Merge: mptcp: fix DSS map corruption

MR: https://gitlab.com/redhat/centos-stream/src/kernel/centos-stream-9/-/merge_requests/5418

JIRA: https://issues.redhat.com/browse/RHEL-55470

Tested: vs self-tests

This series addresses an issue observed in production and sharing the same root cause with an upstream syzkaller report. The main fix is the second patch in the series. The first one ensure that, should similar issue arise in the future, we will get consistent behavior from the kernel, and suitable introspection.

Signed-off-by: Paolo Abeni <pabeni@redhat.com>

Approved-by: Davide Caratti <dcaratti@redhat.com>
Approved-by: Antoine Tenart <atenart@redhat.com>
Approved-by: CKI KWF Bot <cki-ci-bot+kwf-gitlab-com@redhat.com>

Merged-by: Rado Vrbovsky <rvrbovsk@redhat.com>

Author: emmil

net/ipv4/tcp_output.c

@@ -2310,9 +2310,7 @@ static bool tcp_can_coalesce_send_queue_head(struct sock *sk, int len)
 		if (len <= skb->len)
 			break;
 
-		if (unlikely(TCP_SKB_CB(skb)->eor) ||
-		    tcp_has_tx_tstamp(skb) ||
-		    !skb_pure_zcopy_same(skb, next))
+		if (tcp_has_tx_tstamp(skb) || !tcp_skb_can_collapse(skb, next))
 			return false;
 
 		len -= skb->len;

net/mptcp/mib.c

@@ -24,6 +24,8 @@ static const struct snmp_mib mptcp_snmp_list[] = {
 	SNMP_MIB_ITEM("MPJoinAckRx", MPTCP_MIB_JOINACKRX),
 	SNMP_MIB_ITEM("MPJoinAckHMacFailure", MPTCP_MIB_JOINACKMAC),
 	SNMP_MIB_ITEM("DSSNotMatching", MPTCP_MIB_DSSNOMATCH),
+	SNMP_MIB_ITEM("DSSCorruptionFallback", MPTCP_MIB_DSSCORRUPTIONFALLBACK),
+	SNMP_MIB_ITEM("DSSCorruptionReset", MPTCP_MIB_DSSCORRUPTIONRESET),
 	SNMP_MIB_ITEM("InfiniteMapTx", MPTCP_MIB_INFINITEMAPTX),
 	SNMP_MIB_ITEM("InfiniteMapRx", MPTCP_MIB_INFINITEMAPRX),
 	SNMP_MIB_ITEM("DSSNoMatchTCP", MPTCP_MIB_DSSTCPMISMATCH),

net/mptcp/mib.h

@@ -17,6 +17,8 @@ enum linux_mptcp_mib_field {
 	MPTCP_MIB_JOINACKRX,		/* Received an ACK + MP_JOIN */
 	MPTCP_MIB_JOINACKMAC,		/* HMAC was wrong on ACK + MP_JOIN */
 	MPTCP_MIB_DSSNOMATCH,		/* Received a new mapping that did not match the previous one */
+	MPTCP_MIB_DSSCORRUPTIONFALLBACK,/* DSS corruption detected, fallback */
+	MPTCP_MIB_DSSCORRUPTIONRESET,	/* DSS corruption detected, MPJ subflow reset */
 	MPTCP_MIB_INFINITEMAPTX,	/* Sent an infinite mapping */
 	MPTCP_MIB_INFINITEMAPRX,	/* Received an infinite mapping */
 	MPTCP_MIB_DSSTCPMISMATCH,	/* DSS-mapping did not map with TCP's sequence numbers */

net/mptcp/protocol.c

@@ -618,6 +618,18 @@ static bool mptcp_check_data_fin(struct sock *sk)
 	return ret;
 }
 
+static void mptcp_dss_corruption(struct mptcp_sock *msk, struct sock *ssk)
+{
+	if (READ_ONCE(msk->allow_infinite_fallback)) {
+		MPTCP_INC_STATS(sock_net(ssk),
+				MPTCP_MIB_DSSCORRUPTIONFALLBACK);
+		mptcp_do_fallback(ssk);
+	} else {
+		MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_DSSCORRUPTIONRESET);
+		mptcp_subflow_reset(ssk);
+	}
+}
+
 static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
 					   struct sock *ssk,
 					   unsigned int *bytes)
@@ -690,10 +702,16 @@ static bool __mptcp_move_skbs_from_subflow(struct mptcp_sock *msk,
 				moved += len;
 			seq += len;
 
-			if (WARN_ON_ONCE(map_remaining < len))
-				break;
+			if (unlikely(map_remaining < len)) {
+				DEBUG_NET_WARN_ON_ONCE(1);
+				mptcp_dss_corruption(msk, ssk);
+			}
 		} else {
-			WARN_ON_ONCE(!fin);
+			if (unlikely(!fin)) {
+				DEBUG_NET_WARN_ON_ONCE(1);
+				mptcp_dss_corruption(msk, ssk);
+			}
+
 			sk_eat_skb(ssk, skb);
 			done = true;
 		}

net/mptcp/subflow.c

@@ -952,8 +952,10 @@ static bool skb_is_fully_mapped(struct sock *ssk, struct sk_buff *skb)
 	unsigned int skb_consumed;
 
 	skb_consumed = tcp_sk(ssk)->copied_seq - TCP_SKB_CB(skb)->seq;
-	if (WARN_ON_ONCE(skb_consumed >= skb->len))
+	if (unlikely(skb_consumed >= skb->len)) {
+		DEBUG_NET_WARN_ON_ONCE(1);
 		return true;
+	}
 
 	return skb->len - skb_consumed <= subflow->map_data_len -
 					  mptcp_subflow_get_map_offset(subflow);