summaryrefslogtreecommitdiff
path: root/net/rds/send.c
diff options
context:
space:
mode:
authorSowmini Varadhan <sowmini.varadhan@oracle.com>2016-07-14 13:51:03 +0300
committerDavid S. Miller <davem@davemloft.net>2016-07-15 21:36:58 +0300
commit5916e2c1554f3e36f770401c989c3c7fadf619ca (patch)
treee2f85d0d6ad83d6835b131956324d6e30c4eda3e /net/rds/send.c
parentac3615e7f3cffe2a1a6b25172dfd09e138593d82 (diff)
downloadlinux-5916e2c1554f3e36f770401c989c3c7fadf619ca.tar.xz
RDS: TCP: Enable multipath RDS for TCP
Use RDS probe-ping to compute how many paths may be used with the peer, and to synchronously start the multiple paths. If mprds is supported, hash outgoing traffic to one of multiple paths in rds_sendmsg() when multipath RDS is supported by the transport. CC: Santosh Shilimkar <santosh.shilimkar@oracle.com> Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com> Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/rds/send.c')
-rw-r--r--net/rds/send.c71
1 files changed, 67 insertions, 4 deletions
diff --git a/net/rds/send.c b/net/rds/send.c
index 5a9caf1da896..896626b9a0ef 100644
--- a/net/rds/send.c
+++ b/net/rds/send.c
@@ -963,6 +963,29 @@ static int rds_cmsg_send(struct rds_sock *rs, struct rds_message *rm,
return ret;
}
+static void rds_send_ping(struct rds_connection *conn);
+
+static int rds_send_mprds_hash(struct rds_sock *rs, struct rds_connection *conn)
+{
+ int hash;
+
+ if (conn->c_npaths == 0)
+ hash = RDS_MPATH_HASH(rs, RDS_MPATH_WORKERS);
+ else
+ hash = RDS_MPATH_HASH(rs, conn->c_npaths);
+ if (conn->c_npaths == 0 && hash != 0) {
+ rds_send_ping(conn);
+
+ if (conn->c_npaths == 0) {
+ wait_event_interruptible(conn->c_hs_waitq,
+ (conn->c_npaths != 0));
+ }
+ if (conn->c_npaths == 1)
+ hash = 0;
+ }
+ return hash;
+}
+
int rds_sendmsg(struct socket *sock, struct msghdr *msg, size_t payload_len)
{
struct sock *sk = sock->sk;
@@ -1075,7 +1098,10 @@ int rds_sendmsg(struct socket *sock, struct msghdr *msg, size_t payload_len)
goto out;
}
- cpath = &conn->c_path[0];
+ if (conn->c_trans->t_mp_capable)
+ cpath = &conn->c_path[rds_send_mprds_hash(rs, conn)];
+ else
+ cpath = &conn->c_path[0];
rds_conn_path_connect_if_down(cpath);
@@ -1135,10 +1161,16 @@ out:
}
/*
- * Reply to a ping packet.
+ * send out a probe. Can be shared by rds_send_ping,
+ * rds_send_pong, rds_send_hb.
+ * rds_send_hb should use h_flags
+ * RDS_FLAG_HB_PING|RDS_FLAG_ACK_REQUIRED
+ * or
+ * RDS_FLAG_HB_PONG|RDS_FLAG_ACK_REQUIRED
*/
int
-rds_send_pong(struct rds_conn_path *cp, __be16 dport)
+rds_send_probe(struct rds_conn_path *cp, __be16 sport,
+ __be16 dport, u8 h_flags)
{
struct rds_message *rm;
unsigned long flags;
@@ -1166,9 +1198,18 @@ rds_send_pong(struct rds_conn_path *cp, __be16 dport)
rm->m_inc.i_conn = cp->cp_conn;
rm->m_inc.i_conn_path = cp;
- rds_message_populate_header(&rm->m_inc.i_hdr, 0, dport,
+ rds_message_populate_header(&rm->m_inc.i_hdr, sport, dport,
cp->cp_next_tx_seq);
+ rm->m_inc.i_hdr.h_flags |= h_flags;
cp->cp_next_tx_seq++;
+
+ if (RDS_HS_PROBE(sport, dport) && cp->cp_conn->c_trans->t_mp_capable) {
+ u16 npaths = RDS_MPATH_WORKERS;
+
+ rds_message_add_extension(&rm->m_inc.i_hdr,
+ RDS_EXTHDR_NPATHS, &npaths,
+ sizeof(npaths));
+ }
spin_unlock_irqrestore(&cp->cp_lock, flags);
rds_stats_inc(s_send_queued);
@@ -1185,3 +1226,25 @@ out:
rds_message_put(rm);
return ret;
}
+
+int
+rds_send_pong(struct rds_conn_path *cp, __be16 dport)
+{
+ return rds_send_probe(cp, 0, dport, 0);
+}
+
+void
+rds_send_ping(struct rds_connection *conn)
+{
+ unsigned long flags;
+ struct rds_conn_path *cp = &conn->c_path[0];
+
+ spin_lock_irqsave(&cp->cp_lock, flags);
+ if (conn->c_ping_triggered) {
+ spin_unlock_irqrestore(&cp->cp_lock, flags);
+ return;
+ }
+ conn->c_ping_triggered = 1;
+ spin_unlock_irqrestore(&cp->cp_lock, flags);
+ rds_send_probe(&conn->c_path[0], RDS_FLAG_PROBE_PORT, 0, 0);
+}