From: Sowmini Varadhan <sowmini.varadhan@xxxxxxxxxx>
[ Upstream commit 10beea7d7408d0b1c9208757f445c5c710239e0e ]
Each time we get an incoming SYN to the RDS_TCP_PORT, the TCP
layer accepts the connection and then the rds_tcp_accept_one()
callback is invoked to process the incoming connection.
rds_tcp_accept_one() may reject the incoming syn for a number of
reasons, e.g., commit 1a0e100fb2c9 ("RDS: TCP: Force every connection
to be initiated by numerically smaller IP address"), or because
we are getting spammed by a malicious node that is triggering
a flood of connection attempts to RDS_TCP_PORT. If the incoming
syn is rejected, no data would have been sent on the TCP socket,
and we do not need to be in TIME_WAIT state, so we set linger on
the TCP socket before closing, thereby closing the socket efficiently
with a RST.
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@xxxxxxxxxx>
Tested-by: Imanti Mendez <imanti.mendez@xxxxxxxxxx>
Acked-by: Santosh Shilimkar <santosh.shilimkar@xxxxxxxxxx>
Signed-off-by: David S. Miller <davem@xxxxxxxxxxxxx>
Signed-off-by: Sasha Levin <alexander.levin@xxxxxxxxxxxxx>
---
net/rds/tcp_listen.c | 19 ++++++++++++++++++-
1 file changed, 18 insertions(+), 1 deletion(-)
diff --git a/net/rds/tcp_listen.c b/net/rds/tcp_listen.c
index 185a56b1e29c..1bdf1932c8bf 100644
--- a/net/rds/tcp_listen.c
+++ b/net/rds/tcp_listen.c
@@ -116,6 +116,17 @@ struct rds_tcp_connection *rds_tcp_accept_one_path(struct rds_connection *conn)
return NULL;
}
+static void rds_tcp_set_linger(struct socket *sock)
+{
+ struct linger no_linger = {
+ .l_onoff = 1,
+ .l_linger = 0,
+ };
+
+ kernel_setsockopt(sock, SOL_SOCKET, SO_LINGER,
+ (char *)&no_linger, sizeof(no_linger));
+}
+
int rds_tcp_accept_one(struct socket *sock)
{
struct socket *new_sock = NULL;
@@ -198,7 +209,13 @@ int rds_tcp_accept_one(struct socket *sock)
ret = 0;
goto out;
rst_nsk:
- /* reset the newly returned accept sock and bail */
+ /* reset the newly returned accept sock and bail.
+ * It is safe to set linger on new_sock because the RDS connection
+ * has not been brought up on new_sock, so no RDS-level data could
+ * be pending on it. By setting linger, we achieve the side-effect
+ * of avoiding TIME_WAIT state on new_sock.
+ */
+ rds_tcp_set_linger(new_sock);
kernel_sock_shutdown(new_sock, SHUT_RDWR);
ret = 0;
out:
--
2.15.1