1 files changed, 80 insertions, 59 deletions

diff --git a/net/ipv4/tcp_metrics.c b/net/ipv4/tcp_metrics.c
index 82f4575f9cd9..45b6ecd16412 100644
--- a/net/ipv4/tcp_metrics.c
+++ b/net/ipv4/tcp_metrics.c
@@ -40,7 +40,7 @@ struct tcp_fastopen_metrics {
 
 struct tcp_metrics_block {
 	struct tcp_metrics_block __rcu	*tcpm_next;
-	possible_net_t			tcpm_net;
+	struct net			*tcpm_net;
 	struct inetpeer_addr		tcpm_saddr;
 	struct inetpeer_addr		tcpm_daddr;
 	unsigned long			tcpm_stamp;
@@ -51,34 +51,38 @@ struct tcp_metrics_block {
 	struct rcu_head			rcu_head;
 };
 
-static inline struct net *tm_net(struct tcp_metrics_block *tm)
+static inline struct net *tm_net(const struct tcp_metrics_block *tm)
 {
-	return read_pnet(&tm->tcpm_net);
+	/* Paired with the WRITE_ONCE() in tcpm_new() */
+	return READ_ONCE(tm->tcpm_net);
 }
 
 static bool tcp_metric_locked(struct tcp_metrics_block *tm,
 			      enum tcp_metric_index idx)
 {
-	return tm->tcpm_lock & (1 << idx);
+	/* Paired with WRITE_ONCE() in tcpm_suck_dst() */
+	return READ_ONCE(tm->tcpm_lock) & (1 << idx);
 }
 
-static u32 tcp_metric_get(struct tcp_metrics_block *tm,
+static u32 tcp_metric_get(const struct tcp_metrics_block *tm,
 			  enum tcp_metric_index idx)
 {
-	return tm->tcpm_vals[idx];
+	/* Paired with WRITE_ONCE() in tcp_metric_set() */
+	return READ_ONCE(tm->tcpm_vals[idx]);
 }
 
 static void tcp_metric_set(struct tcp_metrics_block *tm,
 			   enum tcp_metric_index idx,
 			   u32 val)
 {
-	tm->tcpm_vals[idx] = val;
+	/* Paired with READ_ONCE() in tcp_metric_get() */
+	WRITE_ONCE(tm->tcpm_vals[idx], val);
 }
 
 static bool addr_same(const struct inetpeer_addr *a,
 		      const struct inetpeer_addr *b)
 {
-	return inetpeer_addr_cmp(a, b) == 0;
+	return (a->family == b->family) && !inetpeer_addr_cmp(a, b);
 }
 
 struct tcpm_hash_bucket {
@@ -89,6 +93,7 @@ static struct tcpm_hash_bucket	*tcp_metrics_hash __read_mostly;
 static unsigned int		tcp_metrics_hash_log __read_mostly;
 
 static DEFINE_SPINLOCK(tcp_metrics_lock);
+static DEFINE_SEQLOCK(fastopen_seqlock);
 
 static void tcpm_suck_dst(struct tcp_metrics_block *tm,
 			  const struct dst_entry *dst,
@@ -97,7 +102,7 @@ static void tcpm_suck_dst(struct tcp_metrics_block *tm,
 	u32 msval;
 	u32 val;
 
-	tm->tcpm_stamp = jiffies;
+	WRITE_ONCE(tm->tcpm_stamp, jiffies);
 
 	val = 0;
 	if (dst_metric_locked(dst, RTAX_RTT))
@@ -110,30 +115,42 @@ static void tcpm_suck_dst(struct tcp_metrics_block *tm,
 		val |= 1 << TCP_METRIC_CWND;
 	if (dst_metric_locked(dst, RTAX_REORDERING))
 		val |= 1 << TCP_METRIC_REORDERING;
-	tm->tcpm_lock = val;
+	/* Paired with READ_ONCE() in tcp_metric_locked() */
+	WRITE_ONCE(tm->tcpm_lock, val);
 
 	msval = dst_metric_raw(dst, RTAX_RTT);
-	tm->tcpm_vals[TCP_METRIC_RTT] = msval * USEC_PER_MSEC;
+	tcp_metric_set(tm, TCP_METRIC_RTT, msval * USEC_PER_MSEC);
 
 	msval = dst_metric_raw(dst, RTAX_RTTVAR);
-	tm->tcpm_vals[TCP_METRIC_RTTVAR] = msval * USEC_PER_MSEC;
-	tm->tcpm_vals[TCP_METRIC_SSTHRESH] = dst_metric_raw(dst, RTAX_SSTHRESH);
-	tm->tcpm_vals[TCP_METRIC_CWND] = dst_metric_raw(dst, RTAX_CWND);
-	tm->tcpm_vals[TCP_METRIC_REORDERING] = dst_metric_raw(dst, RTAX_REORDERING);
+	tcp_metric_set(tm, TCP_METRIC_RTTVAR, msval * USEC_PER_MSEC);
+	tcp_metric_set(tm, TCP_METRIC_SSTHRESH,
+		       dst_metric_raw(dst, RTAX_SSTHRESH));
+	tcp_metric_set(tm, TCP_METRIC_CWND,
+		       dst_metric_raw(dst, RTAX_CWND));
+	tcp_metric_set(tm, TCP_METRIC_REORDERING,
+		       dst_metric_raw(dst, RTAX_REORDERING));
 	if (fastopen_clear) {
+		write_seqlock(&fastopen_seqlock);
 		tm->tcpm_fastopen.mss = 0;
 		tm->tcpm_fastopen.syn_loss = 0;
 		tm->tcpm_fastopen.try_exp = 0;
 		tm->tcpm_fastopen.cookie.exp = false;
 		tm->tcpm_fastopen.cookie.len = 0;
+		write_sequnlock(&fastopen_seqlock);
 	}
 }
 
 #define TCP_METRICS_TIMEOUT		(60 * 60 * HZ)
 
-static void tcpm_check_stamp(struct tcp_metrics_block *tm, struct dst_entry *dst)
+static void tcpm_check_stamp(struct tcp_metrics_block *tm,
+			     const struct dst_entry *dst)
 {
-	if (tm && unlikely(time_after(jiffies, tm->tcpm_stamp + TCP_METRICS_TIMEOUT)))
+	unsigned long limit;
+
+	if (!tm)
+		return;
+	limit = READ_ONCE(tm->tcpm_stamp) + TCP_METRICS_TIMEOUT;
+	if (unlikely(time_after(jiffies, limit)))
 		tcpm_suck_dst(tm, dst, false);
 }
 
@@ -149,11 +166,11 @@ static struct tcp_metrics_block *tcpm_new(struct dst_entry *dst,
 					  unsigned int hash)
 {
 	struct tcp_metrics_block *tm;
-	struct net *net;
 	bool reclaim = false;
+	struct net *net;
 
 	spin_lock_bh(&tcp_metrics_lock);
-	net = dev_net(dst->dev);
+	net = dst_dev_net_rcu(dst);
 
 	/* While waiting for the spin-lock the cache might have been populated
 	 * with this entry and so we have to check again.
@@ -174,20 +191,23 @@ static struct tcp_metrics_block *tcpm_new(struct dst_entry *dst,
 		oldest = deref_locked(tcp_metrics_hash[hash].chain);
 		for (tm = deref_locked(oldest->tcpm_next); tm;
 		     tm = deref_locked(tm->tcpm_next)) {
-			if (time_before(tm->tcpm_stamp, oldest->tcpm_stamp))
+			if (time_before(READ_ONCE(tm->tcpm_stamp),
+					READ_ONCE(oldest->tcpm_stamp)))
 				oldest = tm;
 		}
 		tm = oldest;
 	} else {
-		tm = kmalloc(sizeof(*tm), GFP_ATOMIC);
+		tm = kzalloc(sizeof(*tm), GFP_ATOMIC);
 		if (!tm)
 			goto out_unlock;
 	}
-	write_pnet(&tm->tcpm_net, net);
+	/* Paired with the READ_ONCE() in tm_net() */
+	WRITE_ONCE(tm->tcpm_net, net);
+
 	tm->tcpm_saddr = *saddr;
 	tm->tcpm_daddr = *daddr;
 
-	tcpm_suck_dst(tm, dst, true);
+	tcpm_suck_dst(tm, dst, reclaim);
 
 	if (likely(!reclaim)) {
 		tm->tcpm_next = tcp_metrics_hash[hash].chain;
@@ -253,7 +273,7 @@ static struct tcp_metrics_block *__tcp_get_metrics_req(struct request_sock *req,
 		return NULL;
 	}
 
-	net = dev_net(dst->dev);
+	net = dst_dev_net_rcu(dst);
 	hash ^= net_hash_mix(net);
 	hash = hash_32(hash, tcp_metrics_hash_log);
 
@@ -298,7 +318,7 @@ static struct tcp_metrics_block *tcp_get_metrics(struct sock *sk,
 	else
 		return NULL;
 
-	net = dev_net(dst->dev);
+	net = dst_dev_net_rcu(dst);
 	hash ^= net_hash_mix(net);
 	hash = hash_32(hash, tcp_metrics_hash_log);
 
@@ -434,7 +454,7 @@ void tcp_update_metrics(struct sock *sk)
 					       tp->reordering);
 		}
 	}
-	tm->tcpm_stamp = jiffies;
+	WRITE_ONCE(tm->tcpm_stamp, jiffies);
 out_unlock:
 	rcu_read_unlock();
 }
@@ -450,11 +470,15 @@ void tcp_init_metrics(struct sock *sk)
 	u32 val, crtt = 0; /* cached RTT scaled by 8 */
 
 	sk_dst_confirm(sk);
+	/* ssthresh may have been reduced unnecessarily during.
+	 * 3WHS. Restore it back to its initial default.
+	 */
+	tp->snd_ssthresh = TCP_INFINITE_SSTHRESH;
 	if (!dst)
 		goto reset;
 
 	rcu_read_lock();
-	tm = tcp_get_metrics(sk, dst, true);
+	tm = tcp_get_metrics(sk, dst, false);
 	if (!tm) {
 		rcu_read_unlock();
 		goto reset;
@@ -469,11 +493,6 @@ void tcp_init_metrics(struct sock *sk)
 		tp->snd_ssthresh = val;
 		if (tp->snd_ssthresh > tp->snd_cwnd_clamp)
 			tp->snd_ssthresh = tp->snd_cwnd_clamp;
-	} else {
-		/* ssthresh may have been reduced unnecessarily during.
-		 * 3WHS. Restore it back to its initial default.
-		 */
-		tp->snd_ssthresh = TCP_INFINITE_SSTHRESH;
 	}
 	val = tcp_metric_get(tm, TCP_METRIC_REORDERING);
 	if (val && tp->reordering != val)
@@ -539,8 +558,6 @@ bool tcp_peer_is_proven(struct request_sock *req, struct dst_entry *dst)
 	return ret;
 }
 
-static DEFINE_SEQLOCK(fastopen_seqlock);
-
 void tcp_fastopen_cache_get(struct sock *sk, u16 *mss,
 			    struct tcp_fastopen_cookie *cookie)
 {
@@ -600,8 +617,13 @@ static struct genl_family tcp_metrics_nl_family;
 
 static const struct nla_policy tcp_metrics_nl_policy[TCP_METRICS_ATTR_MAX + 1] = {
 	[TCP_METRICS_ATTR_ADDR_IPV4]	= { .type = NLA_U32, },
-	[TCP_METRICS_ATTR_ADDR_IPV6]	= { .type = NLA_BINARY,
-					    .len = sizeof(struct in6_addr), },
+	[TCP_METRICS_ATTR_ADDR_IPV6]	=
+		NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
+
+	[TCP_METRICS_ATTR_SADDR_IPV4]	= { .type = NLA_U32, },
+	[TCP_METRICS_ATTR_SADDR_IPV6]	=
+		NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
+
 	/* Following attributes are not received for GET/DEL,
 	 * we keep them for reference
 	 */
@@ -647,7 +669,7 @@ static int tcp_metrics_fill_info(struct sk_buff *msg,
 	}
 
 	if (nla_put_msecs(msg, TCP_METRICS_ATTR_AGE,
-			  jiffies - tm->tcpm_stamp,
+			  jiffies - READ_ONCE(tm->tcpm_stamp),
 			  TCP_METRICS_ATTR_PAD) < 0)
 		goto nla_put_failure;
 
@@ -658,7 +680,7 @@ static int tcp_metrics_fill_info(struct sk_buff *msg,
 		if (!nest)
 			goto nla_put_failure;
 		for (i = 0; i < TCP_METRIC_MAX_KERNEL + 1; i++) {
-			u32 val = tm->tcpm_vals[i];
+			u32 val = tcp_metric_get(tm, i);
 
 			if (!val)
 				continue;
@@ -749,6 +771,7 @@ static int tcp_metrics_nl_dump(struct sk_buff *skb,
 	unsigned int max_rows = 1U << tcp_metrics_hash_log;
 	unsigned int row, s_row = cb->args[0];
 	int s_col = cb->args[1], col = s_col;
+	int res = 0;
 
 	for (row = s_row; row < max_rows; row++, s_col = 0) {
 		struct tcp_metrics_block *tm;
@@ -761,7 +784,8 @@ static int tcp_metrics_nl_dump(struct sk_buff *skb,
 				continue;
 			if (col < s_col)
 				continue;
-			if (tcp_metrics_dump_info(skb, cb, tm) < 0) {
+			res = tcp_metrics_dump_info(skb, cb, tm);
+			if (res < 0) {
 				rcu_read_unlock();
 				goto done;
 			}
@@ -772,7 +796,7 @@ static int tcp_metrics_nl_dump(struct sk_buff *skb,
 done:
 	cb->args[0] = row;
 	cb->args[1] = col;
-	return skb->len;
+	return res;
 }
 
 static int __parse_nl_addr(struct genl_info *info, struct inetpeer_addr *addr,
@@ -791,8 +815,6 @@ static int __parse_nl_addr(struct genl_info *info, struct inetpeer_addr *addr,
 	if (a) {
 		struct in6_addr in6;
 
-		if (nla_len(a) != sizeof(struct in6_addr))
-			return -EINVAL;
 		in6 = nla_get_in6_addr(a);
 		inetpeer_set_addr_v6(addr, &in6);
 		if (hash)
@@ -881,22 +903,25 @@ static void tcp_metrics_flush_all(struct net *net)
 	unsigned int row;
 
 	for (row = 0; row < max_rows; row++, hb++) {
-		struct tcp_metrics_block __rcu **pp;
+		struct tcp_metrics_block __rcu **pp = &hb->chain;
 		bool match;
 
+		if (!rcu_access_pointer(*pp))
+			continue;
+
 		spin_lock_bh(&tcp_metrics_lock);
-		pp = &hb->chain;
 		for (tm = deref_locked(*pp); tm; tm = deref_locked(*pp)) {
 			match = net ? net_eq(tm_net(tm), net) :
-				!refcount_read(&tm_net(tm)->ns.count);
+				!check_net(tm_net(tm));
 			if (match) {
-				*pp = tm->tcpm_next;
+				rcu_assign_pointer(*pp, tm->tcpm_next);
 				kfree_rcu(tm, rcu_head);
 			} else {
 				pp = &tm->tcpm_next;
 			}
 		}
 		spin_unlock_bh(&tcp_metrics_lock);
+		cond_resched();
 	}
 }
 
@@ -931,7 +956,7 @@ static int tcp_metrics_nl_cmd_del(struct sk_buff *skb, struct genl_info *info)
 		if (addr_same(&tm->tcpm_daddr, &daddr) &&
 		    (!src || addr_same(&tm->tcpm_saddr, &saddr)) &&
 		    net_eq(tm_net(tm), net)) {
-			*pp = tm->tcpm_next;
+			rcu_assign_pointer(*pp, tm->tcpm_next);
 			kfree_rcu(tm, rcu_head);
 			found = true;
 		} else {
@@ -966,13 +991,14 @@ static struct genl_family tcp_metrics_nl_family __ro_after_init = {
 	.maxattr	= TCP_METRICS_ATTR_MAX,
 	.policy = tcp_metrics_nl_policy,
 	.netnsok	= true,
+	.parallel_ops	= true,
 	.module		= THIS_MODULE,
 	.small_ops	= tcp_metrics_nl_ops,
 	.n_small_ops	= ARRAY_SIZE(tcp_metrics_nl_ops),
 	.resv_start_op	= TCP_METRICS_CMD_DEL + 1,
 };
 
-static unsigned int tcpmhash_entries;
+static unsigned int tcpmhash_entries __initdata;
 static int __init set_tcpmhash_entries(char *str)
 {
 	ssize_t ret;
@@ -988,15 +1014,11 @@ static int __init set_tcpmhash_entries(char *str)
 }
 __setup("tcpmhash_entries=", set_tcpmhash_entries);
 
-static int __net_init tcp_net_metrics_init(struct net *net)
+static void __init tcp_metrics_hash_alloc(void)
 {
+	unsigned int slots = tcpmhash_entries;
 	size_t size;
-	unsigned int slots;
-
-	if (!net_eq(net, &init_net))
-		return 0;
 
-	slots = tcpmhash_entries;
 	if (!slots) {
 		if (totalram_pages() >= 128 * 1024)
 			slots = 16 * 1024;
@@ -1009,9 +1031,7 @@ static int __net_init tcp_net_metrics_init(struct net *net)
 
 	tcp_metrics_hash = kvzalloc(size, GFP_KERNEL);
 	if (!tcp_metrics_hash)
-		return -ENOMEM;
-
-	return 0;
+		panic("Could not allocate the tcp_metrics hash table\n");
 }
 
 static void __net_exit tcp_net_metrics_exit_batch(struct list_head *net_exit_list)
@@ -1020,7 +1040,6 @@ static void __net_exit tcp_net_metrics_exit_batch(struct list_head *net_exit_lis
 }
 
 static __net_initdata struct pernet_operations tcp_net_metrics_ops = {
-	.init		=	tcp_net_metrics_init,
 	.exit_batch	=	tcp_net_metrics_exit_batch,
 };
 
@@ -1028,9 +1047,11 @@ void __init tcp_metrics_init(void)
 {
 	int ret;
 
+	tcp_metrics_hash_alloc();
+
 	ret = register_pernet_subsys(&tcp_net_metrics_ops);
 	if (ret < 0)
-		panic("Could not allocate the tcp_metrics hash table\n");
+		panic("Could not register tcp_net_metrics_ops\n");
 
 	ret = genl_register_family(&tcp_metrics_nl_family);
 	if (ret < 0)