnd __read_mostly;
+static siphash_key_t nf_conntrack_hash_rnd __read_mostly;
 
 static u32 hash_conntrack_raw(const struct nf_conntrack_tuple *tuple,
 			      const struct net *net)
 {
-	unsigned int n;
-	u32 seed;
+	struct {
+		struct nf_conntrack_man src;
+		union nf_inet_addr dst_addr;
+		u32 net_mix;
+		u16 dport;
+		u16 proto;
+	} __aligned(SIPHASH_ALIGNMENT) combined;
 
 	get_random_once(&nf_conntrack_hash_rnd, sizeof(nf_conntrack_hash_rnd));
 
-	/* The direction must be ignored, so we hash everything up to the
-	 * destination ports (which is a multiple of 4) and treat the last
-	 * three bytes manually.
-	 */
-	seed = nf_conntrack_hash_rnd ^ net_hash_mix(net);
-	n = (sizeof(tuple->src) + sizeof(tuple->dst.u3)) / sizeof(u32);
-	return jhash2((u32 *)tuple, n, seed ^
-		      (((__force __u16)tuple->dst.u.all << 16) |
-		      tuple->dst.protonum));
+	memset(&combined, 0, sizeof(combined));
+
+	/* The direction must be ignored, so handle usable members manually. */
+	combined.src = tuple->src;
+	combined.dst_addr = tuple->dst.u3;
+	combined.net_mix = net_hash_mix(net);
+	combined.dport = (__force __u16)tuple->dst.u.all;
+	combined.proto = tuple->dst.protonum;
+
+	return (u32)siphash(&combined, sizeof(combined), &nf_conntrack_hash_rnd);
 }
 
 static u32 scale_hash(u32 hash)
@@ -662,8 +672,13 @@ bool nf_ct_delete(struct nf_conn *ct, u32 portid, int report)
 		return false;
 
 	tstamp = nf_conn_tstamp_find(ct);
-	if (tstamp && tstamp->stop == 0)
+	if (tstamp) {
+		s32 timeout = ct->timeout - nfct_time_stamp;
+
 		tstamp->stop = ktime_get_real_ns();
+		if (timeout < 0)
+			tstamp->stop -= jiffies_to_nsecs(-timeout);
+	}
 
 	if (nf_conntrack_event_report(IPCT_DESTROY, ct,
 				    portid, report) < 0) {
@@ -827,7 +842,9 @@ nf_conntrack_hash_check_insert(struct nf_conn *ct)
 	unsigned int hash, reply_hash;
 	struct nf_conntrack_tuple_hash *h;
 	struct hlist_nulls_node *n;
+	unsigned int chainlen = 0;
 	unsigned int sequence;
+	int err = -EEXIST;
 
 	zone = nf_ct_zone(ct);
 
@@ -841,15 +858,24 @@ nf_conntrack_hash_check_insert(struct nf_conn *ct)
 	} while (nf_conntrack_double_lock(net, hash, reply_hash, sequence));
 
 	/* See if there's one in the list already, including reverse */
-	hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[hash], hnnode)
+	hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[hash], hnnode) {
 		if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
 				    zone, net))
 			goto out;
 
-	hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[reply_hash], hnnode)
+		if (chainlen++ > MAX_CHAINLEN)
+			goto chaintoolong;
+	}
+
+	chainlen = 0;
+
+	hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[reply_hash], hnnode) {
 		if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_REPLY].tuple,
 				    zone, net))
 			goto out;
+		if (chainlen++ > MAX_CHAINLEN)
+			goto chaintoolong;
+	}
 
 	smp_wmb();
 	/* The caller holds a reference to this object */
@@ -859,11 +885,13 @@ nf_conntrack_hash_check_insert(struct nf_conn *ct)
 	NF_CT_STAT_INC(net, insert);
 	local_bh_enable();
 	return 0;
-
+chaintoolong:
+	NF_CT_STAT_INC(net, chaintoolong);
+	err = -ENOSPC;
 out:
 	nf_conntrack_double_unlock(hash, reply_hash);
 	local_bh_enable();
-	return -EEXIST;
+	return err;
 }
 EXPORT_SYMBOL_GPL(nf_conntrack_hash_check_insert);
 
@@ -1076,6 +1104,7 @@ int
 __nf_conntrack_confirm(struct sk_buff *skb)
 {
 	const struct nf_conntrack_zone *zone;
+	unsigned int chainlen = 0, sequence;
 	unsigned int hash, reply_hash;
 	struct nf_conntrack_tuple_hash *h;
 	struct nf_conn *ct;
@@ -1083,7 +1112,6 @@ __nf_conntrack_confirm(struct sk_buff *skb)
 	struct hlist_nulls_node *n;
 	enum ip_conntrack_info ctinfo;
 	struct net *net;
-	unsigned int sequence;
 	int ret = NF_DROP;
 
 	ct = nf_ct_get(skb, &ctinfo);
@@ -1143,15 +1171,28 @@ __nf_conntrack_confirm(struct sk_buff *skb)
 	/* See if there's one in the list already, including reverse:
 	   NAT could have grabbed it without realizing, since we're
 	   not in the hash.  If there is, we lost race. */
-	hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[hash], hnnode)
+	hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[hash], hnnode) {
 		if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
 				    zone, net))
 			goto out;
+		if (chainlen++ > MAX_CHAINLEN)
+			goto chaintoolong;
+	}
 
-	hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[reply_hash], hnnode)
+	chainlen = 0;
+	hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[reply_hash], hnnode) {
 		if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_REPLY].tuple,
 				    zone, net))
 			goto out;
+		if (chainlen++ > MAX_CHAINLEN) {
+chaintoolong:
+			nf_ct_add_to_dying_list(ct);
+			NF_CT_STAT_INC(net, chaintoolong);
+			NF_CT_STAT_INC(net, insert_failed);
+			ret = NF_DROP;
+			goto dying;
+		}
+	}
 
 	/* Timer relative to confirmation time, not original
 	   setting time, otherwise we'd get timer wrap in
@@ -1350,17 +1391,13 @@ static bool gc_worker_can_early_drop(const struct nf_conn *ct)
 
 static void gc_worker(struct work_struct *work)
 {
-	unsigned int min_interval = max(HZ / GC_MAX_BUCKETS_DIV, 1u);
-	unsigned int i, goal, buckets = 0, expired_count = 0;
-	unsigned int nf_conntrack_max95 = 0;
+	unsigned long end_time = jiffies + GC_SCAN_MAX_DURATION;
+	unsigned int i, hashsz, nf_conntrack_max95 = 0;
+	unsigned long next_run = GC_SCAN_INTERVAL;
 	struct conntrack_gc_work *gc_work;
-	unsigned int ratio, scanned = 0;
-	unsigned long next_run;
-
 	gc_work = container_of(work, struct conntrack_gc_work, dwork.work);
 
-	goal = nf_conntrack_htable_size / GC_MAX_BUCKETS_DIV;
-	i = gc_work->last_bucket;
+	i = gc_work->next_bucket;
 	if (gc_work->early_drop)
 		nf_conntrack_max95 = nf_conntrack_max / 100u * 95u;
 
@@ -1368,15 +1405,15 @@ static void gc_worker(struct work_struct *work)
 		struct nf_conntrack_tuple_hash *h;
 		struct hlist_nulls_head *ct_hash;
 		struct hlist_nulls_node *n;
-		unsigned int hashsz;
 		struct nf_conn *tmp;
 
-		i++;
 		rcu_read_lock();
 
 		nf_conntrack_get_ht(&ct_hash, &hashsz);
-		if (i >= hashsz)
-			i = 0;
+		if (i >= hashsz) {
+			rcu_read_unlock();
+			break;
+		}
 
 		hlist_nulls_for_each_entry_rcu(h, n, &ct_hash[i], hnnode) {
 			struct nf_conntrack_net *cnet;
@@ -1384,7 +1421,6 @@ static void gc_worker(struct work_struct *work)
 
 			tmp = nf_ct_tuplehash_to_ctrack(h);
 
-			scanned++;
 			if (test_bit(IPS_OFFLOAD_BIT, &tmp->status)) {
 				nf_ct_offload_timeout(tmp);
 				continue;
@@ -1392,7 +1428,6 @@ static void gc_worker(struct work_struct *work)
 
 			if (nf_ct_is_expired(tmp)) {
 				nf_ct_gc_expired(tmp);
-				expired_count++;
 				continue;
 			}
 
@@ -1425,7 +1460,14 @@ static void gc_worker(struct work_struct *work)
 		 */
 		rcu_read_unlock();
 		cond_resched();
-	} while (++buckets < goal);
+		i++;
+
+		if (time_after(jiffies, end_time) && i < hashsz) {
+			gc_work->next_bucket = i;
+			next_run = 0;
+			break;
+		}
+	} while (i < hashsz);
 
 	if (gc_work->exiting)
 		return;
@@ -1436,40 +1478,17 @@ static void gc_worker(struct work_struct *work)
 	 *
 	 * This worker is only here to reap expired entries when system went
 	 * idle after a busy period.
-	 *
-	 * The heuristics below are supposed to balance conflicting goals:
-	 *
-	 * 1. Minimize time until we notice a stale entry
-	 * 2. Maximize scan intervals to not waste cycles
-	 *
-	 * Normally, expire ratio will be close to 0.
-	 *
-	 * As soon as a sizeable fraction of the entries have expired
-	 * increase scan frequency.
 	 */
-	ratio = scanned ? expired_count * 100 / scanned : 0;
-	if (ratio > GC_EVICT_RATIO) {
-		gc_work->next_gc_run = min_interval;
-	} else {
-		unsigned int max = GC_MAX_SCAN_JIFFIES / GC_MAX_BUCKETS_DIV;
-
-		BUILD_BUG_ON((GC_MAX_SCAN_JIFFIES / GC_MAX_BUCKETS_DIV) == 0);
-
-		gc_work->next_gc_run += min_interval;
-		if (gc_work->next_gc_run > max)
-			gc_work->next_gc_run = max;
+	if (next_run) {
+		gc_work->early_drop = false;
+		gc_work->next_bucket = 0;
 	}
-
-	next_run = gc_work->next_gc_run;
-	gc_work->last_bucket = i;
-	gc_work->early_drop = false;
 	queue_delayed_work(system_power_efficient_wq, &gc_work->dwork, next_run);
 }
 
 static void conntrack_gc_work_init(struct conntrack_gc_work *gc_work)
 {
 	INIT_DEFERRABLE_WORK(&gc_work->dwork, gc_worker);
-	gc_work->next_gc_run = HZ;
 	gc_work->exiting = false;
 }
 
@@ -2457,7 +2476,6 @@ i_see_dead_people:
 	}
 
 	list_for_each_entry(net, net_exit_list, exit_list) {
-		nf_conntrack_proto_pernet_fini(net);
 		nf_conntrack_ecache_pernet_fini(net);
 		nf_conntrack_expect_pernet_fini(net);
 		free_percpu(net->ct.stat);
@@ -2609,26 +2627,24 @@ int nf_conntrack_init_start(void)
 		spin_lock_init(&nf_conntrack_locks[i]);
 
 	if (!nf_conntrack_htable_size) {
-		/* Idea from tcp.c: use 1/16384 of memory.
-		 * On i386: 32MB machine has 512 buckets.
-		 * >= 1GB machines have 16384 buckets.
-		 * >= 4GB machines have 65536 buckets.
-		 */
 		nf_conntrack_htable_size
 			= (((nr_pages << PAGE_SHIFT) / 16384)
 			   / sizeof(struct hlist_head));
-		if (nr_pages > (4 * (1024 * 1024 * 1024 / PAGE_SIZE)))
-			nf_conntrack_htable_size = 65536;
+		if (BITS_PER_LONG >= 64 &&
+		    nr_pages > (4 * (1024 * 1024 * 1024 / PAGE_SIZE)))
+			nf_conntrack_htable_size = 262144;
 		else if (nr_pages > (1024 * 1024 * 1024 / PAGE_SIZE))
-			nf_conntrack_htable_size = 16384;
-		if (nf_conntrack_htable_size < 32)
-			nf_conntrack_htable_size = 32;
-
-		/* Use a max. factor of four by default to get the same max as
-		 * with the old struct list_heads. When a table size is given
-		 * we use the old value of 8 to avoid reducing the max.
-		 * entries. */
-		max_factor = 4;
+			nf_conntrack_htable_size = 65536;
+
+		if (nf_conntrack_htable_size < 1024)
+			nf_conntrack_htable_size = 1024;
+		/* Use a max. factor of one by default to keep the average
+		 * hash chain length at 2 entries.  Each entry has to be added
+		 * twice (once for original direction, once for reply).
+		 * When a table size is given we use the old value of 8 to
+		 * avoid implicit reduction of the max entries setting.
+		 */
+		max_factor = 1;
 	}
 
 	nf_conntrack_hash = nf_ct_alloc_hashtable(&nf_conntrack_htable_size, 1);