Merge branch 'rds-next'

Sowmini Varadhan says:

====================
net/rds: SOL_RDS socket option to explicitly select transport

Today the underlying transport (TCP or IB) for a PF_RDS socket is
implicitly selected based on the local address used to bind(2) the
PF_RDS socket. This results in some non-deterministic behavior when
there are un-numbered and IPoIB interfaces sharing the same IP address.
It also places the constraint that the IB interface must have an IP
address (and thus, IPoIB) configured on it.

The non-determinism may be avoided by providing the user-space application
a socket option that allows it to explicitly select the transport
prior to bind(2).

Patch 1 of this series provides the constant definitions needed by
the application via <linux/rds.h>.

Patch 2 provides the setsockopt support, and Patch 3 provides the
getsockopt support.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

5 files changed, 77 insertions, 5 deletions

diff --git a/include/uapi/linux/rds.h b/include/uapi/linux/rds.h
index 91950950aa59..0f9265cb2a96 100644
--- a/include/uapi/linux/rds.h
+++ b/include/uapi/linux/rds.h
@@ -38,6 +38,8 @@
 
 #define RDS_IB_ABI_VERSION		0x301
 
+#define	SOL_RDS		276
+
 /*
  * setsockopt/getsockopt for SOL_RDS
  */
@@ -48,6 +50,14 @@
 #define RDS_RECVERR			5
 #define RDS_CONG_MONITOR		6
 #define RDS_GET_MR_FOR_DEST		7
+#define SO_RDS_TRANSPORT		8
+
+/* supported values for SO_RDS_TRANSPORT */
+#define	RDS_TRANS_IB	0
+#define	RDS_TRANS_IWARP	1
+#define	RDS_TRANS_TCP	2
+#define RDS_TRANS_COUNT	3
+#define	RDS_TRANS_NONE	(~0)
 
 /*
  * Control message types for SOL_RDS.
diff --git a/net/rds/af_rds.c b/net/rds/af_rds.c
index 3d83641f2861..2ad9032372b2 100644
--- a/net/rds/af_rds.c
+++ b/net/rds/af_rds.c
@@ -270,6 +270,28 @@ static int rds_cong_monitor(struct rds_sock *rs, char __user *optval,
 	return ret;
 }
 
+static int rds_set_transport(struct rds_sock *rs, char __user *optval,
+			     int optlen)
+{
+	int t_type;
+
+	if (rs->rs_transport)
+		return -EOPNOTSUPP; /* previously attached to transport */
+
+	if (optlen != sizeof(int))
+		return -EINVAL;
+
+	if (copy_from_user(&t_type, (int __user *)optval, sizeof(t_type)))
+		return -EFAULT;
+
+	if (t_type < 0 || t_type >= RDS_TRANS_COUNT)
+		return -EINVAL;
+
+	rs->rs_transport = rds_trans_get(t_type);
+
+	return rs->rs_transport ? 0 : -ENOPROTOOPT;
+}
+
 static int rds_setsockopt(struct socket *sock, int level, int optname,
 			  char __user *optval, unsigned int optlen)
 {
@@ -300,6 +322,11 @@ static int rds_setsockopt(struct socket *sock, int level, int optname,
 	case RDS_CONG_MONITOR:
 		ret = rds_cong_monitor(rs, optval, optlen);
 		break;
+	case SO_RDS_TRANSPORT:
+		lock_sock(sock->sk);
+		ret = rds_set_transport(rs, optval, optlen);
+		release_sock(sock->sk);
+		break;
 	default:
 		ret = -ENOPROTOOPT;
 	}
@@ -312,6 +339,7 @@ static int rds_getsockopt(struct socket *sock, int level, int optname,
 {
 	struct rds_sock *rs = rds_sk_to_rs(sock->sk);
 	int ret = -ENOPROTOOPT, len;
+	int trans;
 
 	if (level != SOL_RDS)
 		goto out;
@@ -337,6 +365,19 @@ static int rds_getsockopt(struct socket *sock, int level, int optname,
 		else
 			ret = 0;
 		break;
+	case SO_RDS_TRANSPORT:
+		if (len < sizeof(int)) {
+			ret = -EINVAL;
+			break;
+		}
+		trans = (rs->rs_transport ? rs->rs_transport->t_type :
+			 RDS_TRANS_NONE); /* unbound */
+		if (put_user(trans, (int __user *)optval) ||
+		    put_user(sizeof(int), optlen))
+			ret = -EFAULT;
+		else
+			ret = 0;
+		break;
 	default:
 		break;
 	}
diff --git a/net/rds/bind.c b/net/rds/bind.c
index a2e6562da751..4ebd29c128b6 100644
--- a/net/rds/bind.c
+++ b/net/rds/bind.c
@@ -181,6 +181,10 @@ int rds_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 	if (ret)
 		goto out;
 
+	if (rs->rs_transport) { /* previously bound */
+		ret = 0;
+		goto out;
+	}
 	trans = rds_trans_get_preferred(sin->sin_addr.s_addr);
 	if (!trans) {
 		ret = -EADDRNOTAVAIL;
diff --git a/net/rds/rds.h b/net/rds/rds.h
index 0d41155a2258..a33fb4ad3535 100644
--- a/net/rds/rds.h
+++ b/net/rds/rds.h
@@ -408,11 +408,6 @@ struct rds_notifier {
  *                 should try hard not to block.
  */
 
-#define RDS_TRANS_IB	0
-#define RDS_TRANS_IWARP	1
-#define RDS_TRANS_TCP	2
-#define RDS_TRANS_COUNT	3
-
 struct rds_transport {
 	char			t_name[TRANSNAMSIZ];
 	struct list_head	t_item;
@@ -803,6 +798,7 @@ struct rds_transport *rds_trans_get_preferred(__be32 addr);
 void rds_trans_put(struct rds_transport *trans);
 unsigned int rds_trans_stats_info_copy(struct rds_info_iterator *iter,
 				       unsigned int avail);
+struct rds_transport *rds_trans_get(int t_type);
 int rds_trans_init(void);
 void rds_trans_exit(void);
 
diff --git a/net/rds/transport.c b/net/rds/transport.c
index 7f2ac4fec367..8b4a6cd2c3a7 100644
--- a/net/rds/transport.c
+++ b/net/rds/transport.c
@@ -101,6 +101,27 @@ struct rds_transport *rds_trans_get_preferred(__be32 addr)
 	return ret;
 }
 
+struct rds_transport *rds_trans_get(int t_type)
+{
+	struct rds_transport *ret = NULL;
+	struct rds_transport *trans;
+	unsigned int i;
+
+	down_read(&rds_trans_sem);
+	for (i = 0; i < RDS_TRANS_COUNT; i++) {
+		trans = transports[i];
+
+		if (trans && trans->t_type == t_type &&
+		    (!trans->t_owner || try_module_get(trans->t_owner))) {
+			ret = trans;
+			break;
+		}
+	}
+	up_read(&rds_trans_sem);
+
+	return ret;
+}
+
 /*
  * This returns the number of stats entries in the snapshot and only
  * copies them using the iter if there is enough space for them.  The