Skip to contentSkip to conversation

tcp: fix race condition when creating child sockets from syncookies

PATCH Superseded Ricardo Dias 2 participants 2 messages Nov 19, 2020

Subsystems (3)

v7 of 8 Superseded Latest is v8 →

lore
Share & tools
lore
#0 lore raw

PATCH v7 1/1 tcp: fix race condition when creating child sockets from syncookies

When the TCP stack is in SYN flood mode, the server child socket is created from the SYN cookie received in a TCP packet with the ACK flag set.

The child socket is created when the server receives the first TCP packet with a valid SYN cookie from the client. Usually, this packet corresponds to the final step of the TCP 3-way handshake, the ACK packet. But is also possible to receive a valid SYN cookie from the first TCP data packet sent by the client, and thus create a child socket from that SYN cookie.

Since a client socket is ready to send data as soon as it receives the SYN+ACK packet from the server, the client can send the ACK packet (sent by the TCP stack code), and the first data packet (sent by the userspace program) almost at the same time, and thus the server will equally receive the two TCP packets with valid SYN cookies almost at the same instant.

When such event happens, the TCP stack code has a race condition that occurs between the momement a lookup is done to the established connections hashtable to check for the existence of a connection for the same client, and the moment that the child socket is added to the established connections hashtable. As a consequence, this race condition can lead to a situation where we add two child sockets to the established connections hashtable and deliver two sockets to the userspace program to the same client.

This patch fixes the race condition by checking if an existing child socket exists for the same client when we are adding the second child socket to the established connections socket. If an existing child socket exists, we return that socket and use it to process the TCP packet received, and discard the second child socket to the same client.

---
v7 (2020-11-19):
* Changed the approach to re-use the first (existing) socket created from thge syncookie. Instead of returning the existing socket in tcp_(v4|v6)_syn_recv_sock and continue the protocol state machine execution, tcp_(v4|v6)_syn_recv_sock signals that already exists a socket, and tells tcp_(v4|v6)_rcv to lookup the socket again in the established connections table.
This new approach fixes the errors reported by Eric for the previous version of the patch.
* Also fixes the memory leaks by making sure that the newly created socket in syn_recv_sock is destroyed in case an already existing socket exists.

v6 (2020-11-17):
* Moved the ehash bucket list search for its own helper function.

v5 (2020-11-16):
- Not considered for review -

v4 (2020-11-12):
* Added struct sock **esk) parameter to inet_ehash_insert`.
* Fixed ref count increment in inet_ehash_insert.
* Fixed callers of inet_ehash_nolisten.

v3 (2020-11-11):
* Fixed IPv6 handling in inet_ehash_insert
* Removed unecessary comparison while traversing the ehash bucket list.

v2 (2020-11-09):
* Changed the author's email domain.
* Removed the helper function inet_ehash_insert_chk_dup and moved the logic to the existing inet_ehash_insert.
* Updated the callers of iner_ehash_nolisten to deal with the new logic.

 include/net/inet_hashtables.h   |  5 ++-
 include/net/tcp.h               |  9 +++--
 net/dccp/ipv4.c                 |  2 +-
 net/dccp/ipv6.c                 |  2 +-
 net/ipv4/inet_connection_sock.c |  2 +-
 net/ipv4/inet_hashtables.c      | 68 +++++++++++++++++++++++++++++----
 net/ipv4/syncookies.c           | 10 +++--
 net/ipv4/tcp_ipv4.c             | 41 +++++++++++++++++---
 net/ipv6/syncookies.c           |  5 ++-
 net/ipv6/tcp_ipv6.c             | 39 +++++++++++++++++--
 10 files changed, 153 insertions(+), 30 deletions(-)
include/net/inet_hashtables.h +3 -2
index 92560974ea67..ca6a3ea9057e 100644
@@ -247,8 +247,9 @@ void inet_hashinfo2_init(struct inet_hashinfo *h, const char *name,
 			 unsigned long high_limit);
 int inet_hashinfo2_init_mod(struct inet_hashinfo *h);
 
-bool inet_ehash_insert(struct sock *sk, struct sock *osk);
-bool inet_ehash_nolisten(struct sock *sk, struct sock *osk);
+bool inet_ehash_insert(struct sock *sk, struct sock *osk, bool *found_dup_sk);
+bool inet_ehash_nolisten(struct sock *sk, struct sock *osk,
+			 bool *found_dup_sk);
 int __inet_hash(struct sock *sk, struct sock *osk);
 int inet_hash(struct sock *sk);
 void inet_unhash(struct sock *sk);
include/net/tcp.h +6 -3
index eab6c7510b5b..ff144d3ead2f 100644
@@ -465,10 +465,12 @@ void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb);
 /* From syncookies.c */
 struct sock *tcp_get_cookie_sock(struct sock *sk, struct sk_buff *skb,
 				 struct request_sock *req,
-				 struct dst_entry *dst, u32 tsoff);
+				 struct dst_entry *dst, u32 tsoff,
+				 bool *found_dup_sk);
 int __cookie_v4_check(const struct iphdr *iph, const struct tcphdr *th,
 		      u32 cookie);
-struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb);
+struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
+			     bool *found_dup_sk);
 struct request_sock *cookie_tcp_reqsk_alloc(const struct request_sock_ops *ops,
 					    struct sock *sk, struct sk_buff *skb);
 #ifdef CONFIG_SYN_COOKIES
@@ -562,7 +564,8 @@ bool cookie_ecn_ok(const struct tcp_options_received *opt,
 /* From net/ipv6/syncookies.c */
 int __cookie_v6_check(const struct ipv6hdr *iph, const struct tcphdr *th,
 		      u32 cookie);
-struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
+struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb,
+			     bool *found_dup_sk);
 
 u32 __cookie_v6_init_sequence(const struct ipv6hdr *iph,
 			      const struct tcphdr *th, u16 *mssp);
net/dccp/ipv4.c +1 -1
index 9c28c8251125..098bae35ab76 100644
@@ -427,7 +427,7 @@ struct sock *dccp_v4_request_recv_sock(const struct sock *sk,
 
 	if (__inet_inherit_port(sk, newsk) < 0)
 		goto put_and_exit;
-	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
+	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash), NULL);
 	if (*own_req)
 		ireq->ireq_opt = NULL;
 	else
net/dccp/ipv6.c +1 -1
index ef4ab28cfde0..78ee1b5acf1f 100644
@@ -533,7 +533,7 @@ static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
 		dccp_done(newsk);
 		goto out;
 	}
-	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
+	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash), NULL);
 	/* Clone pktoptions received with SYN, if we own the req */
 	if (*own_req && ireq->pktopts) {
 		newnp->pktoptions = skb_clone(ireq->pktopts, GFP_ATOMIC);
net/ipv4/inet_connection_sock.c +1 -1
index b457dd2d6c75..df26489e4f6c 100644
@@ -787,7 +787,7 @@ static void reqsk_queue_hash_req(struct request_sock *req,
 	timer_setup(&req->rsk_timer, reqsk_timer_handler, TIMER_PINNED);
 	mod_timer(&req->rsk_timer, jiffies + timeout);
 
-	inet_ehash_insert(req_to_sk(req), NULL);
+	inet_ehash_insert(req_to_sk(req), NULL, NULL);
 	/* before letting lookups find us, make sure all req fields
 	 * are committed to memory and refcnt initialized.
 	 */
net/ipv4/inet_hashtables.c +60 -8

113 lines not on this page · raw message

net/ipv4/syncookies.c +7 -3
index e03756631541..5026aded8e42 100644
@@ -200,7 +200,8 @@ EXPORT_SYMBOL_GPL(__cookie_v4_check);
 
 struct sock *tcp_get_cookie_sock(struct sock *sk, struct sk_buff *skb,
 				 struct request_sock *req,
-				 struct dst_entry *dst, u32 tsoff)
+				 struct dst_entry *dst, u32 tsoff,
+				 bool *found_dup_sk)
 {
 	struct inet_connection_sock *icsk = inet_csk(sk);
 	struct sock *child;
@@ -223,6 +224,8 @@ struct sock *tcp_get_cookie_sock(struct sock *sk, struct sk_buff *skb,
 
 		bh_unlock_sock(child);
 		sock_put(child);
+	}  else if (!own_req) {
+		*found_dup_sk = true;
 	}
 	__reqsk_free(req);
 
@@ -320,7 +323,8 @@ EXPORT_SYMBOL_GPL(cookie_tcp_reqsk_alloc);
  * Output is listener if incoming packet would not create a child
  *           NULL if memory could not be allocated.
  */
-struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb)
+struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
+			     bool *found_dup_sk)
 {
 	struct ip_options *opt = &TCP_SKB_CB(skb)->header.h4.opt;
 	struct tcp_options_received tcp_opt;
@@ -436,7 +440,7 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb)
 	ireq->rcv_wscale  = rcv_wscale;
 	ireq->ecn_ok = cookie_ecn_ok(&tcp_opt, sock_net(sk), &rt->dst);
 
-	ret = tcp_get_cookie_sock(sk, skb, req, &rt->dst, tsoff);
+	ret = tcp_get_cookie_sock(sk, skb, req, &rt->dst, tsoff, found_dup_sk);
 	/* ip_queue_xmit() depends on our flow being setup
 	 * Normal sockets get it right from inet_csk_route_child_sock()
 	 */
net/ipv4/tcp_ipv4.c +36 -5

96 lines not on this page · raw message

net/ipv6/syncookies.c +3 -2
index e796a64be308..3994d0396691 100644
@@ -125,7 +125,8 @@ int __cookie_v6_check(const struct ipv6hdr *iph, const struct tcphdr *th,
 }
 EXPORT_SYMBOL_GPL(__cookie_v6_check);
 
-struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)
+struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb,
+			     bool *found_dup_sk)
 {
 	struct tcp_options_received tcp_opt;
 	struct inet_request_sock *ireq;
@@ -249,7 +250,7 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)
 	ireq->rcv_wscale = rcv_wscale;
 	ireq->ecn_ok = cookie_ecn_ok(&tcp_opt, sock_net(sk), dst);
 
-	ret = tcp_get_cookie_sock(sk, skb, req, dst, tsoff);
+	ret = tcp_get_cookie_sock(sk, skb, req, dst, tsoff, found_dup_sk);
 out:
 	return ret;
 out_free:
net/ipv6/tcp_ipv6.c +35 -4
index 305870a72352..cdf2b94e3e43 100644
@@ -1125,13 +1125,14 @@ static void tcp_v6_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
 }
 
 
-static struct sock *tcp_v6_cookie_check(struct sock *sk, struct sk_buff *skb)
+static struct sock *tcp_v6_cookie_check(struct sock *sk, struct sk_buff *skb,
+					bool *found_dup_sk)
 {
 #ifdef CONFIG_SYN_COOKIES
 	const struct tcphdr *th = tcp_hdr(skb);
 
 	if (!th->syn)
-		sk = cookie_v6_check(sk, skb);
+		sk = cookie_v6_check(sk, skb, found_dup_sk);
 #endif
 	return sk;
 }
@@ -1188,6 +1189,7 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *
 	const struct ipv6_pinfo *np = tcp_inet6_sk(sk);
 	struct ipv6_txoptions *opt;
 	struct inet_sock *newinet;
+	bool found_dup_sk = false;
 	struct tcp_sock *newtp;
 	struct sock *newsk;
 #ifdef CONFIG_TCP_MD5SIG
@@ -1359,7 +1361,8 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *
 		tcp_done(newsk);
 		goto out;
 	}
-	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
+	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash),
+				       &found_dup_sk);
 	if (*own_req) {
 		tcp_move_syn(newtp, req);
 
@@ -1374,6 +1377,15 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *
 				skb_set_owner_r(newnp->pktoptions, newsk);
 			}
 		}
+	} else {
+		if (!req_unhash && found_dup_sk) {
+			/* This code path should only be executed in the
+			 * syncookie case only
+			 */
+			bh_unlock_sock(newsk);
+			sock_put(newsk);
+			newsk = NULL;
+		}
 	}
 
 	return newsk;
@@ -1399,6 +1411,7 @@ static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
 {
 	struct ipv6_pinfo *np = tcp_inet6_sk(sk);
 	struct sk_buff *opt_skb = NULL;
+	bool found_dup_sk = false;
 	struct tcp_sock *tp;
 
 	/* Imagine: socket is IPv6. IPv4 packet arrives,
@@ -1456,7 +1469,16 @@ static int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
 		goto csum_err;
 
 	if (sk->sk_state == TCP_LISTEN) {
-		struct sock *nsk = tcp_v6_cookie_check(sk, skb);
+		struct sock *nsk = tcp_v6_cookie_check(sk, skb, &found_dup_sk);
+
+		if (found_dup_sk) {
+			/* Returning 1 here means that tcp_v6_rcv should lookup
+			 * for an established socket again because we just
+			 * found out that a previous socket already exists for
+			 * the same SYN cookie.
+			 */
+			return 1;
+		}
 
 		if (!nsk)
 			goto discard;
@@ -1554,6 +1576,7 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 	int dif = inet6_iif(skb);
 	const struct tcphdr *th;
 	const struct ipv6hdr *hdr;
+	struct net_device *dev;
 	bool refcounted;
 	struct sock *sk;
 	int ret;
@@ -1664,10 +1687,18 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 	hdr = ipv6_hdr(skb);
 	tcp_v6_fill_cb(skb, hdr, th);
 
+	dev = skb->dev;
 	skb->dev = NULL;
 
 	if (sk->sk_state == TCP_LISTEN) {
 		ret = tcp_v6_do_rcv(sk, skb);
+		if (ret == 1) {
+			skb->dev = dev;
+			tcp_v6_restore_cb(skb);
+			if (refcounted)
+				sock_put(sk);
+			goto lookup;
+		}
 		goto put_and_return;
 	}
 
Signature · 1 line
2.25.1
#1 lore raw
Quoted 35 lines · Ricardo Dias

When the TCP stack is in SYN flood mode, the server child socket is created from the SYN cookie received in a TCP packet with the ACK flag set.

The child socket is created when the server receives the first TCP packet with a valid SYN cookie from the client. Usually, this packet corresponds to the final step of the TCP 3-way handshake, the ACK packet. But is also possible to receive a valid SYN cookie from the first TCP data packet sent by the client, and thus create a child socket from that SYN cookie.

Since a client socket is ready to send data as soon as it receives the SYN+ACK packet from the server, the client can send the ACK packet (sent by the TCP stack code), and the first data packet (sent by the userspace program) almost at the same time, and thus the server will equally receive the two TCP packets with valid SYN cookies almost at the same instant.

When such event happens, the TCP stack code has a race condition that occurs between the momement a lookup is done to the established connections hashtable to check for the existence of a connection for the same client, and the moment that the child socket is added to the established connections hashtable. As a consequence, this race condition can lead to a situation where we add two child sockets to the established connections hashtable and deliver two sockets to the userspace program to the same client.

This patch fixes the race condition by checking if an existing child socket exists for the same client when we are adding the second child socket to the established connections socket. If an existing child socket exists, we return that socket and use it to process the TCP packet received, and discard the second child socket to the same client.

Quoted 13 lines

---
v7 (2020-11-19):
* Changed the approach to re-use the first (existing) socket created from thge syncookie. Instead of returning the existing socket in tcp_(v4|v6)_syn_recv_sock and continue the protocol state machine execution, tcp_(v4|v6)_syn_recv_sock signals that already exists a socket, and tells tcp_(v4|v6)_rcv to lookup the socket again in the established connections table.
This new approach fixes the errors reported by Eric for the previous version of the patch.
* Also fixes the memory leaks by making sure that the newly created socket in syn_recv_sock is destroyed in case an already existing socket exists.

I think this is going too far. Your patch is too complex/risky, and will be hard to backport to old kernels, because TCP stack has changed a lot.

Alternative approach would be to detect the race and simply drop the packet that lost the battle.