Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (76 commits)
  IB: Update MAINTAINERS with Hal's new email address
  IB/mlx4: Implement query SRQ
  IB/mlx4: Implement query QP
  IB/cm: Send no match if a SIDR REQ does not match a listen
  IB/cm: Fix handling of duplicate SIDR REQs
  IB/cm: cm_msgs.h should include ib_cm.h
  IB/cm: Include HCA ACK delay in local ACK timeout
  IB/cm: Use spin_lock_irq() instead of spin_lock_irqsave() when possible
  IB/sa: Make sure SA queries use default P_Key
  IPoIB: Recycle loopback skbs instead of freeing and reallocating
  IB/mthca: Replace memset(<addr>, 0, PAGE_SIZE) with clear_page(<addr>)
  IPoIB/cm: Fix warning if IPV6 is not enabled
  IB/core: Take sizeof the correct pointer when calling kmalloc()
  IB/ehca: Improve latency by unlocking after triggering the hardware
  IB/ehca: Notify consumers of LID/PKEY/SM changes after nondisruptive events
  IB/ehca: Return QP pointer in poll_cq()
  IB/ehca: Change idr spinlocks into rwlocks
  IB/ehca: Refactor sync between completions and destroy_cq using atomic_t
  IB/ehca: Lock renaming, static initializers
  IB/ehca: Report RDMA atomic attributes in query_qp()
  ...

13 files changed, 262 insertions, 169 deletions

diff --git a/drivers/infiniband/core/agent.c b/drivers/infiniband/core/agent.c
index ecd1a3057c6..db2633e4aae 100644
--- a/drivers/infiniband/core/agent.c
+++ b/drivers/infiniband/core/agent.c
@@ -3,7 +3,7 @@
  * Copyright (c) 2004, 2005 Infinicon Corporation.  All rights reserved.
  * Copyright (c) 2004, 2005 Intel Corporation.  All rights reserved.
  * Copyright (c) 2004, 2005 Topspin Corporation.  All rights reserved.
- * Copyright (c) 2004, 2005 Voltaire Corporation.  All rights reserved.
+ * Copyright (c) 2004-2007 Voltaire Corporation.  All rights reserved.
  * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved.
  *
  * This software is available to you under a choice of one of two
@@ -34,7 +34,6 @@
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  *
- * $Id: agent.c 1389 2004-12-27 22:56:47Z roland $
  */
 
 #include <linux/slab.h>
@@ -42,6 +41,7 @@
 
 #include "agent.h"
 #include "smi.h"
+#include "mad_priv.h"
 
 #define SPFX "ib_agent: "
 
@@ -87,8 +87,13 @@ int agent_send_response(struct ib_mad *mad, struct ib_grh *grh,
 	struct ib_mad_send_buf *send_buf;
 	struct ib_ah *ah;
 	int ret;
+	struct ib_mad_send_wr_private *mad_send_wr;
+
+	if (device->node_type == RDMA_NODE_IB_SWITCH)
+		port_priv = ib_get_agent_port(device, 0);
+	else
+		port_priv = ib_get_agent_port(device, port_num);
 
-	port_priv = ib_get_agent_port(device, port_num);
 	if (!port_priv) {
 		printk(KERN_ERR SPFX "Unable to find port agent\n");
 		return -ENODEV;
@@ -113,6 +118,14 @@ int agent_send_response(struct ib_mad *mad, struct ib_grh *grh,
 
 	memcpy(send_buf->mad, mad, sizeof *mad);
 	send_buf->ah = ah;
+
+	if (device->node_type == RDMA_NODE_IB_SWITCH) {
+		mad_send_wr = container_of(send_buf,
+					   struct ib_mad_send_wr_private,
+					   send_buf);
+		mad_send_wr->send_wr.wr.ud.port_num = port_num;
+	}
+
 	if ((ret = ib_post_send_mad(send_buf, NULL))) {
 		printk(KERN_ERR SPFX "ib_post_send_mad error:%d\n", ret);
 		goto err2;
diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
index 40c004a2697..9820c67ba47 100644
--- a/drivers/infiniband/core/cm.c
+++ b/drivers/infiniband/core/cm.c
@@ -87,6 +87,7 @@ struct cm_port {
 struct cm_device {
 	struct list_head list;
 	struct ib_device *device;
+	u8 ack_delay;
 	struct cm_port port[0];
 };
 
@@ -95,7 +96,7 @@ struct cm_av {
 	union ib_gid dgid;
 	struct ib_ah_attr ah_attr;
 	u16 pkey_index;
-	u8 packet_life_time;
+	u8 timeout;
 };
 
 struct cm_work {
@@ -154,6 +155,7 @@ struct cm_id_private {
 	u8 retry_count;
 	u8 rnr_retry_count;
 	u8 service_timeout;
+	u8 target_ack_delay;
 
 	struct list_head work_list;
 	atomic_t work_count;
@@ -293,7 +295,7 @@ static int cm_init_av_by_path(struct ib_sa_path_rec *path, struct cm_av *av)
 	av->port = port;
 	ib_init_ah_from_path(cm_dev->device, port->port_num, path,
 			     &av->ah_attr);
-	av->packet_life_time = path->packet_life_time;
+	av->timeout = path->packet_life_time + 1;
 	return 0;
 }
 
@@ -318,12 +320,10 @@ static int cm_alloc_id(struct cm_id_private *cm_id_priv)
 
 static void cm_free_id(__be32 local_id)
 {
-	unsigned long flags;
-
-	spin_lock_irqsave(&cm.lock, flags);
+	spin_lock_irq(&cm.lock);
 	idr_remove(&cm.local_id_table,
 		   (__force int) (local_id ^ cm.random_id_operand));
-	spin_unlock_irqrestore(&cm.lock, flags);
+	spin_unlock_irq(&cm.lock);
 }
 
 static struct cm_id_private * cm_get_id(__be32 local_id, __be32 remote_id)
@@ -345,11 +345,10 @@ static struct cm_id_private * cm_get_id(__be32 local_id, __be32 remote_id)
 static struct cm_id_private * cm_acquire_id(__be32 local_id, __be32 remote_id)
 {
 	struct cm_id_private *cm_id_priv;
-	unsigned long flags;
 
-	spin_lock_irqsave(&cm.lock, flags);
+	spin_lock_irq(&cm.lock);
 	cm_id_priv = cm_get_id(local_id, remote_id);
-	spin_unlock_irqrestore(&cm.lock, flags);
+	spin_unlock_irq(&cm.lock);
 
 	return cm_id_priv;
 }
@@ -646,6 +645,25 @@ static inline int cm_convert_to_ms(int iba_time)
 	return 1 << max(iba_time - 8, 0);
 }
 
+/*
+ * calculate: 4.096x2^ack_timeout = 4.096x2^ack_delay + 2x4.096x2^life_time
+ * Because of how ack_timeout is stored, adding one doubles the timeout.
+ * To avoid large timeouts, select the max(ack_delay, life_time + 1), and
+ * increment it (round up) only if the other is within 50%.
+ */
+static u8 cm_ack_timeout(u8 ca_ack_delay, u8 packet_life_time)
+{
+	int ack_timeout = packet_life_time + 1;
+
+	if (ack_timeout >= ca_ack_delay)
+		ack_timeout += (ca_ack_delay >= (ack_timeout - 1));
+	else
+		ack_timeout = ca_ack_delay +
+			      (ack_timeout >= (ca_ack_delay - 1));
+
+	return min(31, ack_timeout);
+}
+
 static void cm_cleanup_timewait(struct cm_timewait_info *timewait_info)
 {
 	if (timewait_info->inserted_remote_id) {
@@ -689,7 +707,7 @@ static void cm_enter_timewait(struct cm_id_private *cm_id_priv)
 	 * timewait before notifying the user that we've exited timewait.
 	 */
 	cm_id_priv->id.state = IB_CM_TIMEWAIT;
-	wait_time = cm_convert_to_ms(cm_id_priv->av.packet_life_time + 1);
+	wait_time = cm_convert_to_ms(cm_id_priv->av.timeout);
 	queue_delayed_work(cm.wq, &cm_id_priv->timewait_info->work.work,
 			   msecs_to_jiffies(wait_time));
 	cm_id_priv->timewait_info = NULL;
@@ -713,31 +731,30 @@ static void cm_destroy_id(struct ib_cm_id *cm_id, int err)
 {
 	struct cm_id_private *cm_id_priv;
 	struct cm_work *work;
-	unsigned long flags;
 
 	cm_id_priv = container_of(cm_id, struct cm_id_private, id);
 retest:
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	switch (cm_id->state) {
 	case IB_CM_LISTEN:
 		cm_id->state = IB_CM_IDLE;
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
-		spin_lock_irqsave(&cm.lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
+		spin_lock_irq(&cm.lock);
 		rb_erase(&cm_id_priv->service_node, &cm.listen_service_table);
-		spin_unlock_irqrestore(&cm.lock, flags);
+		spin_unlock_irq(&cm.lock);
 		break;
 	case IB_CM_SIDR_REQ_SENT:
 		cm_id->state = IB_CM_IDLE;
 		ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg);
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		break;
 	case IB_CM_SIDR_REQ_RCVD:
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		cm_reject_sidr_req(cm_id_priv, IB_SIDR_REJECT);
 		break;
 	case IB_CM_REQ_SENT:
 		ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg);
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		ib_send_cm_rej(cm_id, IB_CM_REJ_TIMEOUT,
 			       &cm_id_priv->id.device->node_guid,
 			       sizeof cm_id_priv->id.device->node_guid,
@@ -747,9 +764,9 @@ retest:
 		if (err == -ENOMEM) {
 			/* Do not reject to allow future retries. */
 			cm_reset_to_idle(cm_id_priv);
-			spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+			spin_unlock_irq(&cm_id_priv->lock);
 		} else {
-			spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+			spin_unlock_irq(&cm_id_priv->lock);
 			ib_send_cm_rej(cm_id, IB_CM_REJ_CONSUMER_DEFINED,
 				       NULL, 0, NULL, 0);
 		}
@@ -762,25 +779,25 @@ retest:
 	case IB_CM_MRA_REQ_SENT:
 	case IB_CM_REP_RCVD:
 	case IB_CM_MRA_REP_SENT:
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		ib_send_cm_rej(cm_id, IB_CM_REJ_CONSUMER_DEFINED,
 			       NULL, 0, NULL, 0);
 		break;
 	case IB_CM_ESTABLISHED:
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		ib_send_cm_dreq(cm_id, NULL, 0);
 		goto retest;
 	case IB_CM_DREQ_SENT:
 		ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg);
 		cm_enter_timewait(cm_id_priv);
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		break;
 	case IB_CM_DREQ_RCVD:
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		ib_send_cm_drep(cm_id, NULL, 0);
 		break;
 	default:
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		break;
 	}
 
@@ -912,7 +929,8 @@ static void cm_format_req(struct cm_req_msg *req_msg,
 	cm_req_set_primary_sl(req_msg, param->primary_path->sl);
 	cm_req_set_primary_subnet_local(req_msg, 1); /* local only... */
 	cm_req_set_primary_local_ack_timeout(req_msg,
-		min(31, param->primary_path->packet_life_time + 1));
+		cm_ack_timeout(cm_id_priv->av.port->cm_dev->ack_delay,
+			       param->primary_path->packet_life_time));
 
 	if (param->alternate_path) {
 		req_msg->alt_local_lid = param->alternate_path->slid;
@@ -927,7 +945,8 @@ static void cm_format_req(struct cm_req_msg *req_msg,
 		cm_req_set_alt_sl(req_msg, param->alternate_path->sl);
 		cm_req_set_alt_subnet_local(req_msg, 1); /* local only... */
 		cm_req_set_alt_local_ack_timeout(req_msg,
-			min(31, param->alternate_path->packet_life_time + 1));
+			cm_ack_timeout(cm_id_priv->av.port->cm_dev->ack_delay,
+				       param->alternate_path->packet_life_time));
 	}
 
 	if (param->private_data && param->private_data_len)
@@ -1169,7 +1188,6 @@ static void cm_format_req_event(struct cm_work *work,
 static void cm_process_work(struct cm_id_private *cm_id_priv,
 			    struct cm_work *work)
 {
-	unsigned long flags;
 	int ret;
 
 	/* We will typically only have the current event to report. */
@@ -1177,9 +1195,9 @@ static void cm_process_work(struct cm_id_private *cm_id_priv,
 	cm_free_work(work);
 
 	while (!ret && !atomic_add_negative(-1, &cm_id_priv->work_count)) {
-		spin_lock_irqsave(&cm_id_priv->lock, flags);
+		spin_lock_irq(&cm_id_priv->lock);
 		work = cm_dequeue_work(cm_id_priv);
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		BUG_ON(!work);
 		ret = cm_id_priv->id.cm_handler(&cm_id_priv->id,
 						&work->cm_event);
@@ -1250,7 +1268,6 @@ static void cm_dup_req_handler(struct cm_work *work,
 			       struct cm_id_private *cm_id_priv)
 {
 	struct ib_mad_send_buf *msg = NULL;
-	unsigned long flags;
 	int ret;
 
 	/* Quick state check to discard duplicate REQs. */
@@ -1261,7 +1278,7 @@ static void cm_dup_req_handler(struct cm_work *work,
 	if (ret)
 		return;
 
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	switch (cm_id_priv->id.state) {
 	case IB_CM_MRA_REQ_SENT:
 		cm_format_mra((struct cm_mra_msg *) msg->mad, cm_id_priv,
@@ -1276,14 +1293,14 @@ static void cm_dup_req_handler(struct cm_work *work,
 	default:
 		goto unlock;
 	}
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret)
 		goto free;
 	return;
 
-unlock:	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+unlock:	spin_unlock_irq(&cm_id_priv->lock);
 free:	cm_free_msg(msg);
 }
 
@@ -1293,17 +1310,16 @@ static struct cm_id_private * cm_match_req(struct cm_work *work,
 	struct cm_id_private *listen_cm_id_priv, *cur_cm_id_priv;
 	struct cm_timewait_info *timewait_info;
 	struct cm_req_msg *req_msg;
-	unsigned long flags;
 
 	req_msg = (struct cm_req_msg *)work->mad_recv_wc->recv_buf.mad;
 
 	/* Check for possible duplicate REQ. */
-	spin_lock_irqsave(&cm.lock, flags);
+	spin_lock_irq(&cm.lock);
 	timewait_info = cm_insert_remote_id(cm_id_priv->timewait_info);
 	if (timewait_info) {
 		cur_cm_id_priv = cm_get_id(timewait_info->work.local_id,
 					   timewait_info->work.remote_id);
-		spin_unlock_irqrestore(&cm.lock, flags);
+		spin_unlock_irq(&cm.lock);
 		if (cur_cm_id_priv) {
 			cm_dup_req_handler(work, cur_cm_id_priv);
 			cm_deref_id(cur_cm_id_priv);
@@ -1315,7 +1331,7 @@ static struct cm_id_private * cm_match_req(struct cm_work *work,
 	timewait_info = cm_insert_remote_qpn(cm_id_priv->timewait_info);
 	if (timewait_info) {
 		cm_cleanup_timewait(cm_id_priv->timewait_info);
-		spin_unlock_irqrestore(&cm.lock, flags);
+		spin_unlock_irq(&cm.lock);
 		cm_issue_rej(work->port, work->mad_recv_wc,
 			     IB_CM_REJ_STALE_CONN, CM_MSG_RESPONSE_REQ,
 			     NULL, 0);
@@ -1328,7 +1344,7 @@ static struct cm_id_private * cm_match_req(struct cm_work *work,
 					   req_msg->private_data);
 	if (!listen_cm_id_priv) {
 		cm_cleanup_timewait(cm_id_priv->timewait_info);
-		spin_unlock_irqrestore(&cm.lock, flags);
+		spin_unlock_irq(&cm.lock);
 		cm_issue_rej(work->port, work->mad_recv_wc,
 			     IB_CM_REJ_INVALID_SERVICE_ID, CM_MSG_RESPONSE_REQ,
 			     NULL, 0);
@@ -1338,7 +1354,7 @@ static struct cm_id_private * cm_match_req(struct cm_work *work,
 	atomic_inc(&cm_id_priv->refcount);
 	cm_id_priv->id.state = IB_CM_REQ_RCVD;
 	atomic_inc(&cm_id_priv->work_count);
-	spin_unlock_irqrestore(&cm.lock, flags);
+	spin_unlock_irq(&cm.lock);
 out:
 	return listen_cm_id_priv;
 }
@@ -1440,7 +1456,8 @@ static void cm_format_rep(struct cm_rep_msg *rep_msg,
 	cm_rep_set_starting_psn(rep_msg, cpu_to_be32(param->starting_psn));
 	rep_msg->resp_resources = param->responder_resources;
 	rep_msg->initiator_depth = param->initiator_depth;
-	cm_rep_set_target_ack_delay(rep_msg, param->target_ack_delay);
+	cm_rep_set_target_ack_delay(rep_msg,
+				    cm_id_priv->av.port->cm_dev->ack_delay);
 	cm_rep_set_failover(rep_msg, param->failover_accepted);
 	cm_rep_set_flow_ctrl(rep_msg, param->flow_control);
 	cm_rep_set_rnr_retry_count(rep_msg, param->rnr_retry_count);
@@ -1591,7 +1608,6 @@ static void cm_dup_rep_handler(struct cm_work *work)
 	struct cm_id_private *cm_id_priv;
 	struct cm_rep_msg *rep_msg;
 	struct ib_mad_send_buf *msg = NULL;
-	unsigned long flags;
 	int ret;
 
 	rep_msg = (struct cm_rep_msg *) work->mad_recv_wc->recv_buf.mad;
@@ -1604,7 +1620,7 @@ static void cm_dup_rep_handler(struct cm_work *work)
 	if (ret)
 		goto deref;
 
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	if (cm_id_priv->id.state == IB_CM_ESTABLISHED)
 		cm_format_rtu((struct cm_rtu_msg *) msg->mad, cm_id_priv,
 			      cm_id_priv->private_data,
@@ -1616,14 +1632,14 @@ static void cm_dup_rep_handler(struct cm_work *work)
 			      cm_id_priv->private_data_len);
 	else
 		goto unlock;
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret)
 		goto free;
 	goto deref;
 
-unlock:	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+unlock:	spin_unlock_irq(&cm_id_priv->lock);
 free:	cm_free_msg(msg);
 deref:	cm_deref_id(cm_id_priv);
 }
@@ -1632,7 +1648,6 @@ static int cm_rep_handler(struct cm_work *work)
 {
 	struct cm_id_private *cm_id_priv;
 	struct cm_rep_msg *rep_msg;
-	unsigned long flags;
 	int ret;
 
 	rep_msg = (struct cm_rep_msg *)work->mad_recv_wc->recv_buf.mad;
@@ -1644,13 +1659,13 @@ static int cm_rep_handler(struct cm_work *work)
 
 	cm_format_rep_event(work);
 
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	switch (cm_id_priv->id.state) {
 	case IB_CM_REQ_SENT:
 	case IB_CM_MRA_REQ_RCVD:
 		break;
 	default:
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		ret = -EINVAL;
 		goto error;
 	}
@@ -1663,7 +1678,7 @@ static int cm_rep_handler(struct cm_work *work)
 	/* Check for duplicate REP. */
 	if (cm_insert_remote_id(cm_id_priv->timewait_info)) {
 		spin_unlock(&cm.lock);
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		ret = -EINVAL;
 		goto error;
 	}
@@ -1673,7 +1688,7 @@ static int cm_rep_handler(struct cm_work *work)
 			 &cm.remote_id_table);
 		cm_id_priv->timewait_info->inserted_remote_id = 0;
 		spin_unlock(&cm.lock);
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		cm_issue_rej(work->port, work->mad_recv_wc,
 			     IB_CM_REJ_STALE_CONN, CM_MSG_RESPONSE_REP,
 			     NULL, 0);
@@ -1689,6 +1704,13 @@ static int cm_rep_handler(struct cm_work *work)
 	cm_id_priv->responder_resources = rep_msg->initiator_depth;
 	cm_id_priv->sq_psn = cm_rep_get_starting_psn(rep_msg);
 	cm_id_priv->rnr_retry_count = cm_rep_get_rnr_retry_count(rep_msg);
+	cm_id_priv->target_ack_delay = cm_rep_get_target_ack_delay(rep_msg);
+	cm_id_priv->av.timeout =
+			cm_ack_timeout(cm_id_priv->target_ack_delay,
+				       cm_id_priv->av.timeout - 1);
+	cm_id_priv->alt_av.timeout =
+			cm_ack_timeout(cm_id_priv->target_ack_delay,
+				       cm_id_priv->alt_av.timeout - 1);
 
 	/* todo: handle peer_to_peer */
 
@@ -1696,7 +1718,7 @@ static int cm_rep_handler(struct cm_work *work)
 	ret = atomic_inc_and_test(&cm_id_priv->work_count);
 	if (!ret)
 		list_add_tail(&work->list, &cm_id_priv->work_list);
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 
 	if (ret)
 		cm_process_work(cm_id_priv, work);
@@ -1712,7 +1734,6 @@ error:
 static int cm_establish_handler(struct cm_work *work)
 {
 	struct cm_id_private *cm_id_priv;
-	unsigned long flags;
 	int ret;
 
 	/* See comment in cm_establish about lookup. */
@@ -1720,9 +1741,9 @@ static int cm_establish_handler(struct cm_work *work)
 	if (!cm_id_priv)
 		return -EINVAL;
 
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	if (cm_id_priv->id.state != IB_CM_ESTABLISHED) {
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		goto out;
 	}
 
@@ -1730,7 +1751,7 @@ static int cm_establish_handler(struct cm_work *work)
 	ret = atomic_inc_and_test(&cm_id_priv->work_count);
 	if (!ret)
 		list_add_tail(&work->list, &cm_id_priv->work_list);
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 
 	if (ret)
 		cm_process_work(cm_id_priv, work);
@@ -1746,7 +1767,6 @@ static int cm_rtu_handler(struct cm_work *work)
 {
 	struct cm_id_private *cm_id_priv;
 	struct cm_rtu_msg *rtu_msg;
-	unsigned long flags;
 	int ret;
 
 	rtu_msg = (struct cm_rtu_msg *)work->mad_recv_wc->recv_buf.mad;
@@ -1757,10 +1777,10 @@ static int cm_rtu_handler(struct cm_work *work)
 
 	work->cm_event.private_data = &rtu_msg->private_data;
 
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	if (cm_id_priv->id.state != IB_CM_REP_SENT &&
 	    cm_id_priv->id.state != IB_CM_MRA_REP_RCVD) {
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		goto out;
 	}
 	cm_id_priv->id.state = IB_CM_ESTABLISHED;
@@ -1769,7 +1789,7 @@ static int cm_rtu_handler(struct cm_work *work)
 	ret = atomic_inc_and_test(&cm_id_priv->work_count);
 	if (!ret)
 		list_add_tail(&work->list, &cm_id_priv->work_list);
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 
 	if (ret)
 		cm_process_work(cm_id_priv, work);
@@ -1932,7 +1952,6 @@ static int cm_dreq_handler(struct cm_work *work)
 	struct cm_id_private *cm_id_priv;
 	struct cm_dreq_msg *dreq_msg;
 	struct ib_mad_send_buf *msg = NULL;
-	unsigned long flags;
 	int ret;
 
 	dreq_msg = (struct cm_dreq_msg *)work->mad_recv_wc->recv_buf.mad;
@@ -1945,7 +1964,7 @@ static int cm_dreq_handler(struct cm_work *work)
 
 	work->cm_event.private_data = &dreq_msg->private_data;
 
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	if (cm_id_priv->local_qpn != cm_dreq_get_remote_qpn(dreq_msg))
 		goto unlock;
 
@@ -1964,7 +1983,7 @@ static int cm_dreq_handler(struct cm_work *work)
 		cm_format_drep((struct cm_drep_msg *) msg->mad, cm_id_priv,
 			       cm_id_priv->private_data,
 			       cm_id_priv->private_data_len);
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 
 		if (ib_post_send_mad(msg, NULL))
 			cm_free_msg(msg);
@@ -1977,7 +1996,7 @@ static int cm_dreq_handler(struct cm_work *work)
 	ret = atomic_inc_and_test(&cm_id_priv->work_count);
 	if (!ret)
 		list_add_tail(&work->list, &cm_id_priv->work_list);
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 
 	if (ret)
 		cm_process_work(cm_id_priv, work);
@@ -1985,7 +2004,7 @@ static int cm_dreq_handler(struct cm_work *work)
 		cm_deref_id(cm_id_priv);
 	return 0;
 
-unlock:	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+unlock:	spin_unlock_irq(&cm_id_priv->lock);
 deref:	cm_deref_id(cm_id_priv);
 	return -EINVAL;
 }
@@ -1994,7 +2013,6 @@ static int cm_drep_handler(struct cm_work *work)
 {
 	struct cm_id_private *cm_id_priv;
 	struct cm_drep_msg *drep_msg;
-	unsigned long flags;
 	int ret;
 
 	drep_msg = (struct cm_drep_msg *)work->mad_recv_wc->recv_buf.mad;
@@ -2005,10 +2023,10 @@ static int cm_drep_handler(struct cm_work *work)
 
 	work->cm_event.private_data = &drep_msg->private_data;
 
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	if (cm_id_priv->id.state != IB_CM_DREQ_SENT &&
 	    cm_id_priv->id.state != IB_CM_DREQ_RCVD) {
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		goto out;
 	}
 	cm_enter_timewait(cm_id_priv);
@@ -2017,7 +2035,7 @@ static int cm_drep_handler(struct cm_work *work)
 	ret = atomic_inc_and_test(&cm_id_priv->work_count);
 	if (!ret)
 		list_add_tail(&work->list, &cm_id_priv->work_list);
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 
 	if (ret)
 		cm_process_work(cm_id_priv, work);
@@ -2107,17 +2125,16 @@ static struct cm_id_private * cm_acquire_rejected_id(struct cm_rej_msg *rej_msg)
 {
 	struct cm_timewait_info *timewait_info;
 	struct cm_id_private *cm_id_priv;
-	unsigned long flags;
 	__be32 remote_id;
 
 	remote_id = rej_msg->local_comm_id;
 
 	if (__be16_to_cpu(rej_msg->reason) == IB_CM_REJ_TIMEOUT) {
-		spin_lock_irqsave(&cm.lock, flags);
+		spin_lock_irq(&cm.lock);
 		timewait_info = cm_find_remote_id( *((__be64 *) rej_msg->ari),
 						  remote_id);
 		if (!timewait_info) {
-			spin_unlock_irqrestore(&cm.lock, flags);
+			spin_unlock_irq(&cm.lock);
 			return NULL;
 		}
 		cm_id_priv = idr_find(&cm.local_id_table, (__force int)
@@ -2129,7 +2146,7 @@ static struct cm_id_private * cm_acquire_rejected_id(struct cm_rej_msg *rej_msg)
 			else
 				cm_id_priv = NULL;
 		}
-		spin_unlock_irqrestore(&cm.lock, flags);
+		spin_unlock_irq(&cm.lock);
 	} else if (cm_rej_get_msg_rejected(rej_msg) == CM_MSG_RESPONSE_REQ)
 		cm_id_priv = cm_acquire_id(rej_msg->remote_comm_id, 0);
 	else
@@ -2142,7 +2159,6 @@ static int cm_rej_handler(struct cm_work *work)
 {
 	struct cm_id_private *cm_id_priv;
 	struct cm_rej_msg *rej_msg;
-	unsigned long flags;
 	int ret;
 
 	rej_msg = (struct cm_rej_msg *)work->mad_recv_wc->recv_buf.mad;
@@ -2152,7 +2168,7 @@ static int cm_rej_handler(struct cm_work *work)
 
 	cm_format_rej_event(work);
 
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	switch (cm_id_priv->id.state) {
 	case IB_CM_REQ_SENT:
 	case IB_CM_MRA_REQ_RCVD:
@@ -2176,7 +2192,7 @@ static int cm_rej_handler(struct cm_work *work)
 		cm_enter_timewait(cm_id_priv);
 		break;
 	default:
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		ret = -EINVAL;
 		goto out;
 	}
@@ -2184,7 +2200,7 @@ static int cm_rej_handler(struct cm_work *work)
 	ret = atomic_inc_and_test(&cm_id_priv->work_count);
 	if (!ret)
 		list_add_tail(&work->list, &cm_id_priv->work_list);
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 
 	if (ret)
 		cm_process_work(cm_id_priv, work);
@@ -2295,7 +2311,6 @@ static int cm_mra_handler(struct cm_work *work)
 {
 	struct cm_id_private *cm_id_priv;
 	struct cm_mra_msg *mra_msg;
-	unsigned long flags;
 	int timeout, ret;
 
 	mra_msg = (struct cm_mra_msg *)work->mad_recv_wc->recv_buf.mad;
@@ -2307,9 +2322,9 @@ static int cm_mra_handler(struct cm_work *work)
 	work->cm_event.param.mra_rcvd.service_timeout =
 					cm_mra_get_service_timeout(mra_msg);
 	timeout = cm_convert_to_ms(cm_mra_get_service_timeout(mra_msg)) +
-		  cm_convert_to_ms(cm_id_priv->av.packet_life_time);
+		  cm_convert_to_ms(cm_id_priv->av.timeout);
 
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	switch (cm_id_priv->id.state) {
 	case IB_CM_REQ_SENT:
 		if (cm_mra_get_msg_mraed(mra_msg) != CM_MSG_RESPONSE_REQ ||
@@ -2342,7 +2357,7 @@ static int cm_mra_handler(struct cm_work *work)
 	ret = atomic_inc_and_test(&cm_id_priv->work_count);
 	if (!ret)
 		list_add_tail(&work->list, &cm_id_priv->work_list);
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 
 	if (ret)
 		cm_process_work(cm_id_priv, work);
@@ -2350,7 +2365,7 @@ static int cm_mra_handler(struct cm_work *work)
 		cm_deref_id(cm_id_priv);
 	return 0;
 out:
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 	cm_deref_id(cm_id_priv);
 	return -EINVAL;
 }
@@ -2379,7 +2394,8 @@ static void cm_format_lap(struct cm_lap_msg *lap_msg,
 	cm_lap_set_sl(lap_msg, alternate_path->sl);
 	cm_lap_set_subnet_local(lap_msg, 1); /* local only... */
 	cm_lap_set_local_ack_timeout(lap_msg,
-		min(31, alternate_path->packet_life_time + 1));
+		cm_ack_timeout(cm_id_priv->av.port->cm_dev->ack_delay,
+			       alternate_path->packet_life_time));
 
 	if (private_data && private_data_len)
 		memcpy(lap_msg->private_data, private_data, private_data_len);
@@ -2410,6 +2426,9 @@ int ib_send_cm_lap(struct ib_cm_id *cm_id,
 	ret = cm_init_av_by_path(alternate_path, &cm_id_priv->alt_av);
 	if (ret)
 		goto out;
+	cm_id_priv->alt_av.timeout =
+			cm_ack_timeout(cm_id_priv->target_ack_delay,
+				       cm_id_priv->alt_av.timeout - 1);
 
 	ret = cm_alloc_msg(cm_id_priv, &msg);
 	if (ret)
@@ -2465,7 +2484,6 @@ static int cm_lap_handler(struct cm_work *work)
 	struct cm_lap_msg *lap_msg;
 	struct ib_cm_lap_event_param *param;
 	struct ib_mad_send_buf *msg = NULL;
-	unsigned long flags;
 	int ret;
 
 	/* todo: verify LAP request and send reject APR if invalid. */
@@ -2480,7 +2498,7 @@ static int cm_lap_handler(struct cm_work *work)
 	cm_format_path_from_lap(cm_id_priv, param->alternate_path, lap_msg);
 	work->cm_event.private_data = &lap_msg->private_data;
 
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	if (cm_id_priv->id.state != IB_CM_ESTABLISHED)
 		goto unlock;
 
@@ -2497,7 +2515,7 @@ static int cm_lap_handler(struct cm_work *work)
 			      cm_id_priv->service_timeout,
 			      cm_id_priv->private_data,
 			      cm_id_priv->private_data_len);
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 
 		if (ib_post_send_mad(msg, NULL))
 			cm_free_msg(msg);
@@ -2515,7 +2533,7 @@ static int cm_lap_handler(struct cm_work *work)
 	ret = atomic_inc_and_test(&cm_id_priv->work_count);
 	if (!ret)
 		list_add_tail(&work->list, &cm_id_priv->work_list);
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 
 	if (ret)
 		cm_process_work(cm_id_priv, work);
@@ -2523,7 +2541,7 @@ static int cm_lap_handler(struct cm_work *work)
 		cm_deref_id(cm_id_priv);
 	return 0;
 
-unlock:	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+unlock:	spin_unlock_irq(&cm_id_priv->lock);
 deref:	cm_deref_id(cm_id_priv);
 	return -EINVAL;
 }
@@ -2598,7 +2616,6 @@ static int cm_apr_handler(struct cm_work *work)
 {
 	struct cm_id_private *cm_id_priv;
 	struct cm_apr_msg *apr_msg;
-	unsigned long flags;
 	int ret;
 
 	apr_msg = (struct cm_apr_msg *)work->mad_recv_wc->recv_buf.mad;
@@ -2612,11 +2629,11 @@ static int cm_apr_handler(struct cm_work *work)
 	work->cm_event.param.apr_rcvd.info_len = apr_msg->info_length;
 	work->cm_event.private_data = &apr_msg->private_data;
 
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	if (cm_id_priv->id.state != IB_CM_ESTABLISHED ||
 	    (cm_id_priv->id.lap_state != IB_CM_LAP_SENT &&
 	     cm_id_priv->id.lap_state != IB_CM_MRA_LAP_RCVD)) {
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		goto out;
 	}
 	cm_id_priv->id.lap_state = IB_CM_LAP_IDLE;
@@ -2626,7 +2643,7 @@ static int cm_apr_handler(struct cm_work *work)
 	ret = atomic_inc_and_test(&cm_id_priv->work_count);
 	if (!ret)
 		list_add_tail(&work->list, &cm_id_priv->work_list);
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 
 	if (ret)
 		cm_process_work(cm_id_priv, work);
@@ -2761,7 +2778,6 @@ static int cm_sidr_req_handler(struct cm_work *work)
 	struct cm_id_private *cm_id_priv, *cur_cm_id_priv;
 	struct cm_sidr_req_msg *sidr_req_msg;
 	struct ib_wc *wc;
-	unsigned long flags;
 
 	cm_id = ib_create_cm_id(work->port->cm_dev->device, NULL, NULL);
 	if (IS_ERR(cm_id))
@@ -2778,27 +2794,26 @@ static int cm_sidr_req_handler(struct cm_work *work)
 				work->mad_recv_wc->recv_buf.grh,
 				&cm_id_priv->av);
 	cm_id_priv->id.remote_id = sidr_req_msg->request_id;
-	cm_id_priv->id.state = IB_CM_SIDR_REQ_RCVD;
 	cm_id_priv->tid = sidr_req_msg->hdr.tid;
 	atomic_inc(&cm_id_priv->work_count);
 
-	spin_lock_irqsave(&cm.lock, flags);
+	spin_lock_irq(&cm.lock);
 	cur_cm_id_priv = cm_insert_remote_sidr(cm_id_priv);
 	if (cur_cm_id_priv) {
-		spin_unlock_irqrestore(&cm.lock, flags);
+		spin_unlock_irq(&cm.lock);
 		goto out; /* Duplicate message. */
 	}
+	cm_id_priv->id.state = IB_CM_SIDR_REQ_RCVD;
 	cur_cm_id_priv = cm_find_listen(cm_id->device,
 					sidr_req_msg->service_id,
 					sidr_req_msg->private_data);
 	if (!cur_cm_id_priv) {
-		rb_erase(&cm_id_priv->sidr_id_node, &cm.remote_sidr_table);
-		spin_unlock_irqrestore(&cm.lock, flags);
-		/* todo: reply with no match */
+		spin_unlock_irq(&cm.lock);
+		cm_reject_sidr_req(cm_id_priv, IB_SIDR_UNSUPPORTED);
 		goto out; /* No match. */
 	}
 	atomic_inc(&cur_cm_id_priv->refcount);
-	spin_unlock_irqrestore(&cm.lock, flags);
+	spin_unlock_irq(&cm.lock);
 
 	cm_id_priv->id.cm_handler = cur_cm_id_priv->id.cm_handler;
 	cm_id_priv->id.context = cur_cm_id_priv->id.context;
@@ -2899,7 +2914,6 @@ static int cm_sidr_rep_handler(struct cm_work *work)
 {
 	struct cm_sidr_rep_msg *sidr_rep_msg;
 	struct cm_id_private *cm_id_priv;
-	unsigned long flags;
 
 	sidr_rep_msg = (struct cm_sidr_rep_msg *)
 				work->mad_recv_wc->recv_buf.mad;
@@ -2907,14 +2921,14 @@ static int cm_sidr_rep_handler(struct cm_work *work)
 	if (!cm_id_priv)
 		return -EINVAL; /* Unmatched reply. */
 
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	if (cm_id_priv->id.state != IB_CM_SIDR_REQ_SENT) {
-		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+		spin_unlock_irq(&cm_id_priv->lock);
 		goto out;
 	}
 	cm_id_priv->id.state = IB_CM_IDLE;
 	ib_cancel_mad(cm_id_priv->av.port->mad_agent, cm_id_priv->msg);
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 
 	cm_format_sidr_rep_event(work);
 	cm_process_work(cm_id_priv, work);
@@ -2930,14 +2944,13 @@ static void cm_process_send_error(struct ib_mad_send_buf *msg,
 	struct cm_id_private *cm_id_priv;
 	struct ib_cm_event cm_event;
 	enum ib_cm_state state;
-	unsigned long flags;
 	int ret;
 
 	memset(&cm_event, 0, sizeof cm_event);
 	cm_id_priv = msg->context[0];
 
 	/* Discard old sends or ones without a response. */
-	spin_lock_irqsave(&cm_id_priv->lock, flags);
+	spin_lock_irq(&cm_id_priv->lock);
 	state = (enum ib_cm_state) (unsigned long) msg->context[1];
 	if (msg != cm_id_priv->msg || state != cm_id_priv->id.state)
 		goto discard;
@@ -2964,7 +2977,7 @@ static void cm_process_send_error(struct ib_mad_send_buf *msg,
 	default:
 		goto discard;
 	}
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 	cm_event.param.send_status = wc_status;
 
 	/* No other events can occur on the cm_id at this point. */
@@ -2974,7 +2987,7 @@ static void cm_process_send_error(struct ib_mad_send_buf *msg,
 		ib_destroy_cm_id(&cm_id_priv->id);
 	return;
 discard:
-	spin_unlock_irqrestore(&cm_id_priv->lock, flags);
+	spin_unlock_irq(&cm_id_priv->lock);
 	cm_free_msg(msg);
 }
 
@@ -3269,8 +3282,7 @@ static int cm_init_qp_rtr_attr(struct cm_id_private *cm_id_priv,
 			*qp_attr_mask |= IB_QP_ALT_PATH;
 			qp_attr->alt_port_num = cm_id_priv->alt_av.port->port_num;
 			qp_attr->alt_pkey_index = cm_id_priv->alt_av.pkey_index;
-			qp_attr->alt_timeout =
-					cm_id_priv->alt_av.packet_life_time + 1;
+			qp_attr->alt_timeout = cm_id_priv->alt_av.timeout;
 			qp_attr->alt_ah_attr = cm_id_priv->alt_av.ah_attr;
 		}
 		ret = 0;
@@ -3308,8 +3320,7 @@ static int cm_init_qp_rts_attr(struct cm_id_private *cm_id_priv,
 				*qp_attr_mask |= IB_QP_TIMEOUT | IB_QP_RETRY_CNT |
 						 IB_QP_RNR_RETRY |
 						 IB_QP_MAX_QP_RD_ATOMIC;
-				qp_attr->timeout =
-					cm_id_priv->av.packet_life_time + 1;
+				qp_attr->timeout = cm_id_priv->av.timeout;
 				qp_attr->retry_cnt = cm_id_priv->retry_count;
 				qp_attr->rnr_retry = cm_id_priv->rnr_retry_count;
 				qp_attr->max_rd_atomic =
@@ -3323,8 +3334,7 @@ static int cm_init_qp_rts_attr(struct cm_id_private *cm_id_priv,
 			*qp_attr_mask = IB_QP_ALT_PATH | IB_QP_PATH_MIG_STATE;
 			qp_attr->alt_port_num = cm_id_priv->alt_av.port->port_num;
 			qp_attr->alt_pkey_index = cm_id_priv->alt_av.pkey_index;
-			qp_attr->alt_timeout =
-				cm_id_priv->alt_av.packet_life_time + 1;
+			qp_attr->alt_timeout = cm_id_priv->alt_av.timeout;
 			qp_attr->alt_ah_attr = cm_id_priv->alt_av.ah_attr;
 			qp_attr->path_mig_state = IB_MIG_REARM;
 		}
@@ -3364,6 +3374,16 @@ int ib_cm_init_qp_attr(struct ib_cm_id *cm_id,
 }
 EXPORT_SYMBOL(ib_cm_init_qp_attr);
 
+void cm_get_ack_delay(struct cm_device *cm_dev)
+{
+	struct ib_device_attr attr;
+
+	if (ib_query_device(cm_dev->device, &attr))
+		cm_dev->ack_delay = 0; /* acks will rely on packet life time */
+	else
+		cm_dev->ack_delay = attr.local_ca_ack_delay;
+}
+
 static void cm_add_one(struct ib_device *device)
 {
 	struct cm_device *cm_dev;
@@ -3388,6 +3408,7 @@ static void cm_add_one(struct ib_device *device)
 		return;
 
 	cm_dev->device = device;
+	cm_get_ack_delay(cm_dev);
 
 	set_bit(IB_MGMT_METHOD_SEND, reg_req.method_mask);
 	for (i = 1; i <= device->phys_port_cnt; i++) {
diff --git a/drivers/infiniband/core/cm_msgs.h b/drivers/infiniband/core/cm_msgs.h
index 4d3aee90c24..aec9c7af825 100644
--- a/drivers/infiniband/core/cm_msgs.h
+++ b/drivers/infiniband/core/cm_msgs.h
@@ -35,6 +35,7 @@
 #define CM_MSGS_H
 
 #include <rdma/ib_mad.h>
+#include <rdma/ib_cm.h>
 
 /*
  * Parameters to routines below should be in network-byte order, and values
diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c
index 32a0e66d2a2..23af7a032a0 100644
--- a/drivers/infiniband/core/cma.c
+++ b/drivers/infiniband/core/cma.c
@@ -2326,7 +2326,6 @@ static int cma_accept_ib(struct rdma_id_private *id_priv,
 	rep.private_data_len = conn_param->private_data_len;
 	rep.responder_resources = conn_param->responder_resources;
 	rep.initiator_depth = conn_param->initiator_depth;
-	rep.target_ack_delay = CMA_CM_RESPONSE_TIMEOUT;
 	rep.failover_accepted = 0;
 	rep.flow_control = conn_param->flow_control;
 	rep.rnr_retry_count = conn_param->rnr_retry_count;
diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c
index 85ccf13b804..6b8faca02f8 100644
--- a/drivers/infiniband/core/mad.c
+++ b/drivers/infiniband/core/mad.c
@@ -675,10 +675,16 @@ static int handle_outgoing_dr_smp(struct ib_mad_agent_private *mad_agent_priv,
 	struct ib_mad_port_private *port_priv;
 	struct ib_mad_agent_private *recv_mad_agent = NULL;
 	struct ib_device *device = mad_agent_priv->agent.device;
-	u8 port_num = mad_agent_priv->agent.port_num;
+	u8 port_num;
 	struct ib_wc mad_wc;
 	struct ib_send_wr *send_wr = &mad_send_wr->send_wr;
 
+	if (device->node_type == RDMA_NODE_IB_SWITCH &&
+	    smp->mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE)
+		port_num = send_wr->wr.ud.port_num;
+	else
+		port_num = mad_agent_priv->agent.port_num;
+
 	/*
 	 * Directed route handling starts if the initial LID routed part of
 	 * a request or the ending LID routed part of a response is empty.
@@ -1839,6 +1845,7 @@ static void ib_mad_recv_done_handler(struct ib_mad_port_private *port_priv,
 	struct ib_mad_private *recv, *response;
 	struct ib_mad_list_head *mad_list;
 	struct ib_mad_agent_private *mad_agent;
+	int port_num;
 
 	response = kmem_cache_alloc(ib_mad_cache, GFP_KERNEL);
 	if (!response)
@@ -1872,25 +1879,50 @@ static void ib_mad_recv_done_handler(struct ib_mad_port_private *port_priv,
 	if (!validate_mad(&recv->mad.mad, qp_info->qp->qp_num))
 		goto out;
 
+	if (port_priv->device->node_type == RDMA_NODE_IB_SWITCH)
+		port_num = wc->port_num;
+	else
+		port_num = port_priv->port_num;
+
 	if (recv->mad.mad.mad_hdr.mgmt_class ==
 	    IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) {
+		enum smi_forward_action retsmi;
+
 		if (smi_handle_dr_smp_recv(&recv->mad.smp,
 					   port_priv->device->node_type,
-					   port_priv->port_num,
+					   port_num,
 					   port_priv->device->phys_port_cnt) ==
 					   IB_SMI_DISCARD)
 			goto out;
 
-		if (smi_check_forward_dr_smp(&recv->mad.smp) == IB_SMI_LOCAL)
+		retsmi = smi_check_forward_dr_smp(&recv->mad.smp);
+		if (retsmi == IB_SMI_LOCAL)
 			goto local;
 
-		if (smi_handle_dr_smp_send(&recv->mad.smp,
-					   port_priv->device->node_type,
-					   port_priv->port_num) == IB_SMI_DISCARD)
-			goto out;
+		if (retsmi == IB_SMI_SEND) { /* don't forward */
+			if (smi_handle_dr_smp_send(&recv->mad.smp,
+						   port_priv->device->node_type,
+						   port_num) == IB_SMI_DISCARD)
+				goto out;
+
+			if (smi_check_local_smp(&recv->mad.smp, port_priv->device) == IB_SMI_DISCARD)
+				goto out;
+		} else if (port_priv->device->node_type == RDMA_NODE_IB_SWITCH) {
+			/* forward case for switches */
+			memcpy(response, recv, sizeof(*response));
+			response->header.recv_wc.wc = &response->header.wc;
+			response->header.recv_wc.recv_buf.mad = &response->mad.mad;
+			response->header.recv_wc.recv_buf.grh = &response->grh;
+
+			if (!agent_send_response(&response->mad.mad,
+						 &response->grh, wc,
+						 port_priv->device,
+						 smi_get_fwd_port(&recv->mad.smp),
+						 qp_info->qp->qp_num))
+				response = NULL;
 
-		if (smi_check_local_smp(&recv->mad.smp, port_priv->device) == IB_SMI_DISCARD)
 			goto out;
+		}
 	}
 
 local:
@@ -1919,7 +1951,7 @@ local:
 				agent_send_response(&response->mad.mad,
 						    &recv->grh, wc,
 						    port_priv->device,
-						    port_priv->port_num,
+						    port_num,
 						    qp_info->qp->qp_num);
 				goto out;
 			}
diff --git a/drivers/infiniband/core/multicast.c b/drivers/infiniband/core/multicast.c
index 1e13ab42b70..15b4c4d3606 100644
--- a/drivers/infiniband/core/multicast.c
+++ b/drivers/infiniband/core/multicast.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2006 Intel Corporation.  All rights reserved.
+ * Copyright (c) 2006 Intel Corporation.  All rights reserved.
  *
  * This software is available to you under a choice of one of two
  * licenses.  You may choose to be licensed under the terms of the GNU
diff --git a/drivers/infiniband/core/sa.h b/drivers/infiniband/core/sa.h
index 24c93fd320f..b1d4bbf4ce5 100644
--- a/drivers/infiniband/core/sa.h
+++ b/drivers/infiniband/core/sa.h
@@ -1,6 +1,6 @@
 /*
  * Copyright (c) 2004 Topspin Communications.  All rights reserved.
- * Copyright (c) 2005 Voltaire, Inc.  All rights reserved.
+ * Copyright (c) 2005 Voltaire, Inc.  All rights reserved.
  * Copyright (c) 2006 Intel Corporation.  All rights reserved.
  *
  * This software is available to you under a choice of one of two
diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c
index 6469406ea9d..20ab6b3e484 100644
--- a/drivers/infiniband/core/sa_query.c
+++ b/drivers/infiniband/core/sa_query.c
@@ -1,6 +1,6 @@
 /*
  * Copyright (c) 2004 Topspin Communications.  All rights reserved.
- * Copyright (c) 2005 Voltaire, Inc.  All rights reserved.
+ * Copyright (c) 2005 Voltaire, Inc.  All rights reserved.
  * Copyright (c) 2006 Intel Corporation.  All rights reserved.
  *
  * This software is available to you under a choice of one of two
@@ -56,6 +56,7 @@ MODULE_LICENSE("Dual BSD/GPL");
 struct ib_sa_sm_ah {
 	struct ib_ah        *ah;
 	struct kref          ref;
+	u16		     pkey_index;
 	u8		     src_path_mask;
 };
 
@@ -382,6 +383,13 @@ static void update_sm_ah(struct work_struct *work)
 	kref_init(&new_ah->ref);
 	new_ah->src_path_mask = (1 << port_attr.lmc) - 1;
 
+	new_ah->pkey_index = 0;
+	if (ib_find_pkey(port->agent->device, port->port_num,
+			 IB_DEFAULT_PKEY_FULL, &new_ah->pkey_index) &&
+	    ib_find_pkey(port->agent->device, port->port_num,
+			 IB_DEFAULT_PKEY_PARTIAL, &new_ah->pkey_index))
+		printk(KERN_ERR "Couldn't find index for default PKey\n");
+
 	memset(&ah_attr, 0, sizeof ah_attr);
 	ah_attr.dlid     = port_attr.sm_lid;
 	ah_attr.sl       = port_attr.sm_sl;
@@ -512,6 +520,35 @@ int ib_init_ah_from_path(struct ib_device *device, u8 port_num,
 }
 EXPORT_SYMBOL(ib_init_ah_from_path);
 
+static int alloc_mad(struct ib_sa_query *query, gfp_t gfp_mask)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&query->port->ah_lock, flags);
+	kref_get(&query->port->sm_ah->ref);
+	query->sm_ah = query->port->sm_ah;
+	spin_unlock_irqrestore(&query->port->ah_lock, flags);
+
+	query->mad_buf = ib_create_send_mad(query->port->agent, 1,
+					    query->sm_ah->pkey_index,
+					    0, IB_MGMT_SA_HDR, IB_MGMT_SA_DATA,
+					    gfp_mask);
+	if (!query->mad_buf) {
+		kref_put(&query->sm_ah->ref, free_sm_ah);
+		return -ENOMEM;
+	}
+
+	query->mad_buf->ah = query->sm_ah->ah;
+
+	return 0;
+}
+
+static void free_mad(struct ib_sa_query *query)
+{
+	ib_free_send_mad(query->mad_buf);
+	kref_put(&query->sm_ah->ref, free_sm_ah);
+}
+
 static void init_mad(struct ib_sa_mad *mad, struct ib_mad_agent *agent)
 {
 	unsigned long flags;
@@ -548,20 +585,11 @@ retry:
 	query->mad_buf->context[0] = query;
 	query->id = id;
 
-	spin_lock_irqsave(&query->port->ah_lock, flags);
-	kref_get(&query->port->sm_ah->ref);
-	query->sm_ah = query->port->sm_ah;
-	spin_unlock_irqrestore(&query->port->ah_lock, flags);
-
-	query->mad_buf->ah = query->sm_ah->ah;
-
 	ret = ib_post_send_mad(query->mad_buf, NULL);
 	if (ret) {
 		spin_lock_irqsave(&idr_lock, flags);
 		idr_remove(&query_idr, id);
 		spin_unlock_irqrestore(&idr_lock, flags);
-
-		kref_put(&query->sm_ah->ref, free_sm_ah);
 	}
 
 	/*
@@ -647,13 +675,10 @@ int ib_sa_path_rec_get(struct ib_sa_client *client,
 	if (!query)
 		return -ENOMEM;
 
-	query->sa_query.mad_buf = ib_create_send_mad(agent, 1, 0,
-						     0, IB_MGMT_SA_HDR,
-						     IB_MGMT_SA_DATA, gfp_mask);
-	if (!query->sa_query.mad_buf) {
-		ret = -ENOMEM;
+	query->sa_query.port     = port;
+	ret = alloc_mad(&query->sa_query, gfp_mask);
+	if (ret)
 		goto err1;
-	}
 
 	ib_sa_client_get(client);
 	query->sa_query.client = client;
@@ -665,7 +690,6 @@ int ib_sa_path_rec_get(struct ib_sa_client *client,
 
 	query->sa_query.callback = callback ? ib_sa_path_rec_callback : NULL;
 	query->sa_query.release  = ib_sa_path_rec_release;
-	query->sa_query.port     = port;
 	mad->mad_hdr.method	 = IB_MGMT_METHOD_GET;
 	mad->mad_hdr.attr_id	 = cpu_to_be16(IB_SA_ATTR_PATH_REC);
 	mad->sa_hdr.comp_mask	 = comp_mask;
@@ -683,7 +707,7 @@ int ib_sa_path_rec_get(struct ib_sa_client *client,
 err2:
 	*sa_query = NULL;
 	ib_sa_client_put(query->sa_query.client);
-	ib_free_send_mad(query->sa_query.mad_buf);
+	free_mad(&query->sa_query);
 
 err1:
 	kfree(query);
@@ -773,13 +797,10 @@ int ib_sa_service_rec_query(struct ib_sa_client *client,
 	if (!query)
 		return -ENOMEM;
 
-	query->sa_query.mad_buf = ib_create_send_mad(agent, 1, 0,
-						     0, IB_MGMT_SA_HDR,
-						     IB_MGMT_SA_DATA, gfp_mask);
-	if (!query->sa_query.mad_buf) {
-		ret = -ENOMEM;
+	query->sa_query.port     = port;
+	ret = alloc_mad(&query->sa_query, gfp_mask);
+	if (ret)
 		goto err1;
-	}
 
 	ib_sa_client_get(client);
 	query->sa_query.client = client;
@@ -791,7 +812,6 @@ int ib_sa_service_rec_query(struct ib_sa_client *client,
 
 	query->sa_query.callback = callback ? ib_sa_service_rec_callback : NULL;
 	query->sa_query.release  = ib_sa_service_rec_release;
-	query->sa_query.port     = port;
 	mad->mad_hdr.method	 = method;
 	mad->mad_hdr.attr_id	 = cpu_to_be16(IB_SA_ATTR_SERVICE_REC);
 	mad->sa_hdr.comp_mask	 = comp_mask;
@@ -810,7 +830,7 @@ int ib_sa_service_rec_query(struct ib_sa_client *client,
 err2:
 	*sa_query = NULL;
 	ib_sa_client_put(query->sa_query.client);
-	ib_free_send_mad(query->sa_query.mad_buf);
+	free_mad(&query->sa_query);
 
 err1:
 	kfree(query);
@@ -869,13 +889,10 @@ int ib_sa_mcmember_rec_query(struct ib_sa_client *client,
 	if (!query)
 		return -ENOMEM;
 
-	query->sa_query.mad_buf = ib_create_send_mad(agent, 1, 0,
-						     0, IB_MGMT_SA_HDR,
-						     IB_MGMT_SA_DATA, gfp_mask);
-	if (!query->sa_query.mad_buf) {
-		ret = -ENOMEM;
+	query->sa_query.port     = port;
+	ret = alloc_mad(&query->sa_query, gfp_mask);
+	if (ret)
 		goto err1;
-	}
 
 	ib_sa_client_get(client);
 	query->sa_query.client = client;
@@ -887,7 +904,6 @@ int ib_sa_mcmember_rec_query(struct ib_sa_client *client,
 
 	query->sa_query.callback = callback ? ib_sa_mcmember_rec_callback : NULL;
 	query->sa_query.release  = ib_sa_mcmember_rec_release;
-	query->sa_query.port     = port;
 	mad->mad_hdr.method	 = method;
 	mad->mad_hdr.attr_id	 = cpu_to_be16(IB_SA_ATTR_MC_MEMBER_REC);
 	mad->sa_hdr.comp_mask	 = comp_mask;
@@ -906,7 +922,7 @@ int ib_sa_mcmember_rec_query(struct ib_sa_client *client,
 err2:
 	*sa_query = NULL;
 	ib_sa_client_put(query->sa_query.client);
-	ib_free_send_mad(query->sa_query.mad_buf);
+	free_mad(&query->sa_query);
 
 err1:
 	kfree(query);
@@ -939,8 +955,7 @@ static void send_handler(struct ib_mad_agent *agent,
 	idr_remove(&query_idr, query->id);
 	spin_unlock_irqrestore(&idr_lock, flags);
 
-	ib_free_send_mad(mad_send_wc->send_buf);
-	kref_put(&query->sm_ah->ref, free_sm_ah);
+	free_mad(query);
 	ib_sa_client_put(query->client);
 	query->release(query);
 }
diff --git a/drivers/infiniband/core/smi.c b/drivers/infiniband/core/smi.c
index 2bca753eb62..87236753bce 100644
--- a/drivers/infiniband/core/smi.c
+++ b/drivers/infiniband/core/smi.c
@@ -192,7 +192,7 @@ enum smi_action smi_handle_dr_smp_recv(struct ib_smp *smp, u8 node_type,
 			}
 			/* smp->hop_ptr updated when sending */
 			return (node_type == RDMA_NODE_IB_SWITCH ?
-				IB_SMI_HANDLE: IB_SMI_DISCARD);
+				IB_SMI_HANDLE : IB_SMI_DISCARD);
 		}
 
 		/* C14-13:4 -- hop_ptr = 0 -> give to SM */
@@ -211,7 +211,7 @@ enum smi_forward_action smi_check_forward_dr_smp(struct ib_smp *smp)
 	if (!ib_get_smp_direction(smp)) {
 		/* C14-9:2 -- intermediate hop */
 		if (hop_ptr && hop_ptr < hop_cnt)
-			return IB_SMI_SEND;
+			return IB_SMI_FORWARD;
 
 		/* C14-9:3 -- at the end of the DR segment of path */
 		if (hop_ptr == hop_cnt)
@@ -224,7 +224,7 @@ enum smi_forward_action smi_check_forward_dr_smp(struct ib_smp *smp)
 	} else {
 		/* C14-13:2  -- intermediate hop */
 		if (2 <= hop_ptr && hop_ptr <= hop_cnt)
-			return IB_SMI_SEND;
+			return IB_SMI_FORWARD;
 
 		/* C14-13:3 -- at the end of the DR segment of path */
 		if (hop_ptr == 1)
@@ -233,3 +233,13 @@ enum smi_forward_action smi_check_forward_dr_smp(struct ib_smp *smp)
 	}
 	return IB_SMI_LOCAL;
 }
+
+/*
+ * Return the forwarding port number from initial_path for outgoing SMP and
+ * from return_path for returning SMP
+ */
+int smi_get_fwd_port(struct ib_smp *smp)
+{
+	return (!ib_get_smp_direction(smp) ? smp->initial_path[smp->hop_ptr+1] :
+		smp->return_path[smp->hop_ptr-1]);
+}
diff --git a/drivers/infiniband/core/smi.h b/drivers/infiniband/core/smi.h
index 9a4b349efc3..1cfc2984434 100644
--- a/drivers/infiniband/core/smi.h
+++ b/drivers/infiniband/core/smi.h
@@ -48,10 +48,12 @@ enum smi_action {
 enum smi_forward_action {
 	IB_SMI_LOCAL,	/* SMP should be completed up the stack */
 	IB_SMI_SEND,	/* received DR SMP should be forwarded to the send queue */
+	IB_SMI_FORWARD	/* SMP should be forwarded (for switches only) */
 };
 
 enum smi_action smi_handle_dr_smp_recv(struct ib_smp *smp, u8 node_type,
 				       int port_num, int phys_port_cnt);
+int smi_get_fwd_port(struct ib_smp *smp);
 extern enum smi_forward_action smi_check_forward_dr_smp(struct ib_smp *smp);
 extern enum smi_action smi_handle_dr_smp_send(struct ib_smp *smp,
 					      u8 node_type, int port_num);
diff --git a/drivers/infiniband/core/sysfs.c b/drivers/infiniband/core/sysfs.c
index bf9b9929204..70b77ae6742 100644
--- a/drivers/infiniband/core/sysfs.c
+++ b/drivers/infiniband/core/sysfs.c
@@ -311,7 +311,7 @@ static ssize_t show_pma_counter(struct ib_port *p, struct port_attribute *attr,
 		return sprintf(buf, "N/A (no PMA)\n");
 
 	in_mad  = kzalloc(sizeof *in_mad, GFP_KERNEL);
-	out_mad = kmalloc(sizeof *in_mad, GFP_KERNEL);
+	out_mad = kmalloc(sizeof *out_mad, GFP_KERNEL);
 	if (!in_mad || !out_mad) {
 		ret = -ENOMEM;
 		goto out;
diff --git a/drivers/infiniband/core/ucm.c b/drivers/infiniband/core/ucm.c
index 2586a3ee8eb..424983f5b1e 100644
--- a/drivers/infiniband/core/ucm.c
+++ b/drivers/infiniband/core/ucm.c
@@ -823,7 +823,6 @@ static ssize_t ib_ucm_send_rep(struct ib_ucm_file *file,
 	param.private_data_len    = cmd.len;
 	param.responder_resources = cmd.responder_resources;
 	param.initiator_depth     = cmd.initiator_depth;
-	param.target_ack_delay    = cmd.target_ack_delay;
 	param.failover_accepted   = cmd.failover_accepted;
 	param.flow_control        = cmd.flow_control;
 	param.rnr_retry_count     = cmd.rnr_retry_count;
diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c
index d40652a8015..26d0470eef6 100644
--- a/drivers/infiniband/core/umem.c
+++ b/drivers/infiniband/core/umem.c
@@ -121,6 +121,7 @@ struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
 
 	cur_base = addr & PAGE_MASK;
 
+	ret = 0;
 	while (npages) {
 		ret = get_user_pages(current, current->mm, cur_base,
 				     min_t(int, npages,