Merge branch 'rds-enable-mprds'

Sowmini Varadhan says:

====================
RDS: TCP: Enable mprds for rds-tcp

The third, and final, installment for mprds-tcp changes.

In Patch 3 of this set, if the transport support t_mp_capable,
we hash outgoing traffic across multiple paths.  Additionally, even if
the transport is MP capable, we may be peering with some node that does
not support mprds, or supports a different number of paths. This
necessitates RDS control plane changes so that both peers agree
on the number of paths to be used for the rds-tcp connection.
Patch 3 implements all these changes, which are documented in patch 5
of the series.

Patch 1 of this series is a bug fix for a race-condition
that has always existed, but is now more easily encountered with mprds.
Patch 2 is code refactoring. Patches 4 and 5 are Documentation updates.
====================

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

Documentation/networking/rds.txt

@@ -85,7 +85,8 @@ Socket Interface
 
   bind(fd, &sockaddr_in, ...)
         This binds the socket to a local IP address and port, and a
-        transport.
+        transport, if one has not already been selected via the
+	SO_RDS_TRANSPORT socket option
 
   sendmsg(fd, ...)
         Sends a message to the indicated recipient. The kernel will
@@ -146,6 +147,20 @@ Socket Interface
         operation. In this case, it would use RDS_CANCEL_SENT_TO to
         nuke any pending messages.
 
+  setsockopt(fd, SOL_RDS, SO_RDS_TRANSPORT, (int *)&transport ..)
+  getsockopt(fd, SOL_RDS, SO_RDS_TRANSPORT, (int *)&transport ..)
+	Set or read an integer defining  the underlying
+	encapsulating transport to be used for RDS packets on the
+	socket. When setting the option, integer argument may be
+	one of RDS_TRANS_TCP or RDS_TRANS_IB. When retrieving the
+	value, RDS_TRANS_NONE will be returned on an unbound socket.
+	This socket option may only be set exactly once on the socket,
+	prior to binding it via the bind(2) system call. Attempts to
+	set SO_RDS_TRANSPORT on a socket for which the transport has
+	been previously attached explicitly (by SO_RDS_TRANSPORT) or
+	implicitly (via bind(2)) will return an error of EOPNOTSUPP.
+	An attempt to set SO_RDS_TRANSPPORT to RDS_TRANS_NONE will
+	always return EINVAL.
 
 RDMA for RDS
 ============
@@ -350,4 +365,59 @@ The recv path
     handle CMSGs
     return to application
 
+Multipath RDS (mprds)
+=====================
+  Mprds is multipathed-RDS, primarily intended for RDS-over-TCP
+  (though the concept can be extended to other transports). The classical
+  implementation of RDS-over-TCP is implemented by demultiplexing multiple
+  PF_RDS sockets between any 2 endpoints (where endpoint == [IP address,
+  port]) over a single TCP socket between the 2 IP addresses involved. This
+  has the limitation that it ends up funneling multiple RDS flows over a
+  single TCP flow, thus it is
+  (a) upper-bounded to the single-flow bandwidth,
+  (b) suffers from head-of-line blocking for all the RDS sockets.
+
+  Better throughput (for a fixed small packet size, MTU) can be achieved
+  by having multiple TCP/IP flows per rds/tcp connection, i.e., multipathed
+  RDS (mprds).  Each such TCP/IP flow constitutes a path for the rds/tcp
+  connection. RDS sockets will be attached to a path based on some hash
+  (e.g., of local address and RDS port number) and packets for that RDS
+  socket will be sent over the attached path using TCP to segment/reassemble
+  RDS datagrams on that path.
+
+  Multipathed RDS is implemented by splitting the struct rds_connection into
+  a common (to all paths) part, and a per-path struct rds_conn_path. All
+  I/O workqs and reconnect threads are driven from the rds_conn_path.
+  Transports such as TCP that are multipath capable may then set up a
+  TPC socket per rds_conn_path, and this is managed by the transport via
+  the transport privatee cp_transport_data pointer.
+
+  Transports announce themselves as multipath capable by setting the
+  t_mp_capable bit during registration with the rds core module. When the
+  transport is multipath-capable, rds_sendmsg() hashes outgoing traffic
+  across multiple paths. The outgoing hash is computed based on the
+  local address and port that the PF_RDS socket is bound to.
+
+  Additionally, even if the transport is MP capable, we may be
+  peering with some node that does not support mprds, or supports
+  a different number of paths. As a result, the peering nodes need
+  to agree on the number of paths to be used for the connection.
+  This is done by sending out a control packet exchange before the
+  first data packet. The control packet exchange must have completed
+  prior to outgoing hash completion in rds_sendmsg() when the transport
+  is mutlipath capable.
+
+  The control packet is an RDS ping packet (i.e., packet to rds dest
+  port 0) with the ping packet having a rds extension header option  of
+  type RDS_EXTHDR_NPATHS, length 2 bytes, and the value is the
+  number of paths supported by the sender. The "probe" ping packet will
+  get sent from some reserved port, RDS_FLAG_PROBE_PORT (in <linux/rds.h>)
+  The receiver of a ping from RDS_FLAG_PROBE_PORT will thus immediately
+  be able to compute the min(sender_paths, rcvr_paths). The pong
+  sent in response to a probe-ping should contain the rcvr's npaths
+  when the rcvr is mprds-capable.
+
+  If the rcvr is not mprds-capable, the exthdr in the ping will be
+  ignored.  In this case the pong will not have any exthdrs, so the sender
+  of the probe-ping can default to single-path mprds.
 

net/rds/bind.c

@@ -81,6 +81,8 @@ static int rds_add_bound(struct rds_sock *rs, __be32 addr, __be16 *port)
 
 	if (*port != 0) {
 		rover = be16_to_cpu(*port);
+		if (rover == RDS_FLAG_PROBE_PORT)
+			return -EINVAL;
 		last = rover;
 	} else {
 		rover = max_t(u16, prandom_u32(), 2);
@@ -91,12 +93,16 @@ static int rds_add_bound(struct rds_sock *rs, __be32 addr, __be16 *port)
 		if (rover == 0)
 			rover++;
 
+		if (rover == RDS_FLAG_PROBE_PORT)
+			continue;
 		key = ((u64)addr << 32) | cpu_to_be16(rover);
 		if (rhashtable_lookup_fast(&bind_hash_table, &key, ht_parms))
 			continue;
 
 		rs->rs_bound_key = key;
 		rs->rs_bound_addr = addr;
+		net_get_random_once(&rs->rs_hash_initval,
+				    sizeof(rs->rs_hash_initval));
 		rs->rs_bound_port = cpu_to_be16(rover);
 		rs->rs_bound_node.next = NULL;
 		rds_sock_addref(rs);

net/rds/connection.c

@@ -155,7 +155,7 @@ static struct rds_connection *__rds_conn_create(struct net *net,
 	struct hlist_head *head = rds_conn_bucket(laddr, faddr);
 	struct rds_transport *loop_trans;
 	unsigned long flags;
-	int ret;
+	int ret, i;
 
 	rcu_read_lock();
 	conn = rds_conn_lookup(net, head, laddr, faddr, trans);
@@ -211,6 +211,12 @@ static struct rds_connection *__rds_conn_create(struct net *net,
 
 	conn->c_trans = trans;
 
+	init_waitqueue_head(&conn->c_hs_waitq);
+	for (i = 0; i < RDS_MPATH_WORKERS; i++) {
+		__rds_conn_path_init(conn, &conn->c_path[i],
+				     is_outgoing);
+		conn->c_path[i].cp_index = i;
+	}
 	ret = trans->conn_alloc(conn, gfp);
 	if (ret) {
 		kmem_cache_free(rds_conn_slab, conn);
@@ -263,14 +269,6 @@ static struct rds_connection *__rds_conn_create(struct net *net,
 			kmem_cache_free(rds_conn_slab, conn);
 			conn = found;
 		} else {
-			int i;
-
-			for (i = 0; i < RDS_MPATH_WORKERS; i++) {
-				__rds_conn_path_init(conn, &conn->c_path[i],
-						     is_outgoing);
-				conn->c_path[i].cp_index = i;
-			}
-
 			hlist_add_head_rcu(&conn->c_hash_node, head);
 			rds_cong_add_conn(conn);
 			rds_conn_count++;
@@ -668,6 +666,7 @@ EXPORT_SYMBOL_GPL(rds_conn_path_drop);
 
 void rds_conn_drop(struct rds_connection *conn)
 {
+	WARN_ON(conn->c_trans->t_mp_capable);
 	rds_conn_path_drop(&conn->c_path[0]);
 }
 EXPORT_SYMBOL_GPL(rds_conn_drop);

net/rds/message.c

@@ -41,6 +41,7 @@ static unsigned int	rds_exthdr_size[__RDS_EXTHDR_MAX] = {
 [RDS_EXTHDR_VERSION]	= sizeof(struct rds_ext_header_version),
 [RDS_EXTHDR_RDMA]	= sizeof(struct rds_ext_header_rdma),
 [RDS_EXTHDR_RDMA_DEST]	= sizeof(struct rds_ext_header_rdma_dest),
+[RDS_EXTHDR_NPATHS]	= sizeof(u16),
 };
 
 

net/rds/rds.h

@@ -85,7 +85,9 @@ enum {
 #define RDS_RECV_REFILL		3
 
 /* Max number of multipaths per RDS connection. Must be a power of 2 */
-#define	RDS_MPATH_WORKERS	1
+#define	RDS_MPATH_WORKERS	8
+#define	RDS_MPATH_HASH(rs, n) (jhash_1word((rs)->rs_bound_port, \
+			       (rs)->rs_hash_initval) & ((n) - 1))
 
 /* Per mpath connection state */
 struct rds_conn_path {
@@ -131,7 +133,8 @@ struct rds_connection {
 	__be32			c_laddr;
 	__be32			c_faddr;
 	unsigned int		c_loopback:1,
-				c_pad_to_32:31;
+				c_ping_triggered:1,
+				c_pad_to_32:30;
 	int			c_npaths;
 	struct rds_connection	*c_passive;
 	struct rds_transport	*c_trans;
@@ -147,6 +150,7 @@ struct rds_connection {
 	unsigned long		c_map_queued;
 
 	struct rds_conn_path	c_path[RDS_MPATH_WORKERS];
+	wait_queue_head_t	c_hs_waitq; /* handshake waitq */
 };
 
 static inline
@@ -166,6 +170,17 @@ void rds_conn_net_set(struct rds_connection *conn, struct net *net)
 #define RDS_FLAG_RETRANSMITTED	0x04
 #define RDS_MAX_ADV_CREDIT	255
 
+/* RDS_FLAG_PROBE_PORT is the reserved sport used for sending a ping
+ * probe to exchange control information before establishing a connection.
+ * Currently the control information that is exchanged is the number of
+ * supported paths. If the peer is a legacy (older kernel revision) peer,
+ * it would return a pong message without additional control information
+ * that would then alert the sender that the peer was an older rev.
+ */
+#define RDS_FLAG_PROBE_PORT	1
+#define	RDS_HS_PROBE(sport, dport) \
+		((sport == RDS_FLAG_PROBE_PORT && dport == 0) || \
+		 (sport == 0 && dport == RDS_FLAG_PROBE_PORT))
 /*
  * Maximum space available for extension headers.
  */
@@ -225,6 +240,11 @@ struct rds_ext_header_rdma_dest {
 	__be32			h_rdma_offset;
 };
 
+/* Extension header announcing number of paths.
+ * Implicit length = 2 bytes.
+ */
+#define RDS_EXTHDR_NPATHS	4
+
 #define __RDS_EXTHDR_MAX	16 /* for now */
 
 struct rds_incoming {
@@ -545,6 +565,7 @@ struct rds_sock {
 	/* Socket options - in case there will be more */
 	unsigned char		rs_recverr,
 				rs_cong_monitor;
+	u32			rs_hash_initval;
 };
 
 static inline struct rds_sock *rds_sk_to_rs(const struct sock *sk)

net/rds/recv.c

@@ -156,6 +156,67 @@ static void rds_recv_incoming_exthdrs(struct rds_incoming *inc, struct rds_sock
 	}
 }
 
+static void rds_recv_hs_exthdrs(struct rds_header *hdr,
+				struct rds_connection *conn)
+{
+	unsigned int pos = 0, type, len;
+	union {
+		struct rds_ext_header_version version;
+		u16 rds_npaths;
+	} buffer;
+
+	while (1) {
+		len = sizeof(buffer);
+		type = rds_message_next_extension(hdr, &pos, &buffer, &len);
+		if (type == RDS_EXTHDR_NONE)
+			break;
+		/* Process extension header here */
+		switch (type) {
+		case RDS_EXTHDR_NPATHS:
+			conn->c_npaths = min_t(int, RDS_MPATH_WORKERS,
+					       buffer.rds_npaths);
+			break;
+		default:
+			pr_warn_ratelimited("ignoring unknown exthdr type "
+					     "0x%x\n", type);
+		}
+	}
+	/* if RDS_EXTHDR_NPATHS was not found, default to a single-path */
+	conn->c_npaths = max_t(int, conn->c_npaths, 1);
+}
+
+/* rds_start_mprds() will synchronously start multiple paths when appropriate.
+ * The scheme is based on the following rules:
+ *
+ * 1. rds_sendmsg on first connect attempt sends the probe ping, with the
+ *    sender's npaths (s_npaths)
+ * 2. rcvr of probe-ping knows the mprds_paths = min(s_npaths, r_npaths). It
+ *    sends back a probe-pong with r_npaths. After that, if rcvr is the
+ *    smaller ip addr, it starts rds_conn_path_connect_if_down on all
+ *    mprds_paths.
+ * 3. sender gets woken up, and can move to rds_conn_path_connect_if_down.
+ *    If it is the smaller ipaddr, rds_conn_path_connect_if_down can be
+ *    called after reception of the probe-pong on all mprds_paths.
+ *    Otherwise (sender of probe-ping is not the smaller ip addr): just call
+ *    rds_conn_path_connect_if_down on the hashed path. (see rule 4)
+ * 4. when cp_index > 0, rds_connect_worker must only trigger
+ *    a connection if laddr < faddr.
+ * 5. sender may end up queuing the packet on the cp. will get sent out later.
+ *    when connection is completed.
+ */
+static void rds_start_mprds(struct rds_connection *conn)
+{
+	int i;
+	struct rds_conn_path *cp;
+
+	if (conn->c_npaths > 1 && conn->c_laddr < conn->c_faddr) {
+		for (i = 1; i < conn->c_npaths; i++) {
+			cp = &conn->c_path[i];
+			rds_conn_path_connect_if_down(cp);
+		}
+	}
+}
+
 /*
  * The transport must make sure that this is serialized against other
  * rx and conn reset on this specific conn.
@@ -232,6 +293,20 @@ void rds_recv_incoming(struct rds_connection *conn, __be32 saddr, __be32 daddr,
 		}
 		rds_stats_inc(s_recv_ping);
 		rds_send_pong(cp, inc->i_hdr.h_sport);
+		/* if this is a handshake ping, start multipath if necessary */
+		if (RDS_HS_PROBE(inc->i_hdr.h_sport, inc->i_hdr.h_dport)) {
+			rds_recv_hs_exthdrs(&inc->i_hdr, cp->cp_conn);
+			rds_start_mprds(cp->cp_conn);
+		}
+		goto out;
+	}
+
+	if (inc->i_hdr.h_dport ==  RDS_FLAG_PROBE_PORT &&
+	    inc->i_hdr.h_sport == 0) {
+		rds_recv_hs_exthdrs(&inc->i_hdr, cp->cp_conn);
+		/* if this is a handshake pong, start multipath if necessary */
+		rds_start_mprds(cp->cp_conn);
+		wake_up(&cp->cp_conn->c_hs_waitq);
 		goto out;
 	}