Skip to contentSkip to conversation

RDS: TCP: bug fixes

PATCH In review Sowmini Varadhan 2 participants 4 messages Nov 4, 2016 – Nov 9, 2016

Subsystems (3)
Networking Network drivers RDS - reliable datagram sockets
lore
Share & tools
lore

Sowmini Varadhan

author
#0 lore raw

cover RDS: TCP: bug fixes

A couple of bug fixes identified during testing.

Sowmini Varadhan (2):
RDS: TCP: report addr/port info based on TCP socket in rds-info
RDS: TCP: start multipath acceptor loop at 0

 net/rds/tcp.c        |   20 +++++++++++++-------
 net/rds/tcp_listen.c |    2 +-
 2 files changed, 14 insertions(+), 8 deletions(-)
author
#1 lore raw

PATCH 1/2 RDS: TCP: report addr/port info based on TCP socket in rds-info

The socket argument passed to rds_tcp_tc_info() is a PF_RDS socket, so it is incorrect to report the address port info based on rds_getname() as part of TCP state report.

Invoke inet_getname() for the t_sock associated with the rds_tcp_connection instead.

Signed-off-by Sowmini Varadhan <[email protected]>
Acked-by Santosh Shilimkar <[email protected]>
---
 net/rds/tcp.c |   20 +++++++++++++-------
 1 files changed, 13 insertions(+), 7 deletions(-)
net/rds/tcp.c +13 -7
index fcddacc..3296a6a 100644
@@ -220,7 +220,7 @@ void rds_tcp_set_callbacks(struct socket *sock, struct rds_conn_path *cp)
 	write_unlock_bh(&sock->sk->sk_callback_lock);
 }
 
-static void rds_tcp_tc_info(struct socket *sock, unsigned int len,
+static void rds_tcp_tc_info(struct socket *rds_sock, unsigned int len,
 			    struct rds_info_iterator *iter,
 			    struct rds_info_lengths *lens)
 {

@@ -229,6 +229,7 @@ static void rds_tcp_tc_info(struct socket *sock, unsigned int len,
 	unsigned long flags;
 	struct sockaddr_in sin;
 	int sinlen;
+	struct socket *sock;
 
 	spin_lock_irqsave(&rds_tcp_tc_list_lock, flags);
 

@@ -237,12 +238,17 @@ static void rds_tcp_tc_info(struct socket *sock, unsigned int len,
 
 	list_for_each_entry(tc, &rds_tcp_tc_list, t_list_item) {
 
-		sock->ops->getname(sock, (struct sockaddr *)&sin, &sinlen, 0);
-		tsinfo.local_addr = sin.sin_addr.s_addr;
-		tsinfo.local_port = sin.sin_port;
-		sock->ops->getname(sock, (struct sockaddr *)&sin, &sinlen, 1);
-		tsinfo.peer_addr = sin.sin_addr.s_addr;
-		tsinfo.peer_port = sin.sin_port;
+		sock = tc->t_sock;
+		if (sock) {
+			sock->ops->getname(sock, (struct sockaddr *)&sin,
+					   &sinlen, 0);
+			tsinfo.local_addr = sin.sin_addr.s_addr;
+			tsinfo.local_port = sin.sin_port;
+			sock->ops->getname(sock, (struct sockaddr *)&sin,
+					   &sinlen, 1);
+			tsinfo.peer_addr = sin.sin_addr.s_addr;
+			tsinfo.peer_port = sin.sin_port;
+		}
 
 		tsinfo.hdr_rem = tc->t_tinc_hdr_rem;
 		tsinfo.data_rem = tc->t_tinc_data_rem;
Signature · 1 line
1.7.1
author
#2 lore raw

PATCH 2/2 RDS: TCP: start multipath acceptor loop at 0

In reply to Sowmini Varadhan

The for() loop in rds_tcp_accept_one() assumes that the 0'th rds_tcp_conn_path is UP and starts multipath accepts at index 1. But this assumption may not always be true: if the 0'th path has failed (ERROR or DOWN state) an incoming connection request should be used to resurrect this path.

Signed-off-by Sowmini Varadhan <[email protected]>
Acked-by Santosh Shilimkar <[email protected]>
---
 net/rds/tcp_listen.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
net/rds/tcp_listen.c +1 -1
index e0b23fb..c9c4968 100644
@@ -103,7 +103,7 @@ struct rds_tcp_connection *rds_tcp_accept_one_path(struct rds_connection *conn)
 	if (!peer_is_smaller)
 		return NULL;
 
-	for (i = 1; i < npaths; i++) {
+	for (i = 0; i < npaths; i++) {
 		struct rds_conn_path *cp = &conn->c_path[i];
 
 		if (rds_conn_path_transition(cp, RDS_CONN_DOWN,
Signature · 1 line
1.7.1