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: (52 commits)
  IB/mlx4: Use kzalloc() for new QPs so flags are initialized to 0
  mlx4_core: Use MOD_STAT_CFG command to get minimal page size
  RDMA/cma: Simplify locking needed for serialization of callbacks
  RDMA/addr: Keep pointer to netdevice in struct rdma_dev_addr
  RDMA/cxgb3: Fixes for zero STag
  RDMA/core: Add local DMA L_Key support
  IB/mthca: Fix check of max_send_sge for special QPs
  IB/mthca: Use round_jiffies() for catastrophic error polling timer
  IB/mthca: Remove "stop" flag for catastrophic error polling timer
  IPoIB: Double default RX/TX ring sizes
  IPoIB/cm: Reduce connected mode TX object size
  IB/ipath: Use IEEE OUI for vendor_id reported by ibv_query_device()
  IPoIB: Use dev_set_mtu() to change mtu
  IPoIB: Use rtnl lock/unlock when changing device flags
  IPoIB: Get rid of ipoib_mcast_detach() wrapper
  IPoIB: Only set Q_Key once: after joining broadcast group
  IPoIB: Remove priv->mcast_mutex
  IPoIB: Remove unused IPOIB_MCAST_STARTED code
  RDMA/cxgb3: Set rkey field for new memory windows in iwch_alloc_mw()
  RDMA/nes: Get rid of ring_doorbell parameter of nes_post_cqp_request()
  ...

drivers/infiniband/core/addr.c

@@ -4,28 +4,33 @@
  * Copyright (c) 1999-2005, Mellanox Technologies, Inc. All rights reserved.
  * Copyright (c) 2005 Intel Corporation.  All rights reserved.
  *
- * This Software is licensed under one of the following licenses:
+ * 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
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
  *
- * 1) under the terms of the "Common Public License 1.0" a copy of which is
+ *     Redistribution and use in source and binary forms, with or
- *    available from the Open Source Initiative, see
+ *     without modification, are permitted provided that the following
- *    http://www.opensource.org/licenses/cpl.php.
+ *     conditions are met:
  *
- * 2) under the terms of the "The BSD License" a copy of which is
+ *      - Redistributions of source code must retain the above
- *    available from the Open Source Initiative, see
+ *        copyright notice, this list of conditions and the following
- *    http://www.opensource.org/licenses/bsd-license.php.
+ *        disclaimer.
  *
- * 3) under the terms of the "GNU General Public License (GPL) Version 2" a
+ *      - Redistributions in binary form must reproduce the above
- *    copy of which is available from the Open Source Initiative, see
+ *        copyright notice, this list of conditions and the following
- *    http://www.opensource.org/licenses/gpl-license.php.
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
  *
- * Licensee has the right to choose one of the above licenses.
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
- *
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
- * Redistributions of source code must retain the above copyright
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
- * notice and one of the license notices.
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
- *
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
- * Redistributions in binary form must reproduce both the above copyright
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * notice, one of the license notices in the documentation
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
- * and/or other materials provided with the distribution.
+ * SOFTWARE.
  */
 
 #include <linux/mutex.h>
@@ -100,6 +105,7 @@ int rdma_copy_addr(struct rdma_dev_addr *dev_addr, struct net_device *dev,
 	memcpy(dev_addr->broadcast, dev->broadcast, MAX_ADDR_LEN);
 	if (dst_dev_addr)
 		memcpy(dev_addr->dst_dev_addr, dst_dev_addr, MAX_ADDR_LEN);
+	dev_addr->src_dev = dev;
 	return 0;
 }
 EXPORT_SYMBOL(rdma_copy_addr);

drivers/infiniband/core/agent.h

@@ -32,8 +32,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: agent.h 1389 2004-12-27 22:56:47Z roland $
  */
 
 #ifndef __AGENT_H_

drivers/infiniband/core/cache.c

@@ -31,8 +31,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: cache.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/module.h>

drivers/infiniband/core/cm.c

@@ -31,8 +31,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: cm.c 4311 2005-12-05 18:42:01Z sean.hefty $
  */
 
 #include <linux/completion.h>

drivers/infiniband/core/cma.c

@@ -4,29 +4,33 @@
  * Copyright (c) 1999-2005, Mellanox Technologies, Inc. All rights reserved.
  * Copyright (c) 2005-2006 Intel Corporation.  All rights reserved.
  *
- * This Software is licensed under one of the following licenses:
+ * 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
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
  *
- * 1) under the terms of the "Common Public License 1.0" a copy of which is
+ *     Redistribution and use in source and binary forms, with or
- *    available from the Open Source Initiative, see
+ *     without modification, are permitted provided that the following
- *    http://www.opensource.org/licenses/cpl.php.
+ *     conditions are met:
  *
- * 2) under the terms of the "The BSD License" a copy of which is
+ *      - Redistributions of source code must retain the above
- *    available from the Open Source Initiative, see
+ *        copyright notice, this list of conditions and the following
- *    http://www.opensource.org/licenses/bsd-license.php.
+ *        disclaimer.
  *
- * 3) under the terms of the "GNU General Public License (GPL) Version 2" a
+ *      - Redistributions in binary form must reproduce the above
- *    copy of which is available from the Open Source Initiative, see
+ *        copyright notice, this list of conditions and the following
- *    http://www.opensource.org/licenses/gpl-license.php.
+ *        disclaimer in the documentation and/or other materials
- *
+ *        provided with the distribution.
- * Licensee has the right to choose one of the above licenses.
- *
- * Redistributions of source code must retain the above copyright
- * notice and one of the license notices.
- *
- * Redistributions in binary form must reproduce both the above copyright
- * notice, one of the license notices in the documentation
- * and/or other materials provided with the distribution.
  *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
  */
 
 #include <linux/completion.h>
@@ -126,8 +130,7 @@ struct rdma_id_private {
 
 	struct completion	comp;
 	atomic_t		refcount;
-	wait_queue_head_t	wait_remove;
+	struct mutex		handler_mutex;
-	atomic_t		dev_remove;
 
 	int			backlog;
 	int			timeout_ms;
@@ -351,26 +354,15 @@ static void cma_deref_id(struct rdma_id_private *id_priv)
 		complete(&id_priv->comp);
 }
 
-static int cma_disable_remove(struct rdma_id_private *id_priv,
+static int cma_disable_callback(struct rdma_id_private *id_priv,
 			      enum cma_state state)
 {
-	unsigned long flags;
+	mutex_lock(&id_priv->handler_mutex);
-	int ret;
+	if (id_priv->state != state) {
-
+		mutex_unlock(&id_priv->handler_mutex);
-	spin_lock_irqsave(&id_priv->lock, flags);
+		return -EINVAL;
-	if (id_priv->state == state) {
+	}
-		atomic_inc(&id_priv->dev_remove);
+	return 0;
-		ret = 0;
-	} else
-		ret = -EINVAL;
-	spin_unlock_irqrestore(&id_priv->lock, flags);
-	return ret;
-}
-
-static void cma_enable_remove(struct rdma_id_private *id_priv)
-{
-	if (atomic_dec_and_test(&id_priv->dev_remove))
-		wake_up(&id_priv->wait_remove);
 }
 
 static int cma_has_cm_dev(struct rdma_id_private *id_priv)
@@ -395,8 +387,7 @@ struct rdma_cm_id *rdma_create_id(rdma_cm_event_handler event_handler,
 	mutex_init(&id_priv->qp_mutex);
 	init_completion(&id_priv->comp);
 	atomic_set(&id_priv->refcount, 1);
-	init_waitqueue_head(&id_priv->wait_remove);
+	mutex_init(&id_priv->handler_mutex);
-	atomic_set(&id_priv->dev_remove, 0);
 	INIT_LIST_HEAD(&id_priv->listen_list);
 	INIT_LIST_HEAD(&id_priv->mc_list);
 	get_random_bytes(&id_priv->seq_num, sizeof id_priv->seq_num);
@@ -923,7 +914,7 @@ static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 	struct rdma_cm_event event;
 	int ret = 0;
 
-	if (cma_disable_remove(id_priv, CMA_CONNECT))
+	if (cma_disable_callback(id_priv, CMA_CONNECT))
 		return 0;
 
 	memset(&event, 0, sizeof event);
@@ -970,7 +961,7 @@ static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 		event.param.conn.private_data_len = IB_CM_REJ_PRIVATE_DATA_SIZE;
 		break;
 	default:
-		printk(KERN_ERR "RDMA CMA: unexpected IB CM event: %d",
+		printk(KERN_ERR "RDMA CMA: unexpected IB CM event: %d\n",
 		       ib_event->event);
 		goto out;
 	}
@@ -980,12 +971,12 @@ static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 		/* Destroy the CM ID by returning a non-zero value. */
 		id_priv->cm_id.ib = NULL;
 		cma_exch(id_priv, CMA_DESTROYING);
-		cma_enable_remove(id_priv);
+		mutex_unlock(&id_priv->handler_mutex);
 		rdma_destroy_id(&id_priv->id);
 		return ret;
 	}
 out:
-	cma_enable_remove(id_priv);
+	mutex_unlock(&id_priv->handler_mutex);
 	return ret;
 }
 
@@ -998,6 +989,7 @@ static struct rdma_id_private *cma_new_conn_id(struct rdma_cm_id *listen_id,
 	union cma_ip_addr *src, *dst;
 	__be16 port;
 	u8 ip_ver;
+	int ret;
 
 	if (cma_get_net_info(ib_event->private_data, listen_id->ps,
 			     &ip_ver, &port, &src, &dst))
@@ -1022,10 +1014,11 @@ static struct rdma_id_private *cma_new_conn_id(struct rdma_cm_id *listen_id,
 	if (rt->num_paths == 2)
 		rt->path_rec[1] = *ib_event->param.req_rcvd.alternate_path;
 
-	ib_addr_set_sgid(&rt->addr.dev_addr, &rt->path_rec[0].sgid);
 	ib_addr_set_dgid(&rt->addr.dev_addr, &rt->path_rec[0].dgid);
-	ib_addr_set_pkey(&rt->addr.dev_addr, be16_to_cpu(rt->path_rec[0].pkey));
+	ret = rdma_translate_ip(&id->route.addr.src_addr,
-	rt->addr.dev_addr.dev_type = RDMA_NODE_IB_CA;
+				&id->route.addr.dev_addr);
+	if (ret)
+		goto destroy_id;
 
 	id_priv = container_of(id, struct rdma_id_private, id);
 	id_priv->state = CMA_CONNECT;
@@ -1095,7 +1088,7 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 	int offset, ret;
 
 	listen_id = cm_id->context;
-	if (cma_disable_remove(listen_id, CMA_LISTEN))
+	if (cma_disable_callback(listen_id, CMA_LISTEN))
 		return -ECONNABORTED;
 
 	memset(&event, 0, sizeof event);
@@ -1116,7 +1109,7 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 		goto out;
 	}
 
-	atomic_inc(&conn_id->dev_remove);
+	mutex_lock_nested(&conn_id->handler_mutex, SINGLE_DEPTH_NESTING);
 	mutex_lock(&lock);
 	ret = cma_acquire_dev(conn_id);
 	mutex_unlock(&lock);
@@ -1138,7 +1131,7 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 		    !cma_is_ud_ps(conn_id->id.ps))
 			ib_send_cm_mra(cm_id, CMA_CM_MRA_SETTING, NULL, 0);
 		mutex_unlock(&lock);
-		cma_enable_remove(conn_id);
+		mutex_unlock(&conn_id->handler_mutex);
 		goto out;
 	}
 
@@ -1147,11 +1140,11 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 
 release_conn_id:
 	cma_exch(conn_id, CMA_DESTROYING);
-	cma_enable_remove(conn_id);
+	mutex_unlock(&conn_id->handler_mutex);
 	rdma_destroy_id(&conn_id->id);
 
 out:
-	cma_enable_remove(listen_id);
+	mutex_unlock(&listen_id->handler_mutex);
 	return ret;
 }
 
@@ -1217,7 +1210,7 @@ static int cma_iw_handler(struct iw_cm_id *iw_id, struct iw_cm_event *iw_event)
 	struct sockaddr_in *sin;
 	int ret = 0;
 
-	if (cma_disable_remove(id_priv, CMA_CONNECT))
+	if (cma_disable_callback(id_priv, CMA_CONNECT))
 		return 0;
 
 	memset(&event, 0, sizeof event);
@@ -1261,12 +1254,12 @@ static int cma_iw_handler(struct iw_cm_id *iw_id, struct iw_cm_event *iw_event)
 		/* Destroy the CM ID by returning a non-zero value. */
 		id_priv->cm_id.iw = NULL;
 		cma_exch(id_priv, CMA_DESTROYING);
-		cma_enable_remove(id_priv);
+		mutex_unlock(&id_priv->handler_mutex);
 		rdma_destroy_id(&id_priv->id);
 		return ret;
 	}
 
-	cma_enable_remove(id_priv);
+	mutex_unlock(&id_priv->handler_mutex);
 	return ret;
 }
 
@@ -1282,7 +1275,7 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id,
 	struct ib_device_attr attr;
 
 	listen_id = cm_id->context;
-	if (cma_disable_remove(listen_id, CMA_LISTEN))
+	if (cma_disable_callback(listen_id, CMA_LISTEN))
 		return -ECONNABORTED;
 
 	/* Create a new RDMA id for the new IW CM ID */
@@ -1294,19 +1287,19 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id,
 		goto out;
 	}
 	conn_id = container_of(new_cm_id, struct rdma_id_private, id);
-	atomic_inc(&conn_id->dev_remove);
+	mutex_lock_nested(&conn_id->handler_mutex, SINGLE_DEPTH_NESTING);
 	conn_id->state = CMA_CONNECT;
 
 	dev = ip_dev_find(&init_net, iw_event->local_addr.sin_addr.s_addr);
 	if (!dev) {
 		ret = -EADDRNOTAVAIL;
-		cma_enable_remove(conn_id);
+		mutex_unlock(&conn_id->handler_mutex);
 		rdma_destroy_id(new_cm_id);
 		goto out;
 	}
 	ret = rdma_copy_addr(&conn_id->id.route.addr.dev_addr, dev, NULL);
 	if (ret) {
-		cma_enable_remove(conn_id);
+		mutex_unlock(&conn_id->handler_mutex);
 		rdma_destroy_id(new_cm_id);
 		goto out;
 	}
@@ -1315,7 +1308,7 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id,
 	ret = cma_acquire_dev(conn_id);
 	mutex_unlock(&lock);
 	if (ret) {
-		cma_enable_remove(conn_id);
+		mutex_unlock(&conn_id->handler_mutex);
 		rdma_destroy_id(new_cm_id);
 		goto out;
 	}
@@ -1331,7 +1324,7 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id,
 
 	ret = ib_query_device(conn_id->id.device, &attr);
 	if (ret) {
-		cma_enable_remove(conn_id);
+		mutex_unlock(&conn_id->handler_mutex);
 		rdma_destroy_id(new_cm_id);
 		goto out;
 	}
@@ -1347,14 +1340,17 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id,
 		/* User wants to destroy the CM ID */
 		conn_id->cm_id.iw = NULL;
 		cma_exch(conn_id, CMA_DESTROYING);
-		cma_enable_remove(conn_id);
+		mutex_unlock(&conn_id->handler_mutex);
 		rdma_destroy_id(&conn_id->id);
+		goto out;
 	}
 
+	mutex_unlock(&conn_id->handler_mutex);
+
 out:
 	if (dev)
 		dev_put(dev);
-	cma_enable_remove(listen_id);
+	mutex_unlock(&listen_id->handler_mutex);
 	return ret;
 }
 
@@ -1446,7 +1442,7 @@ static void cma_listen_on_dev(struct rdma_id_private *id_priv,
 	ret = rdma_listen(id, id_priv->backlog);
 	if (ret)
 		printk(KERN_WARNING "RDMA CMA: cma_listen_on_dev, error %d, "
-		       "listening on device %s", ret, cma_dev->device->name);
+		       "listening on device %s\n", ret, cma_dev->device->name);
 }
 
 static void cma_listen_on_all(struct rdma_id_private *id_priv)
@@ -1586,7 +1582,7 @@ static void cma_work_handler(struct work_struct *_work)
 	struct rdma_id_private *id_priv = work->id;
 	int destroy = 0;
 
-	atomic_inc(&id_priv->dev_remove);
+	mutex_lock(&id_priv->handler_mutex);
 	if (!cma_comp_exch(id_priv, work->old_state, work->new_state))
 		goto out;
 
@@ -1595,7 +1591,7 @@ static void cma_work_handler(struct work_struct *_work)
 		destroy = 1;
 	}
 out:
-	cma_enable_remove(id_priv);
+	mutex_unlock(&id_priv->handler_mutex);
 	cma_deref_id(id_priv);
 	if (destroy)
 		rdma_destroy_id(&id_priv->id);
@@ -1758,7 +1754,7 @@ static void addr_handler(int status, struct sockaddr *src_addr,
 	struct rdma_cm_event event;
 
 	memset(&event, 0, sizeof event);
-	atomic_inc(&id_priv->dev_remove);
+	mutex_lock(&id_priv->handler_mutex);
 
 	/*
 	 * Grab mutex to block rdma_destroy_id() from removing the device while
@@ -1787,13 +1783,13 @@ static void addr_handler(int status, struct sockaddr *src_addr,
 
 	if (id_priv->id.event_handler(&id_priv->id, &event)) {
 		cma_exch(id_priv, CMA_DESTROYING);
-		cma_enable_remove(id_priv);
+		mutex_unlock(&id_priv->handler_mutex);
 		cma_deref_id(id_priv);
 		rdma_destroy_id(&id_priv->id);
 		return;
 	}
 out:
-	cma_enable_remove(id_priv);
+	mutex_unlock(&id_priv->handler_mutex);
 	cma_deref_id(id_priv);
 }
 
@@ -2120,7 +2116,7 @@ static int cma_sidr_rep_handler(struct ib_cm_id *cm_id,
 	struct ib_cm_sidr_rep_event_param *rep = &ib_event->param.sidr_rep_rcvd;
 	int ret = 0;
 
-	if (cma_disable_remove(id_priv, CMA_CONNECT))
+	if (cma_disable_callback(id_priv, CMA_CONNECT))
 		return 0;
 
 	memset(&event, 0, sizeof event);
@@ -2151,7 +2147,7 @@ static int cma_sidr_rep_handler(struct ib_cm_id *cm_id,
 		event.status = 0;
 		break;
 	default:
-		printk(KERN_ERR "RDMA CMA: unexpected IB CM event: %d",
+		printk(KERN_ERR "RDMA CMA: unexpected IB CM event: %d\n",
 		       ib_event->event);
 		goto out;
 	}
@@ -2161,12 +2157,12 @@ static int cma_sidr_rep_handler(struct ib_cm_id *cm_id,
 		/* Destroy the CM ID by returning a non-zero value. */
 		id_priv->cm_id.ib = NULL;
 		cma_exch(id_priv, CMA_DESTROYING);
-		cma_enable_remove(id_priv);
+		mutex_unlock(&id_priv->handler_mutex);
 		rdma_destroy_id(&id_priv->id);
 		return ret;
 	}
 out:
-	cma_enable_remove(id_priv);
+	mutex_unlock(&id_priv->handler_mutex);
 	return ret;
 }
 
@@ -2564,8 +2560,8 @@ static int cma_ib_mc_handler(int status, struct ib_sa_multicast *multicast)
 	int ret;
 
 	id_priv = mc->id_priv;
-	if (cma_disable_remove(id_priv, CMA_ADDR_BOUND) &&
+	if (cma_disable_callback(id_priv, CMA_ADDR_BOUND) &&
-	    cma_disable_remove(id_priv, CMA_ADDR_RESOLVED))
+	    cma_disable_callback(id_priv, CMA_ADDR_RESOLVED))
 		return 0;
 
 	mutex_lock(&id_priv->qp_mutex);
@@ -2590,12 +2586,12 @@ static int cma_ib_mc_handler(int status, struct ib_sa_multicast *multicast)
 	ret = id_priv->id.event_handler(&id_priv->id, &event);
 	if (ret) {
 		cma_exch(id_priv, CMA_DESTROYING);
-		cma_enable_remove(id_priv);
+		mutex_unlock(&id_priv->handler_mutex);
 		rdma_destroy_id(&id_priv->id);
 		return 0;
 	}
 
-	cma_enable_remove(id_priv);
+	mutex_unlock(&id_priv->handler_mutex);
 	return 0;
 }
 
@@ -2754,6 +2750,7 @@ static int cma_remove_id_dev(struct rdma_id_private *id_priv)
 {
 	struct rdma_cm_event event;
 	enum cma_state state;
+	int ret = 0;
 
 	/* Record that we want to remove the device */
 	state = cma_exch(id_priv, CMA_DEVICE_REMOVAL);
@@ -2761,15 +2758,18 @@ static int cma_remove_id_dev(struct rdma_id_private *id_priv)
 		return 0;
 
 	cma_cancel_operation(id_priv, state);
-	wait_event(id_priv->wait_remove, !atomic_read(&id_priv->dev_remove));
+	mutex_lock(&id_priv->handler_mutex);
 
 	/* Check for destruction from another callback. */
 	if (!cma_comp(id_priv, CMA_DEVICE_REMOVAL))
-		return 0;
+		goto out;
 
 	memset(&event, 0, sizeof event);
 	event.event = RDMA_CM_EVENT_DEVICE_REMOVAL;
-	return id_priv->id.event_handler(&id_priv->id, &event);
+	ret = id_priv->id.event_handler(&id_priv->id, &event);
+out:
+	mutex_unlock(&id_priv->handler_mutex);
+	return ret;
 }
 
 static void cma_process_remove(struct cma_device *cma_dev)

drivers/infiniband/core/core_priv.h

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: core_priv.h 1349 2004-12-16 21:09:43Z roland $
  */
 
 #ifndef _CORE_PRIV_H

drivers/infiniband/core/device.c

@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: device.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/module.h>

drivers/infiniband/core/fmr_pool.c

@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: fmr_pool.c 2730 2005-06-28 16:43:03Z sean.hefty $
  */
 
 #include <linux/errno.h>

drivers/infiniband/core/mad_priv.h

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mad_priv.h 5596 2006-03-03 01:00:07Z sean.hefty $
  */
 
 #ifndef __IB_MAD_PRIV_H__

drivers/infiniband/core/mad_rmpp.c

@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mad_rmpp.c 1921 2005-03-02 22:58:44Z sean.hefty $
  */
 
 #include "mad_priv.h"

drivers/infiniband/core/mad_rmpp.h

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mad_rmpp.h 1921 2005-02-25 22:58:44Z sean.hefty $
  */
 
 #ifndef __MAD_RMPP_H__

drivers/infiniband/core/packer.c

@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: packer.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/string.h>

drivers/infiniband/core/sa_query.c

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: sa_query.c 2811 2005-07-06 18:11:43Z halr $
  */
 
 #include <linux/module.h>
@@ -361,7 +359,7 @@ static void update_sm_ah(struct work_struct *work)
 {
 	struct ib_sa_port *port =
 		container_of(work, struct ib_sa_port, update_task);
-	struct ib_sa_sm_ah *new_ah, *old_ah;
+	struct ib_sa_sm_ah *new_ah;
 	struct ib_port_attr port_attr;
 	struct ib_ah_attr   ah_attr;
 
@@ -397,12 +395,9 @@ static void update_sm_ah(struct work_struct *work)
 	}
 
 	spin_lock_irq(&port->ah_lock);
-	old_ah = port->sm_ah;
 	port->sm_ah = new_ah;
 	spin_unlock_irq(&port->ah_lock);
 
-	if (old_ah)
-		kref_put(&old_ah->ref, free_sm_ah);
 }
 
 static void ib_sa_event(struct ib_event_handler *handler, struct ib_event *event)
@@ -413,8 +408,17 @@ static void ib_sa_event(struct ib_event_handler *handler, struct ib_event *event
 	    event->event == IB_EVENT_PKEY_CHANGE ||
 	    event->event == IB_EVENT_SM_CHANGE   ||
 	    event->event == IB_EVENT_CLIENT_REREGISTER) {
-		struct ib_sa_device *sa_dev;
+		unsigned long flags;
-		sa_dev = container_of(handler, typeof(*sa_dev), event_handler);
+		struct ib_sa_device *sa_dev =
+			container_of(handler, typeof(*sa_dev), event_handler);
+		struct ib_sa_port *port =
+			&sa_dev->port[event->element.port_num - sa_dev->start_port];
+
+		spin_lock_irqsave(&port->ah_lock, flags);
+		if (port->sm_ah)
+			kref_put(&port->sm_ah->ref, free_sm_ah);
+		port->sm_ah = NULL;
+		spin_unlock_irqrestore(&port->ah_lock, flags);
 
 		schedule_work(&sa_dev->port[event->element.port_num -
 					    sa_dev->start_port].update_task);
@@ -519,6 +523,10 @@ static int alloc_mad(struct ib_sa_query *query, gfp_t gfp_mask)
 	unsigned long flags;
 
 	spin_lock_irqsave(&query->port->ah_lock, flags);
+	if (!query->port->sm_ah) {
+		spin_unlock_irqrestore(&query->port->ah_lock, flags);
+		return -EAGAIN;
+	}
 	kref_get(&query->port->sm_ah->ref);
 	query->sm_ah = query->port->sm_ah;
 	spin_unlock_irqrestore(&query->port->ah_lock, flags);

drivers/infiniband/core/sysfs.c

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: sysfs.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include "core_priv.h"
@@ -665,6 +663,120 @@ static struct class ib_class = {
 	.dev_uevent = ib_device_uevent,
 };
 
+/* Show a given an attribute in the statistics group */
+static ssize_t show_protocol_stat(const struct device *device,
+			    struct device_attribute *attr, char *buf,
+			    unsigned offset)
+{
+	struct ib_device *dev = container_of(device, struct ib_device, dev);
+	union rdma_protocol_stats stats;
+	ssize_t ret;
+
+	ret = dev->get_protocol_stats(dev, &stats);
+	if (ret)
+		return ret;
+
+	return sprintf(buf, "%llu\n",
+		       (unsigned long long) ((u64 *) &stats)[offset]);
+}
+
+/* generate a read-only iwarp statistics attribute */
+#define IW_STATS_ENTRY(name)						\
+static ssize_t show_##name(struct device *device,			\
+			   struct device_attribute *attr, char *buf)	\
+{									\
+	return show_protocol_stat(device, attr, buf,			\
+				  offsetof(struct iw_protocol_stats, name) / \
+				  sizeof (u64));			\
+}									\
+static DEVICE_ATTR(name, S_IRUGO, show_##name, NULL)
+
+IW_STATS_ENTRY(ipInReceives);
+IW_STATS_ENTRY(ipInHdrErrors);
+IW_STATS_ENTRY(ipInTooBigErrors);
+IW_STATS_ENTRY(ipInNoRoutes);
+IW_STATS_ENTRY(ipInAddrErrors);
+IW_STATS_ENTRY(ipInUnknownProtos);
+IW_STATS_ENTRY(ipInTruncatedPkts);
+IW_STATS_ENTRY(ipInDiscards);
+IW_STATS_ENTRY(ipInDelivers);
+IW_STATS_ENTRY(ipOutForwDatagrams);
+IW_STATS_ENTRY(ipOutRequests);
+IW_STATS_ENTRY(ipOutDiscards);
+IW_STATS_ENTRY(ipOutNoRoutes);
+IW_STATS_ENTRY(ipReasmTimeout);
+IW_STATS_ENTRY(ipReasmReqds);
+IW_STATS_ENTRY(ipReasmOKs);
+IW_STATS_ENTRY(ipReasmFails);
+IW_STATS_ENTRY(ipFragOKs);
+IW_STATS_ENTRY(ipFragFails);
+IW_STATS_ENTRY(ipFragCreates);
+IW_STATS_ENTRY(ipInMcastPkts);
+IW_STATS_ENTRY(ipOutMcastPkts);
+IW_STATS_ENTRY(ipInBcastPkts);
+IW_STATS_ENTRY(ipOutBcastPkts);
+IW_STATS_ENTRY(tcpRtoAlgorithm);
+IW_STATS_ENTRY(tcpRtoMin);
+IW_STATS_ENTRY(tcpRtoMax);
+IW_STATS_ENTRY(tcpMaxConn);
+IW_STATS_ENTRY(tcpActiveOpens);
+IW_STATS_ENTRY(tcpPassiveOpens);
+IW_STATS_ENTRY(tcpAttemptFails);
+IW_STATS_ENTRY(tcpEstabResets);
+IW_STATS_ENTRY(tcpCurrEstab);
+IW_STATS_ENTRY(tcpInSegs);
+IW_STATS_ENTRY(tcpOutSegs);
+IW_STATS_ENTRY(tcpRetransSegs);
+IW_STATS_ENTRY(tcpInErrs);
+IW_STATS_ENTRY(tcpOutRsts);
+
+static struct attribute *iw_proto_stats_attrs[] = {
+	&dev_attr_ipInReceives.attr,
+	&dev_attr_ipInHdrErrors.attr,
+	&dev_attr_ipInTooBigErrors.attr,
+	&dev_attr_ipInNoRoutes.attr,
+	&dev_attr_ipInAddrErrors.attr,
+	&dev_attr_ipInUnknownProtos.attr,
+	&dev_attr_ipInTruncatedPkts.attr,
+	&dev_attr_ipInDiscards.attr,
+	&dev_attr_ipInDelivers.attr,
+	&dev_attr_ipOutForwDatagrams.attr,
+	&dev_attr_ipOutRequests.attr,
+	&dev_attr_ipOutDiscards.attr,
+	&dev_attr_ipOutNoRoutes.attr,
+	&dev_attr_ipReasmTimeout.attr,
+	&dev_attr_ipReasmReqds.attr,
+	&dev_attr_ipReasmOKs.attr,
+	&dev_attr_ipReasmFails.attr,
+	&dev_attr_ipFragOKs.attr,
+	&dev_attr_ipFragFails.attr,
+	&dev_attr_ipFragCreates.attr,
+	&dev_attr_ipInMcastPkts.attr,
+	&dev_attr_ipOutMcastPkts.attr,
+	&dev_attr_ipInBcastPkts.attr,
+	&dev_attr_ipOutBcastPkts.attr,
+	&dev_attr_tcpRtoAlgorithm.attr,
+	&dev_attr_tcpRtoMin.attr,
+	&dev_attr_tcpRtoMax.attr,
+	&dev_attr_tcpMaxConn.attr,
+	&dev_attr_tcpActiveOpens.attr,
+	&dev_attr_tcpPassiveOpens.attr,
+	&dev_attr_tcpAttemptFails.attr,
+	&dev_attr_tcpEstabResets.attr,
+	&dev_attr_tcpCurrEstab.attr,
+	&dev_attr_tcpInSegs.attr,
+	&dev_attr_tcpOutSegs.attr,
+	&dev_attr_tcpRetransSegs.attr,
+	&dev_attr_tcpInErrs.attr,
+	&dev_attr_tcpOutRsts.attr,
+	NULL
+};
+
+static struct attribute_group iw_stats_group = {
+	.name	= "proto_stats",
+	.attrs	= iw_proto_stats_attrs,
+};
+
 int ib_device_register_sysfs(struct ib_device *device)
 {
 	struct device *class_dev = &device->dev;
@@ -707,6 +819,12 @@ int ib_device_register_sysfs(struct ib_device *device)
 		}
 	}
 
+	if (device->node_type == RDMA_NODE_RNIC && device->get_protocol_stats) {
+		ret = sysfs_create_group(&class_dev->kobj, &iw_stats_group);
+		if (ret)
+			goto err_put;
+	}
+
 	return 0;
 
 err_put:

drivers/infiniband/core/ucm.c

@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: ucm.c 4311 2005-12-05 18:42:01Z sean.hefty $
  */
 
 #include <linux/completion.h>

drivers/infiniband/core/ud_header.c

@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: ud_header.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/errno.h>

drivers/infiniband/core/umem.c

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: uverbs_mem.c 2743 2005-06-28 22:27:59Z roland $
  */
 
 #include <linux/mm.h>

drivers/infiniband/core/user_mad.c

@@ -31,8 +31,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: user_mad.c 5596 2006-03-03 01:00:07Z sean.hefty $
  */
 
 #include <linux/module.h>

drivers/infiniband/core/uverbs.h

@@ -32,8 +32,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: uverbs.h 2559 2005-06-06 19:43:16Z roland $
  */
 
 #ifndef UVERBS_H

drivers/infiniband/core/uverbs_cmd.c

@@ -31,8 +31,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: uverbs_cmd.c 2708 2005-06-24 17:27:21Z roland $
  */
 
 #include <linux/file.h>
@@ -919,7 +917,7 @@ ssize_t ib_uverbs_poll_cq(struct ib_uverbs_file *file,
 		resp->wc[i].opcode 	   = wc[i].opcode;
 		resp->wc[i].vendor_err 	   = wc[i].vendor_err;
 		resp->wc[i].byte_len 	   = wc[i].byte_len;
-		resp->wc[i].imm_data 	   = (__u32 __force) wc[i].imm_data;
+		resp->wc[i].ex.imm_data    = (__u32 __force) wc[i].ex.imm_data;
 		resp->wc[i].qp_num 	   = wc[i].qp->qp_num;
 		resp->wc[i].src_qp 	   = wc[i].src_qp;
 		resp->wc[i].wc_flags 	   = wc[i].wc_flags;

drivers/infiniband/core/uverbs_main.c

@@ -32,8 +32,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: uverbs_main.c 2733 2005-06-28 19:14:34Z roland $
  */
 
 #include <linux/module.h>

drivers/infiniband/core/verbs.c

@@ -34,8 +34,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: verbs.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/errno.h>
@@ -317,7 +315,6 @@ static const struct {
 } qp_state_table[IB_QPS_ERR + 1][IB_QPS_ERR + 1] = {
 	[IB_QPS_RESET] = {
 		[IB_QPS_RESET] = { .valid = 1 },
-		[IB_QPS_ERR]   = { .valid = 1 },
 		[IB_QPS_INIT]  = {
 			.valid = 1,
 			.req_param = {
@@ -755,6 +752,52 @@ int ib_dereg_mr(struct ib_mr *mr)
 }
 EXPORT_SYMBOL(ib_dereg_mr);
 
+struct ib_mr *ib_alloc_fast_reg_mr(struct ib_pd *pd, int max_page_list_len)
+{
+	struct ib_mr *mr;
+
+	if (!pd->device->alloc_fast_reg_mr)
+		return ERR_PTR(-ENOSYS);
+
+	mr = pd->device->alloc_fast_reg_mr(pd, max_page_list_len);
+
+	if (!IS_ERR(mr)) {
+		mr->device  = pd->device;
+		mr->pd      = pd;
+		mr->uobject = NULL;
+		atomic_inc(&pd->usecnt);
+		atomic_set(&mr->usecnt, 0);
+	}
+
+	return mr;
+}
+EXPORT_SYMBOL(ib_alloc_fast_reg_mr);
+
+struct ib_fast_reg_page_list *ib_alloc_fast_reg_page_list(struct ib_device *device,
+							  int max_page_list_len)
+{
+	struct ib_fast_reg_page_list *page_list;
+
+	if (!device->alloc_fast_reg_page_list)
+		return ERR_PTR(-ENOSYS);
+
+	page_list = device->alloc_fast_reg_page_list(device, max_page_list_len);
+
+	if (!IS_ERR(page_list)) {
+		page_list->device = device;
+		page_list->max_page_list_len = max_page_list_len;
+	}
+
+	return page_list;
+}
+EXPORT_SYMBOL(ib_alloc_fast_reg_page_list);
+
+void ib_free_fast_reg_page_list(struct ib_fast_reg_page_list *page_list)
+{
+	page_list->device->free_fast_reg_page_list(page_list);
+}
+EXPORT_SYMBOL(ib_free_fast_reg_page_list);
+
 /* Memory windows */
 
 struct ib_mw *ib_alloc_mw(struct ib_pd *pd)

drivers/infiniband/hw/amso1100/c2_rnic.c

@@ -454,7 +454,7 @@ int __devinit c2_rnic_init(struct c2_dev *c2dev)
 	    (IB_DEVICE_RESIZE_MAX_WR |
 	     IB_DEVICE_CURR_QP_STATE_MOD |
 	     IB_DEVICE_SYS_IMAGE_GUID |
-	     IB_DEVICE_ZERO_STAG |
+	     IB_DEVICE_LOCAL_DMA_LKEY |
 	     IB_DEVICE_MEM_WINDOW);
 
 	/* Allocate the qptr_array */

drivers/infiniband/hw/cxgb3/cxio_hal.c

@@ -145,7 +145,9 @@ static int cxio_hal_clear_qp_ctx(struct cxio_rdev *rdev_p, u32 qpid)
 	}
 	wqe = (struct t3_modify_qp_wr *) skb_put(skb, sizeof(*wqe));
 	memset(wqe, 0, sizeof(*wqe));
-	build_fw_riwrh((struct fw_riwrh *) wqe, T3_WR_QP_MOD, 3, 0, qpid, 7);
+	build_fw_riwrh((struct fw_riwrh *) wqe, T3_WR_QP_MOD,
+		       T3_COMPLETION_FLAG | T3_NOTIFY_FLAG, 0, qpid, 7,
+		       T3_SOPEOP);
 	wqe->flags = cpu_to_be32(MODQP_WRITE_EC);
 	sge_cmd = qpid << 8 | 3;
 	wqe->sge_cmd = cpu_to_be64(sge_cmd);
@@ -276,7 +278,7 @@ int cxio_create_qp(struct cxio_rdev *rdev_p, u32 kernel_domain,
 	if (!wq->qpid)
 		return -ENOMEM;
 
-	wq->rq = kzalloc(depth * sizeof(u64), GFP_KERNEL);
+	wq->rq = kzalloc(depth * sizeof(struct t3_swrq), GFP_KERNEL);
 	if (!wq->rq)
 		goto err1;
 
@@ -300,6 +302,7 @@ int cxio_create_qp(struct cxio_rdev *rdev_p, u32 kernel_domain,
 	if (!kernel_domain)
 		wq->udb = (u64)rdev_p->rnic_info.udbell_physbase +
 					(wq->qpid << rdev_p->qpshift);
+	wq->rdev = rdev_p;
 	PDBG("%s qpid 0x%x doorbell 0x%p udb 0x%llx\n", __func__,
 	     wq->qpid, wq->doorbell, (unsigned long long) wq->udb);
 	return 0;
@@ -558,7 +561,7 @@ static int cxio_hal_init_ctrl_qp(struct cxio_rdev *rdev_p)
 	wqe = (struct t3_modify_qp_wr *) skb_put(skb, sizeof(*wqe));
 	memset(wqe, 0, sizeof(*wqe));
 	build_fw_riwrh((struct fw_riwrh *) wqe, T3_WR_QP_MOD, 0, 0,
-		       T3_CTL_QP_TID, 7);
+		       T3_CTL_QP_TID, 7, T3_SOPEOP);
 	wqe->flags = cpu_to_be32(MODQP_WRITE_EC);
 	sge_cmd = (3ULL << 56) | FW_RI_SGEEC_START << 8 | 3;
 	wqe->sge_cmd = cpu_to_be64(sge_cmd);
@@ -674,7 +677,7 @@ static int cxio_hal_ctrl_qp_write_mem(struct cxio_rdev *rdev_p, u32 addr,
 		build_fw_riwrh((struct fw_riwrh *) wqe, T3_WR_BP, flag,
 			       Q_GENBIT(rdev_p->ctrl_qp.wptr,
 					T3_CTRL_QP_SIZE_LOG2), T3_CTRL_QP_ID,
-			       wr_len);
+			       wr_len, T3_SOPEOP);
 		if (flag == T3_COMPLETION_FLAG)
 			ring_doorbell(rdev_p->ctrl_qp.doorbell, T3_CTRL_QP_ID);
 		len -= 96;
@@ -816,6 +819,13 @@ int cxio_deallocate_window(struct cxio_rdev *rdev_p, u32 stag)
 			     0, 0);
 }
 
+int cxio_allocate_stag(struct cxio_rdev *rdev_p, u32 *stag, u32 pdid, u32 pbl_size, u32 pbl_addr)
+{
+	*stag = T3_STAG_UNSET;
+	return __cxio_tpt_op(rdev_p, 0, stag, 0, pdid, TPT_NON_SHARED_MR,
+			     0, 0, 0ULL, 0, 0, pbl_size, pbl_addr);
+}
+
 int cxio_rdma_init(struct cxio_rdev *rdev_p, struct t3_rdma_init_attr *attr)
 {
 	struct t3_rdma_init_wr *wqe;
@@ -1257,13 +1267,16 @@ proc_cqe:
 		wq->sq_rptr = CQE_WRID_SQ_WPTR(*hw_cqe);
 		PDBG("%s completing sq idx %ld\n", __func__,
 		     Q_PTR2IDX(wq->sq_rptr, wq->sq_size_log2));
-		*cookie = (wq->sq +
+		*cookie = wq->sq[Q_PTR2IDX(wq->sq_rptr, wq->sq_size_log2)].wr_id;
-			   Q_PTR2IDX(wq->sq_rptr, wq->sq_size_log2))->wr_id;
 		wq->sq_rptr++;
 	} else {
 		PDBG("%s completing rq idx %ld\n", __func__,
 		     Q_PTR2IDX(wq->rq_rptr, wq->rq_size_log2));
-		*cookie = *(wq->rq + Q_PTR2IDX(wq->rq_rptr, wq->rq_size_log2));
+		*cookie = wq->rq[Q_PTR2IDX(wq->rq_rptr, wq->rq_size_log2)].wr_id;
+		if (wq->rq[Q_PTR2IDX(wq->rq_rptr, wq->rq_size_log2)].pbl_addr)
+			cxio_hal_pblpool_free(wq->rdev,
+				wq->rq[Q_PTR2IDX(wq->rq_rptr,
+				wq->rq_size_log2)].pbl_addr, T3_STAG0_PBL_SIZE);
 		wq->rq_rptr++;
 	}
 

drivers/infiniband/hw/cxgb3/cxio_hal.h

@@ -45,15 +45,17 @@
 #define T3_CTRL_QP_SIZE_LOG2  8
 #define T3_CTRL_CQ_ID    0
 
-/* TBD */
 #define T3_MAX_NUM_RI (1<<15)
 #define T3_MAX_NUM_QP (1<<15)
 #define T3_MAX_NUM_CQ (1<<15)
 #define T3_MAX_NUM_PD (1<<15)
 #define T3_MAX_PBL_SIZE 256
 #define T3_MAX_RQ_SIZE 1024
+#define T3_MAX_QP_DEPTH (T3_MAX_RQ_SIZE-1)
+#define T3_MAX_CQ_DEPTH 8192
 #define T3_MAX_NUM_STAG (1<<15)
 #define T3_MAX_MR_SIZE 0x100000000ULL
+#define T3_PAGESIZE_MASK 0xffff000  /* 4KB-128MB */
 
 #define T3_STAG_UNSET 0xffffffff
 
@@ -165,6 +167,7 @@ int cxio_reregister_phys_mem(struct cxio_rdev *rdev, u32 * stag, u32 pdid,
 int cxio_dereg_mem(struct cxio_rdev *rdev, u32 stag, u32 pbl_size,
 		   u32 pbl_addr);
 int cxio_allocate_window(struct cxio_rdev *rdev, u32 * stag, u32 pdid);
+int cxio_allocate_stag(struct cxio_rdev *rdev, u32 *stag, u32 pdid, u32 pbl_size, u32 pbl_addr);
 int cxio_deallocate_window(struct cxio_rdev *rdev, u32 stag);
 int cxio_rdma_init(struct cxio_rdev *rdev, struct t3_rdma_init_attr *attr);
 void cxio_register_ev_cb(cxio_hal_ev_callback_func_t ev_cb);

drivers/infiniband/hw/cxgb3/cxio_wr.h

@@ -39,6 +39,9 @@
 
 #define T3_MAX_SGE      4
 #define T3_MAX_INLINE	64
+#define T3_STAG0_PBL_SIZE (2 * T3_MAX_SGE << 3)
+#define T3_STAG0_MAX_PBE_LEN (128 * 1024 * 1024)
+#define T3_STAG0_PAGE_SHIFT 15
 
 #define Q_EMPTY(rptr,wptr) ((rptr)==(wptr))
 #define Q_FULL(rptr,wptr,size_log2)  ( (((wptr)-(rptr))>>(size_log2)) && \
@@ -72,7 +75,8 @@ enum t3_wr_opcode {
 	T3_WR_BIND = FW_WROPCODE_RI_BIND_MW,
 	T3_WR_RCV = FW_WROPCODE_RI_RECEIVE,
 	T3_WR_INIT = FW_WROPCODE_RI_RDMA_INIT,
-	T3_WR_QP_MOD = FW_WROPCODE_RI_MODIFY_QP
+	T3_WR_QP_MOD = FW_WROPCODE_RI_MODIFY_QP,
+	T3_WR_FASTREG = FW_WROPCODE_RI_FASTREGISTER_MR
 } __attribute__ ((packed));
 
 enum t3_rdma_opcode {
@@ -89,7 +93,8 @@ enum t3_rdma_opcode {
 	T3_FAST_REGISTER,
 	T3_LOCAL_INV,
 	T3_QP_MOD,
-	T3_BYPASS
+	T3_BYPASS,
+	T3_RDMA_READ_REQ_WITH_INV,
 } __attribute__ ((packed));
 
 static inline enum t3_rdma_opcode wr2opcode(enum t3_wr_opcode wrop)
@@ -103,6 +108,7 @@ static inline enum t3_rdma_opcode wr2opcode(enum t3_wr_opcode wrop)
 		case T3_WR_BIND: return T3_BIND_MW;
 		case T3_WR_INIT: return T3_RDMA_INIT;
 		case T3_WR_QP_MOD: return T3_QP_MOD;
+		case T3_WR_FASTREG: return T3_FAST_REGISTER;
 		default: break;
 	}
 	return -1;
@@ -170,11 +176,54 @@ struct t3_send_wr {
 	struct t3_sge sgl[T3_MAX_SGE];	/* 4+ */
 };
 
+#define T3_MAX_FASTREG_DEPTH 24
+#define T3_MAX_FASTREG_FRAG 10
+
+struct t3_fastreg_wr {
+	struct fw_riwrh wrh;	/* 0 */
+	union t3_wrid wrid;	/* 1 */
+	__be32 stag;		/* 2 */
+	__be32 len;
+	__be32 va_base_hi;	/* 3 */
+	__be32 va_base_lo_fbo;
+	__be32 page_type_perms; /* 4 */
+	__be32 reserved1;
+	__be64 pbl_addrs[0];	/* 5+ */
+};
+
+/*
+ * If a fastreg wr spans multiple wqes, then the 2nd fragment look like this.
+ */
+struct t3_pbl_frag {
+	struct fw_riwrh wrh;	/* 0 */
+	__be64 pbl_addrs[14];	/* 1..14 */
+};
+
+#define S_FR_PAGE_COUNT		24
+#define M_FR_PAGE_COUNT		0xff
+#define V_FR_PAGE_COUNT(x)	((x) << S_FR_PAGE_COUNT)
+#define G_FR_PAGE_COUNT(x)	((((x) >> S_FR_PAGE_COUNT)) & M_FR_PAGE_COUNT)
+
+#define S_FR_PAGE_SIZE		16
+#define M_FR_PAGE_SIZE		0x1f
+#define V_FR_PAGE_SIZE(x)	((x) << S_FR_PAGE_SIZE)
+#define G_FR_PAGE_SIZE(x)	((((x) >> S_FR_PAGE_SIZE)) & M_FR_PAGE_SIZE)
+
+#define S_FR_TYPE		8
+#define M_FR_TYPE		0x1
+#define V_FR_TYPE(x)		((x) << S_FR_TYPE)
+#define G_FR_TYPE(x)		((((x) >> S_FR_TYPE)) & M_FR_TYPE)
+
+#define S_FR_PERMS		0
+#define M_FR_PERMS		0xff
+#define V_FR_PERMS(x)		((x) << S_FR_PERMS)
+#define G_FR_PERMS(x)		((((x) >> S_FR_PERMS)) & M_FR_PERMS)
+
 struct t3_local_inv_wr {
 	struct fw_riwrh wrh;	/* 0 */
 	union t3_wrid wrid;	/* 1 */
 	__be32 stag;		/* 2 */
-	__be32 reserved3;
+	__be32 reserved;
 };
 
 struct t3_rdma_write_wr {
@@ -193,7 +242,8 @@ struct t3_rdma_read_wr {
 	struct fw_riwrh wrh;	/* 0 */
 	union t3_wrid wrid;	/* 1 */
 	u8 rdmaop;		/* 2 */
-	u8 reserved[3];
+	u8 local_inv;
+	u8 reserved[2];
 	__be32 rem_stag;
 	__be64 rem_to;		/* 3 */
 	__be32 local_stag;	/* 4 */
@@ -201,18 +251,6 @@ struct t3_rdma_read_wr {
 	__be64 local_to;	/* 5 */
 };
 
-enum t3_addr_type {
-	T3_VA_BASED_TO = 0x0,
-	T3_ZERO_BASED_TO = 0x1
-} __attribute__ ((packed));
-
-enum t3_mem_perms {
-	T3_MEM_ACCESS_LOCAL_READ = 0x1,
-	T3_MEM_ACCESS_LOCAL_WRITE = 0x2,
-	T3_MEM_ACCESS_REM_READ = 0x4,
-	T3_MEM_ACCESS_REM_WRITE = 0x8
-} __attribute__ ((packed));
-
 struct t3_bind_mw_wr {
 	struct fw_riwrh wrh;	/* 0 */
 	union t3_wrid wrid;	/* 1 */
@@ -336,6 +374,11 @@ struct t3_genbit {
 	__be64 genbit;
 };
 
+struct t3_wq_in_err {
+	u64 flit[13];
+	u64 err;
+};
+
 enum rdma_init_wr_flags {
 	MPA_INITIATOR = (1<<0),
 	PRIV_QP = (1<<1),
@@ -346,13 +389,16 @@ union t3_wr {
 	struct t3_rdma_write_wr write;
 	struct t3_rdma_read_wr read;
 	struct t3_receive_wr recv;
+	struct t3_fastreg_wr fastreg;
+	struct t3_pbl_frag pbl_frag;
 	struct t3_local_inv_wr local_inv;
 	struct t3_bind_mw_wr bind;
 	struct t3_bypass_wr bypass;
 	struct t3_rdma_init_wr init;
 	struct t3_modify_qp_wr qp_mod;
 	struct t3_genbit genbit;
-	u64 flit[16];
+	struct t3_wq_in_err wq_in_err;
+	__be64 flit[16];
 };
 
 #define T3_SQ_CQE_FLIT	  13
@@ -366,12 +412,18 @@ static inline enum t3_wr_opcode fw_riwrh_opcode(struct fw_riwrh *wqe)
 	return G_FW_RIWR_OP(be32_to_cpu(wqe->op_seop_flags));
 }
 
+enum t3_wr_hdr_bits {
+	T3_EOP = 1,
+	T3_SOP = 2,
+	T3_SOPEOP = T3_EOP|T3_SOP,
+};
+
 static inline void build_fw_riwrh(struct fw_riwrh *wqe, enum t3_wr_opcode op,
 				  enum t3_wr_flags flags, u8 genbit, u32 tid,
-				  u8 len)
+				  u8 len, u8 sopeop)
 {
 	wqe->op_seop_flags = cpu_to_be32(V_FW_RIWR_OP(op) |
-					 V_FW_RIWR_SOPEOP(M_FW_RIWR_SOPEOP) |
+					 V_FW_RIWR_SOPEOP(sopeop) |
 					 V_FW_RIWR_FLAGS(flags));
 	wmb();
 	wqe->gen_tid_len = cpu_to_be32(V_FW_RIWR_GEN(genbit) |
@@ -404,6 +456,7 @@ enum tpt_addr_type {
 };
 
 enum tpt_mem_perm {
+	TPT_MW_BIND = 0x10,
 	TPT_LOCAL_READ = 0x8,
 	TPT_LOCAL_WRITE = 0x4,
 	TPT_REMOTE_READ = 0x2,
@@ -615,6 +668,11 @@ struct t3_swsq {
 	int			signaled;
 };
 
+struct t3_swrq {
+	__u64			wr_id;
+	__u32			pbl_addr;
+};
+
 /*
  * A T3 WQ implements both the SQ and RQ.
  */
@@ -631,14 +689,15 @@ struct t3_wq {
 	u32 sq_wptr;			/* sq_wptr - sq_rptr == count of */
 	u32 sq_rptr;			/* pending wrs */
 	u32 sq_size_log2;		/* sq size */
-	u64 *rq;			/* SW RQ (holds consumer wr_ids */
+	struct t3_swrq *rq;		/* SW RQ (holds consumer wr_ids */
 	u32 rq_wptr;			/* rq_wptr - rq_rptr == count of */
 	u32 rq_rptr;			/* pending wrs */
-	u64 *rq_oldest_wr;		/* oldest wr on the SW RQ */
+	struct t3_swrq *rq_oldest_wr;	/* oldest wr on the SW RQ */
 	u32 rq_size_log2;		/* rq size */
 	u32 rq_addr;			/* rq adapter address */
 	void __iomem *doorbell;		/* kernel db */
 	u64 udb;			/* user db if any */
+	struct cxio_rdev *rdev;
 };
 
 struct t3_cq {
@@ -659,7 +718,7 @@ struct t3_cq {
 
 static inline void cxio_set_wq_in_error(struct t3_wq *wq)
 {
-	wq->queue->flit[13] = 1;
+	wq->queue->wq_in_err.err = 1;
 }
 
 static inline struct t3_cqe *cxio_next_hw_cqe(struct t3_cq *cq)

drivers/infiniband/hw/cxgb3/iwch.c

@@ -71,18 +71,16 @@ static void rnic_init(struct iwch_dev *rnicp)
 	idr_init(&rnicp->mmidr);
 	spin_lock_init(&rnicp->lock);
 
-	rnicp->attr.vendor_id = 0x168;
-	rnicp->attr.vendor_part_id = 7;
 	rnicp->attr.max_qps = T3_MAX_NUM_QP - 32;
-	rnicp->attr.max_wrs = (1UL << 24) - 1;
+	rnicp->attr.max_wrs = T3_MAX_QP_DEPTH;
 	rnicp->attr.max_sge_per_wr = T3_MAX_SGE;
 	rnicp->attr.max_sge_per_rdma_write_wr = T3_MAX_SGE;
 	rnicp->attr.max_cqs = T3_MAX_NUM_CQ - 1;
-	rnicp->attr.max_cqes_per_cq = (1UL << 24) - 1;
+	rnicp->attr.max_cqes_per_cq = T3_MAX_CQ_DEPTH;
 	rnicp->attr.max_mem_regs = cxio_num_stags(&rnicp->rdev);
 	rnicp->attr.max_phys_buf_entries = T3_MAX_PBL_SIZE;
 	rnicp->attr.max_pds = T3_MAX_NUM_PD - 1;
-	rnicp->attr.mem_pgsizes_bitmask = 0x7FFF;	/* 4KB-128MB */
+	rnicp->attr.mem_pgsizes_bitmask = T3_PAGESIZE_MASK;
 	rnicp->attr.max_mr_size = T3_MAX_MR_SIZE;
 	rnicp->attr.can_resize_wq = 0;
 	rnicp->attr.max_rdma_reads_per_qp = 8;

drivers/infiniband/hw/cxgb3/iwch.h

@@ -48,8 +48,6 @@ struct iwch_qp;
 struct iwch_mr;
 
 struct iwch_rnic_attributes {
-	u32 vendor_id;
-	u32 vendor_part_id;
 	u32 max_qps;
 	u32 max_wrs;				/* Max for any SQ/RQ */
 	u32 max_sge_per_wr;

drivers/infiniband/hw/cxgb3/iwch_cq.c

@@ -81,6 +81,7 @@ static int iwch_poll_cq_one(struct iwch_dev *rhp, struct iwch_cq *chp,
 	wc->wr_id = cookie;
 	wc->qp = &qhp->ibqp;
 	wc->vendor_err = CQE_STATUS(cqe);
+	wc->wc_flags = 0;
 
 	PDBG("%s qpid 0x%x type %d opcode %d status 0x%x wrid hi 0x%x "
 	     "lo 0x%x cookie 0x%llx\n", __func__,
@@ -94,6 +95,11 @@ static int iwch_poll_cq_one(struct iwch_dev *rhp, struct iwch_cq *chp,
 		else
 			wc->byte_len = 0;
 		wc->opcode = IB_WC_RECV;
+		if (CQE_OPCODE(cqe) == T3_SEND_WITH_INV ||
+		    CQE_OPCODE(cqe) == T3_SEND_WITH_SE_INV) {
+			wc->ex.invalidate_rkey = CQE_WRID_STAG(cqe);
+			wc->wc_flags |= IB_WC_WITH_INVALIDATE;
+		}
 	} else {
 		switch (CQE_OPCODE(cqe)) {
 		case T3_RDMA_WRITE:
@@ -105,17 +111,20 @@ static int iwch_poll_cq_one(struct iwch_dev *rhp, struct iwch_cq *chp,
 			break;
 		case T3_SEND:
 		case T3_SEND_WITH_SE:
+		case T3_SEND_WITH_INV:
+		case T3_SEND_WITH_SE_INV:
 			wc->opcode = IB_WC_SEND;
 			break;
 		case T3_BIND_MW:
 			wc->opcode = IB_WC_BIND_MW;
 			break;
 
-		/* these aren't supported yet */
-		case T3_SEND_WITH_INV:
-		case T3_SEND_WITH_SE_INV:
 		case T3_LOCAL_INV:
+			wc->opcode = IB_WC_LOCAL_INV;
+			break;
 		case T3_FAST_REGISTER:
+			wc->opcode = IB_WC_FAST_REG_MR;
+			break;
 		default:
 			printk(KERN_ERR MOD "Unexpected opcode %d "
 			       "in the CQE received for QPID=0x%0x\n",

drivers/infiniband/hw/cxgb3/iwch_provider.c

@@ -56,6 +56,7 @@
 #include "iwch_provider.h"
 #include "iwch_cm.h"
 #include "iwch_user.h"
+#include "common.h"
 
 static int iwch_modify_port(struct ib_device *ibdev,
 			    u8 port, int port_modify_mask,
@@ -747,6 +748,7 @@ static struct ib_mw *iwch_alloc_mw(struct ib_pd *pd)
 	mhp->attr.type = TPT_MW;
 	mhp->attr.stag = stag;
 	mmid = (stag) >> 8;
+	mhp->ibmw.rkey = stag;
 	insert_handle(rhp, &rhp->mmidr, mhp, mmid);
 	PDBG("%s mmid 0x%x mhp %p stag 0x%x\n", __func__, mmid, mhp, stag);
 	return &(mhp->ibmw);
@@ -768,6 +770,68 @@ static int iwch_dealloc_mw(struct ib_mw *mw)
 	return 0;
 }
 
+static struct ib_mr *iwch_alloc_fast_reg_mr(struct ib_pd *pd, int pbl_depth)
+{
+	struct iwch_dev *rhp;
+	struct iwch_pd *php;
+	struct iwch_mr *mhp;
+	u32 mmid;
+	u32 stag = 0;
+	int ret;
+
+	php = to_iwch_pd(pd);
+	rhp = php->rhp;
+	mhp = kzalloc(sizeof(*mhp), GFP_KERNEL);
+	if (!mhp)
+		return ERR_PTR(-ENOMEM);
+
+	mhp->rhp = rhp;
+	ret = iwch_alloc_pbl(mhp, pbl_depth);
+	if (ret) {
+		kfree(mhp);
+		return ERR_PTR(ret);
+	}
+	mhp->attr.pbl_size = pbl_depth;
+	ret = cxio_allocate_stag(&rhp->rdev, &stag, php->pdid,
+				 mhp->attr.pbl_size, mhp->attr.pbl_addr);
+	if (ret) {
+		iwch_free_pbl(mhp);
+		kfree(mhp);
+		return ERR_PTR(ret);
+	}
+	mhp->attr.pdid = php->pdid;
+	mhp->attr.type = TPT_NON_SHARED_MR;
+	mhp->attr.stag = stag;
+	mhp->attr.state = 1;
+	mmid = (stag) >> 8;
+	mhp->ibmr.rkey = mhp->ibmr.lkey = stag;
+	insert_handle(rhp, &rhp->mmidr, mhp, mmid);
+	PDBG("%s mmid 0x%x mhp %p stag 0x%x\n", __func__, mmid, mhp, stag);
+	return &(mhp->ibmr);
+}
+
+static struct ib_fast_reg_page_list *iwch_alloc_fastreg_pbl(
+					struct ib_device *device,
+					int page_list_len)
+{
+	struct ib_fast_reg_page_list *page_list;
+
+	page_list = kmalloc(sizeof *page_list + page_list_len * sizeof(u64),
+			    GFP_KERNEL);
+	if (!page_list)
+		return ERR_PTR(-ENOMEM);
+
+	page_list->page_list = (u64 *)(page_list + 1);
+	page_list->max_page_list_len = page_list_len;
+
+	return page_list;
+}
+
+static void iwch_free_fastreg_pbl(struct ib_fast_reg_page_list *page_list)
+{
+	kfree(page_list);
+}
+
 static int iwch_destroy_qp(struct ib_qp *ib_qp)
 {
 	struct iwch_dev *rhp;
@@ -843,6 +907,15 @@ static struct ib_qp *iwch_create_qp(struct ib_pd *pd,
 	 */
 	sqsize = roundup_pow_of_two(attrs->cap.max_send_wr);
 	wqsize = roundup_pow_of_two(rqsize + sqsize);
+
+	/*
+	 * Kernel users need more wq space for fastreg WRs which can take
+	 * 2 WR fragments.
+	 */
+	ucontext = pd->uobject ? to_iwch_ucontext(pd->uobject->context) : NULL;
+	if (!ucontext && wqsize < (rqsize + (2 * sqsize)))
+		wqsize = roundup_pow_of_two(rqsize +
+				roundup_pow_of_two(attrs->cap.max_send_wr * 2));
 	PDBG("%s wqsize %d sqsize %d rqsize %d\n", __func__,
 	     wqsize, sqsize, rqsize);
 	qhp = kzalloc(sizeof(*qhp), GFP_KERNEL);
@@ -851,7 +924,6 @@ static struct ib_qp *iwch_create_qp(struct ib_pd *pd,
 	qhp->wq.size_log2 = ilog2(wqsize);
 	qhp->wq.rq_size_log2 = ilog2(rqsize);
 	qhp->wq.sq_size_log2 = ilog2(sqsize);
-	ucontext = pd->uobject ? to_iwch_ucontext(pd->uobject->context) : NULL;
 	if (cxio_create_qp(&rhp->rdev, !udata, &qhp->wq,
 			   ucontext ? &ucontext->uctx : &rhp->rdev.uctx)) {
 		kfree(qhp);
@@ -935,10 +1007,10 @@ static struct ib_qp *iwch_create_qp(struct ib_pd *pd,
 	qhp->ibqp.qp_num = qhp->wq.qpid;
 	init_timer(&(qhp->timer));
 	PDBG("%s sq_num_entries %d, rq_num_entries %d "
-	     "qpid 0x%0x qhp %p dma_addr 0x%llx size %d\n",
+	     "qpid 0x%0x qhp %p dma_addr 0x%llx size %d rq_addr 0x%x\n",
 	     __func__, qhp->attr.sq_num_entries, qhp->attr.rq_num_entries,
 	     qhp->wq.qpid, qhp, (unsigned long long) qhp->wq.dma_addr,
-	     1 << qhp->wq.size_log2);
+	     1 << qhp->wq.size_log2, qhp->wq.rq_addr);
 	return &qhp->ibqp;
 }
 
@@ -1023,6 +1095,29 @@ static int iwch_query_gid(struct ib_device *ibdev, u8 port,
 	return 0;
 }
 
+static u64 fw_vers_string_to_u64(struct iwch_dev *iwch_dev)
+{
+	struct ethtool_drvinfo info;
+	struct net_device *lldev = iwch_dev->rdev.t3cdev_p->lldev;
+	char *cp, *next;
+	unsigned fw_maj, fw_min, fw_mic;
+
+	rtnl_lock();
+	lldev->ethtool_ops->get_drvinfo(lldev, &info);
+	rtnl_unlock();
+
+	next = info.fw_version + 1;
+	cp = strsep(&next, ".");
+	sscanf(cp, "%i", &fw_maj);
+	cp = strsep(&next, ".");
+	sscanf(cp, "%i", &fw_min);
+	cp = strsep(&next, ".");
+	sscanf(cp, "%i", &fw_mic);
+
+	return (((u64)fw_maj & 0xffff) << 32) | ((fw_min & 0xffff) << 16) |
+	       (fw_mic & 0xffff);
+}
+
 static int iwch_query_device(struct ib_device *ibdev,
 			     struct ib_device_attr *props)
 {
@@ -1033,7 +1128,10 @@ static int iwch_query_device(struct ib_device *ibdev,
 	dev = to_iwch_dev(ibdev);
 	memset(props, 0, sizeof *props);
 	memcpy(&props->sys_image_guid, dev->rdev.t3cdev_p->lldev->dev_addr, 6);
+	props->hw_ver = dev->rdev.t3cdev_p->type;
+	props->fw_ver = fw_vers_string_to_u64(dev);
 	props->device_cap_flags = dev->device_cap_flags;
+	props->page_size_cap = dev->attr.mem_pgsizes_bitmask;
 	props->vendor_id = (u32)dev->rdev.rnic_info.pdev->vendor;
 	props->vendor_part_id = (u32)dev->rdev.rnic_info.pdev->device;
 	props->max_mr_size = dev->attr.max_mr_size;
@@ -1048,6 +1146,7 @@ static int iwch_query_device(struct ib_device *ibdev,
 	props->max_mr = dev->attr.max_mem_regs;
 	props->max_pd = dev->attr.max_pds;
 	props->local_ca_ack_delay = 0;
+	props->max_fast_reg_page_list_len = T3_MAX_FASTREG_DEPTH;
 
 	return 0;
 }
@@ -1088,6 +1187,28 @@ static ssize_t show_rev(struct device *dev, struct device_attribute *attr,
 	return sprintf(buf, "%d\n", iwch_dev->rdev.t3cdev_p->type);
 }
 
+static int fw_supports_fastreg(struct iwch_dev *iwch_dev)
+{
+	struct ethtool_drvinfo info;
+	struct net_device *lldev = iwch_dev->rdev.t3cdev_p->lldev;
+	char *cp, *next;
+	unsigned fw_maj, fw_min;
+
+	rtnl_lock();
+	lldev->ethtool_ops->get_drvinfo(lldev, &info);
+	rtnl_unlock();
+
+	next = info.fw_version+1;
+	cp = strsep(&next, ".");
+	sscanf(cp, "%i", &fw_maj);
+	cp = strsep(&next, ".");
+	sscanf(cp, "%i", &fw_min);
+
+	PDBG("%s maj %u min %u\n", __func__, fw_maj, fw_min);
+
+	return fw_maj > 6 || (fw_maj == 6 && fw_min > 0);
+}
+
 static ssize_t show_fw_ver(struct device *dev, struct device_attribute *attr, char *buf)
 {
 	struct iwch_dev *iwch_dev = container_of(dev, struct iwch_dev,
@@ -1127,6 +1248,61 @@ static ssize_t show_board(struct device *dev, struct device_attribute *attr,
 		       iwch_dev->rdev.rnic_info.pdev->device);
 }
 
+static int iwch_get_mib(struct ib_device *ibdev,
+			union rdma_protocol_stats *stats)
+{
+	struct iwch_dev *dev;
+	struct tp_mib_stats m;
+	int ret;
+
+	PDBG("%s ibdev %p\n", __func__, ibdev);
+	dev = to_iwch_dev(ibdev);
+	ret = dev->rdev.t3cdev_p->ctl(dev->rdev.t3cdev_p, RDMA_GET_MIB, &m);
+	if (ret)
+		return -ENOSYS;
+
+	memset(stats, 0, sizeof *stats);
+	stats->iw.ipInReceives = ((u64) m.ipInReceive_hi << 32) +
+				m.ipInReceive_lo;
+	stats->iw.ipInHdrErrors = ((u64) m.ipInHdrErrors_hi << 32) +
+				  m.ipInHdrErrors_lo;
+	stats->iw.ipInAddrErrors = ((u64) m.ipInAddrErrors_hi << 32) +
+				   m.ipInAddrErrors_lo;
+	stats->iw.ipInUnknownProtos = ((u64) m.ipInUnknownProtos_hi << 32) +
+				      m.ipInUnknownProtos_lo;
+	stats->iw.ipInDiscards = ((u64) m.ipInDiscards_hi << 32) +
+				 m.ipInDiscards_lo;
+	stats->iw.ipInDelivers = ((u64) m.ipInDelivers_hi << 32) +
+				 m.ipInDelivers_lo;
+	stats->iw.ipOutRequests = ((u64) m.ipOutRequests_hi << 32) +
+				  m.ipOutRequests_lo;
+	stats->iw.ipOutDiscards = ((u64) m.ipOutDiscards_hi << 32) +
+				  m.ipOutDiscards_lo;
+	stats->iw.ipOutNoRoutes = ((u64) m.ipOutNoRoutes_hi << 32) +
+				  m.ipOutNoRoutes_lo;
+	stats->iw.ipReasmTimeout = (u64) m.ipReasmTimeout;
+	stats->iw.ipReasmReqds = (u64) m.ipReasmReqds;
+	stats->iw.ipReasmOKs = (u64) m.ipReasmOKs;
+	stats->iw.ipReasmFails = (u64) m.ipReasmFails;
+	stats->iw.tcpActiveOpens = (u64) m.tcpActiveOpens;
+	stats->iw.tcpPassiveOpens = (u64) m.tcpPassiveOpens;
+	stats->iw.tcpAttemptFails = (u64) m.tcpAttemptFails;
+	stats->iw.tcpEstabResets = (u64) m.tcpEstabResets;
+	stats->iw.tcpOutRsts = (u64) m.tcpOutRsts;
+	stats->iw.tcpCurrEstab = (u64) m.tcpCurrEstab;
+	stats->iw.tcpInSegs = ((u64) m.tcpInSegs_hi << 32) +
+			      m.tcpInSegs_lo;
+	stats->iw.tcpOutSegs = ((u64) m.tcpOutSegs_hi << 32) +
+			       m.tcpOutSegs_lo;
+	stats->iw.tcpRetransSegs = ((u64) m.tcpRetransSeg_hi << 32) +
+				  m.tcpRetransSeg_lo;
+	stats->iw.tcpInErrs = ((u64) m.tcpInErrs_hi << 32) +
+			      m.tcpInErrs_lo;
+	stats->iw.tcpRtoMin = (u64) m.tcpRtoMin;
+	stats->iw.tcpRtoMax = (u64) m.tcpRtoMax;
+	return 0;
+}
+
 static DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL);
 static DEVICE_ATTR(fw_ver, S_IRUGO, show_fw_ver, NULL);
 static DEVICE_ATTR(hca_type, S_IRUGO, show_hca, NULL);
@@ -1136,7 +1312,7 @@ static struct device_attribute *iwch_class_attributes[] = {
 	&dev_attr_hw_rev,
 	&dev_attr_fw_ver,
 	&dev_attr_hca_type,
-	&dev_attr_board_id
+	&dev_attr_board_id,
 };
 
 int iwch_register_device(struct iwch_dev *dev)
@@ -1149,8 +1325,12 @@ int iwch_register_device(struct iwch_dev *dev)
 	memset(&dev->ibdev.node_guid, 0, sizeof(dev->ibdev.node_guid));
 	memcpy(&dev->ibdev.node_guid, dev->rdev.t3cdev_p->lldev->dev_addr, 6);
 	dev->ibdev.owner = THIS_MODULE;
-	dev->device_cap_flags =
+	dev->device_cap_flags = IB_DEVICE_LOCAL_DMA_LKEY | IB_DEVICE_MEM_WINDOW;
-	    (IB_DEVICE_ZERO_STAG | IB_DEVICE_MEM_WINDOW);
+
+	/* cxgb3 supports STag 0. */
+	dev->ibdev.local_dma_lkey = 0;
+	if (fw_supports_fastreg(dev))
+		dev->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS;
 
 	dev->ibdev.uverbs_cmd_mask =
 	    (1ull << IB_USER_VERBS_CMD_GET_CONTEXT) |
@@ -1202,15 +1382,16 @@ int iwch_register_device(struct iwch_dev *dev)
 	dev->ibdev.alloc_mw = iwch_alloc_mw;
 	dev->ibdev.bind_mw = iwch_bind_mw;
 	dev->ibdev.dealloc_mw = iwch_dealloc_mw;
-
+	dev->ibdev.alloc_fast_reg_mr = iwch_alloc_fast_reg_mr;
+	dev->ibdev.alloc_fast_reg_page_list = iwch_alloc_fastreg_pbl;
+	dev->ibdev.free_fast_reg_page_list = iwch_free_fastreg_pbl;
 	dev->ibdev.attach_mcast = iwch_multicast_attach;
 	dev->ibdev.detach_mcast = iwch_multicast_detach;
 	dev->ibdev.process_mad = iwch_process_mad;
-
 	dev->ibdev.req_notify_cq = iwch_arm_cq;
 	dev->ibdev.post_send = iwch_post_send;
 	dev->ibdev.post_recv = iwch_post_receive;
-
+	dev->ibdev.get_protocol_stats = iwch_get_mib;
 
 	dev->ibdev.iwcm = kmalloc(sizeof(struct iw_cm_verbs), GFP_KERNEL);
 	if (!dev->ibdev.iwcm)

drivers/infiniband/hw/cxgb3/iwch_provider.h

@@ -296,14 +296,6 @@ static inline u32 iwch_ib_to_tpt_access(int acc)
 	       TPT_LOCAL_READ;
 }
 
-static inline u32 iwch_ib_to_mwbind_access(int acc)
-{
-	return (acc & IB_ACCESS_REMOTE_WRITE ? T3_MEM_ACCESS_REM_WRITE : 0) |
-	       (acc & IB_ACCESS_REMOTE_READ ? T3_MEM_ACCESS_REM_READ : 0) |
-	       (acc & IB_ACCESS_LOCAL_WRITE ? T3_MEM_ACCESS_LOCAL_WRITE : 0) |
-	       T3_MEM_ACCESS_LOCAL_READ;
-}
-
 enum iwch_mmid_state {
 	IWCH_STAG_STATE_VALID,
 	IWCH_STAG_STATE_INVALID

drivers/infiniband/hw/cxgb3/iwch_qp.c

@@ -33,10 +33,11 @@
 #include "iwch.h"
 #include "iwch_cm.h"
 #include "cxio_hal.h"
+#include "cxio_resource.h"
 
 #define NO_SUPPORT -1
 
-static int iwch_build_rdma_send(union t3_wr *wqe, struct ib_send_wr *wr,
+static int build_rdma_send(union t3_wr *wqe, struct ib_send_wr *wr,
 				u8 * flit_cnt)
 {
 	int i;
@@ -44,59 +45,44 @@ static int iwch_build_rdma_send(union t3_wr *wqe, struct ib_send_wr *wr,
 
 	switch (wr->opcode) {
 	case IB_WR_SEND:
-	case IB_WR_SEND_WITH_IMM:
 		if (wr->send_flags & IB_SEND_SOLICITED)
 			wqe->send.rdmaop = T3_SEND_WITH_SE;
 		else
 			wqe->send.rdmaop = T3_SEND;
 		wqe->send.rem_stag = 0;
 		break;
-#if 0				/* Not currently supported */
+	case IB_WR_SEND_WITH_INV:
-	case TYPE_SEND_INVALIDATE:
+		if (wr->send_flags & IB_SEND_SOLICITED)
-	case TYPE_SEND_INVALIDATE_IMMEDIATE:
+			wqe->send.rdmaop = T3_SEND_WITH_SE_INV;
-		wqe->send.rdmaop = T3_SEND_WITH_INV;
+		else
-		wqe->send.rem_stag = cpu_to_be32(wr->wr.rdma.rkey);
+			wqe->send.rdmaop = T3_SEND_WITH_INV;
+		wqe->send.rem_stag = cpu_to_be32(wr->ex.invalidate_rkey);
 		break;
-	case TYPE_SEND_SE_INVALIDATE:
-		wqe->send.rdmaop = T3_SEND_WITH_SE_INV;
-		wqe->send.rem_stag = cpu_to_be32(wr->wr.rdma.rkey);
-		break;
-#endif
 	default:
-		break;
+		return -EINVAL;
 	}
 	if (wr->num_sge > T3_MAX_SGE)
 		return -EINVAL;
 	wqe->send.reserved[0] = 0;
 	wqe->send.reserved[1] = 0;
 	wqe->send.reserved[2] = 0;
-	if (wr->opcode == IB_WR_SEND_WITH_IMM) {
+	plen = 0;
-		plen = 4;
+	for (i = 0; i < wr->num_sge; i++) {
-		wqe->send.sgl[0].stag = wr->ex.imm_data;
+		if ((plen + wr->sg_list[i].length) < plen)
-		wqe->send.sgl[0].len = __constant_cpu_to_be32(0);
+			return -EMSGSIZE;
-		wqe->send.num_sgle = __constant_cpu_to_be32(0);
+
-		*flit_cnt = 5;
+		plen += wr->sg_list[i].length;
-	} else {
+		wqe->send.sgl[i].stag = cpu_to_be32(wr->sg_list[i].lkey);
-		plen = 0;
+		wqe->send.sgl[i].len = cpu_to_be32(wr->sg_list[i].length);
-		for (i = 0; i < wr->num_sge; i++) {
+		wqe->send.sgl[i].to = cpu_to_be64(wr->sg_list[i].addr);
-			if ((plen + wr->sg_list[i].length) < plen) {
-				return -EMSGSIZE;
-			}
-			plen += wr->sg_list[i].length;
-			wqe->send.sgl[i].stag =
-			    cpu_to_be32(wr->sg_list[i].lkey);
-			wqe->send.sgl[i].len =
-			    cpu_to_be32(wr->sg_list[i].length);
-			wqe->send.sgl[i].to = cpu_to_be64(wr->sg_list[i].addr);
-		}
-		wqe->send.num_sgle = cpu_to_be32(wr->num_sge);
-		*flit_cnt = 4 + ((wr->num_sge) << 1);
 	}
+	wqe->send.num_sgle = cpu_to_be32(wr->num_sge);
+	*flit_cnt = 4 + ((wr->num_sge) << 1);
 	wqe->send.plen = cpu_to_be32(plen);
 	return 0;
 }
 
-static int iwch_build_rdma_write(union t3_wr *wqe, struct ib_send_wr *wr,
+static int build_rdma_write(union t3_wr *wqe, struct ib_send_wr *wr,
 				 u8 *flit_cnt)
 {
 	int i;
@@ -137,15 +123,18 @@ static int iwch_build_rdma_write(union t3_wr *wqe, struct ib_send_wr *wr,
 	return 0;
 }
 
-static int iwch_build_rdma_read(union t3_wr *wqe, struct ib_send_wr *wr,
+static int build_rdma_read(union t3_wr *wqe, struct ib_send_wr *wr,
 				u8 *flit_cnt)
 {
 	if (wr->num_sge > 1)
 		return -EINVAL;
 	wqe->read.rdmaop = T3_READ_REQ;
+	if (wr->opcode == IB_WR_RDMA_READ_WITH_INV)
+		wqe->read.local_inv = 1;
+	else
+		wqe->read.local_inv = 0;
 	wqe->read.reserved[0] = 0;
 	wqe->read.reserved[1] = 0;
-	wqe->read.reserved[2] = 0;
 	wqe->read.rem_stag = cpu_to_be32(wr->wr.rdma.rkey);
 	wqe->read.rem_to = cpu_to_be64(wr->wr.rdma.remote_addr);
 	wqe->read.local_stag = cpu_to_be32(wr->sg_list[0].lkey);
@@ -155,6 +144,57 @@ static int iwch_build_rdma_read(union t3_wr *wqe, struct ib_send_wr *wr,
 	return 0;
 }
 
+static int build_fastreg(union t3_wr *wqe, struct ib_send_wr *wr,
+				u8 *flit_cnt, int *wr_cnt, struct t3_wq *wq)
+{
+	int i;
+	__be64 *p;
+
+	if (wr->wr.fast_reg.page_list_len > T3_MAX_FASTREG_DEPTH)
+		return -EINVAL;
+	*wr_cnt = 1;
+	wqe->fastreg.stag = cpu_to_be32(wr->wr.fast_reg.rkey);
+	wqe->fastreg.len = cpu_to_be32(wr->wr.fast_reg.length);
+	wqe->fastreg.va_base_hi = cpu_to_be32(wr->wr.fast_reg.iova_start >> 32);
+	wqe->fastreg.va_base_lo_fbo =
+				cpu_to_be32(wr->wr.fast_reg.iova_start & 0xffffffff);
+	wqe->fastreg.page_type_perms = cpu_to_be32(
+		V_FR_PAGE_COUNT(wr->wr.fast_reg.page_list_len) |
+		V_FR_PAGE_SIZE(wr->wr.fast_reg.page_shift-12) |
+		V_FR_TYPE(TPT_VATO) |
+		V_FR_PERMS(iwch_ib_to_tpt_access(wr->wr.fast_reg.access_flags)));
+	p = &wqe->fastreg.pbl_addrs[0];
+	for (i = 0; i < wr->wr.fast_reg.page_list_len; i++, p++) {
+
+		/* If we need a 2nd WR, then set it up */
+		if (i == T3_MAX_FASTREG_FRAG) {
+			*wr_cnt = 2;
+			wqe = (union t3_wr *)(wq->queue +
+				Q_PTR2IDX((wq->wptr+1), wq->size_log2));
+			build_fw_riwrh((void *)wqe, T3_WR_FASTREG, 0,
+			       Q_GENBIT(wq->wptr + 1, wq->size_log2),
+			       0, 1 + wr->wr.fast_reg.page_list_len - T3_MAX_FASTREG_FRAG,
+			       T3_EOP);
+
+			p = &wqe->pbl_frag.pbl_addrs[0];
+		}
+		*p = cpu_to_be64((u64)wr->wr.fast_reg.page_list->page_list[i]);
+	}
+	*flit_cnt = 5 + wr->wr.fast_reg.page_list_len;
+	if (*flit_cnt > 15)
+		*flit_cnt = 15;
+	return 0;
+}
+
+static int build_inv_stag(union t3_wr *wqe, struct ib_send_wr *wr,
+				u8 *flit_cnt)
+{
+	wqe->local_inv.stag = cpu_to_be32(wr->ex.invalidate_rkey);
+	wqe->local_inv.reserved = 0;
+	*flit_cnt = sizeof(struct t3_local_inv_wr) >> 3;
+	return 0;
+}
+
 /*
  * TBD: this is going to be moved to firmware. Missing pdid/qpid check for now.
  */
@@ -205,23 +245,106 @@ static int iwch_sgl2pbl_map(struct iwch_dev *rhp, struct ib_sge *sg_list,
 	return 0;
 }
 
-static int iwch_build_rdma_recv(struct iwch_dev *rhp, union t3_wr *wqe,
+static int build_rdma_recv(struct iwch_qp *qhp, union t3_wr *wqe,
 				struct ib_recv_wr *wr)
 {
-	int i;
+	int i, err = 0;
-	if (wr->num_sge > T3_MAX_SGE)
+	u32 pbl_addr[T3_MAX_SGE];
-		return -EINVAL;
+	u8 page_size[T3_MAX_SGE];
+
+	err = iwch_sgl2pbl_map(qhp->rhp, wr->sg_list, wr->num_sge, pbl_addr,
+			       page_size);
+	if (err)
+		return err;
+	wqe->recv.pagesz[0] = page_size[0];
+	wqe->recv.pagesz[1] = page_size[1];
+	wqe->recv.pagesz[2] = page_size[2];
+	wqe->recv.pagesz[3] = page_size[3];
 	wqe->recv.num_sgle = cpu_to_be32(wr->num_sge);
 	for (i = 0; i < wr->num_sge; i++) {
 		wqe->recv.sgl[i].stag = cpu_to_be32(wr->sg_list[i].lkey);
 		wqe->recv.sgl[i].len = cpu_to_be32(wr->sg_list[i].length);
-		wqe->recv.sgl[i].to = cpu_to_be64(wr->sg_list[i].addr);
+
+		/* to in the WQE == the offset into the page */
+		wqe->recv.sgl[i].to = cpu_to_be64(((u32) wr->sg_list[i].addr) %
+				(1UL << (12 + page_size[i])));
+
+		/* pbl_addr is the adapters address in the PBL */
+		wqe->recv.pbl_addr[i] = cpu_to_be32(pbl_addr[i]);
 	}
 	for (; i < T3_MAX_SGE; i++) {
 		wqe->recv.sgl[i].stag = 0;
 		wqe->recv.sgl[i].len = 0;
 		wqe->recv.sgl[i].to = 0;
+		wqe->recv.pbl_addr[i] = 0;
 	}
+	qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr,
+			     qhp->wq.rq_size_log2)].wr_id = wr->wr_id;
+	qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr,
+			     qhp->wq.rq_size_log2)].pbl_addr = 0;
+	return 0;
+}
+
+static int build_zero_stag_recv(struct iwch_qp *qhp, union t3_wr *wqe,
+				struct ib_recv_wr *wr)
+{
+	int i;
+	u32 pbl_addr;
+	u32 pbl_offset;
+
+
+	/*
+	 * The T3 HW requires the PBL in the HW recv descriptor to reference
+	 * a PBL entry.  So we allocate the max needed PBL memory here and pass
+	 * it to the uP in the recv WR.  The uP will build the PBL and setup
+	 * the HW recv descriptor.
+	 */
+	pbl_addr = cxio_hal_pblpool_alloc(&qhp->rhp->rdev, T3_STAG0_PBL_SIZE);
+	if (!pbl_addr)
+		return -ENOMEM;
+
+	/*
+	 * Compute the 8B aligned offset.
+	 */
+	pbl_offset = (pbl_addr - qhp->rhp->rdev.rnic_info.pbl_base) >> 3;
+
+	wqe->recv.num_sgle = cpu_to_be32(wr->num_sge);
+
+	for (i = 0; i < wr->num_sge; i++) {
+
+		/*
+		 * Use a 128MB page size. This and an imposed 128MB
+		 * sge length limit allows us to require only a 2-entry HW
+		 * PBL for each SGE.  This restriction is acceptable since
+		 * since it is not possible to allocate 128MB of contiguous
+		 * DMA coherent memory!
+		 */
+		if (wr->sg_list[i].length > T3_STAG0_MAX_PBE_LEN)
+			return -EINVAL;
+		wqe->recv.pagesz[i] = T3_STAG0_PAGE_SHIFT;
+
+		/*
+		 * T3 restricts a recv to all zero-stag or all non-zero-stag.
+		 */
+		if (wr->sg_list[i].lkey != 0)
+			return -EINVAL;
+		wqe->recv.sgl[i].stag = 0;
+		wqe->recv.sgl[i].len = cpu_to_be32(wr->sg_list[i].length);
+		wqe->recv.sgl[i].to = cpu_to_be64(wr->sg_list[i].addr);
+		wqe->recv.pbl_addr[i] = cpu_to_be32(pbl_offset);
+		pbl_offset += 2;
+	}
+	for (; i < T3_MAX_SGE; i++) {
+		wqe->recv.pagesz[i] = 0;
+		wqe->recv.sgl[i].stag = 0;
+		wqe->recv.sgl[i].len = 0;
+		wqe->recv.sgl[i].to = 0;
+		wqe->recv.pbl_addr[i] = 0;
+	}
+	qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr,
+			     qhp->wq.rq_size_log2)].wr_id = wr->wr_id;
+	qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr,
+			     qhp->wq.rq_size_log2)].pbl_addr = pbl_addr;
 	return 0;
 }
 
@@ -238,6 +361,7 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 	u32 num_wrs;
 	unsigned long flag;
 	struct t3_swsq *sqp;
+	int wr_cnt = 1;
 
 	qhp = to_iwch_qp(ibqp);
 	spin_lock_irqsave(&qhp->lock, flag);
@@ -262,33 +386,45 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 		t3_wr_flags = 0;
 		if (wr->send_flags & IB_SEND_SOLICITED)
 			t3_wr_flags |= T3_SOLICITED_EVENT_FLAG;
-		if (wr->send_flags & IB_SEND_FENCE)
-			t3_wr_flags |= T3_READ_FENCE_FLAG;
 		if (wr->send_flags & IB_SEND_SIGNALED)
 			t3_wr_flags |= T3_COMPLETION_FLAG;
 		sqp = qhp->wq.sq +
 		      Q_PTR2IDX(qhp->wq.sq_wptr, qhp->wq.sq_size_log2);
 		switch (wr->opcode) {
 		case IB_WR_SEND:
-		case IB_WR_SEND_WITH_IMM:
+		case IB_WR_SEND_WITH_INV:
+			if (wr->send_flags & IB_SEND_FENCE)
+				t3_wr_flags |= T3_READ_FENCE_FLAG;
 			t3_wr_opcode = T3_WR_SEND;
-			err = iwch_build_rdma_send(wqe, wr, &t3_wr_flit_cnt);
+			err = build_rdma_send(wqe, wr, &t3_wr_flit_cnt);
 			break;
 		case IB_WR_RDMA_WRITE:
 		case IB_WR_RDMA_WRITE_WITH_IMM:
 			t3_wr_opcode = T3_WR_WRITE;
-			err = iwch_build_rdma_write(wqe, wr, &t3_wr_flit_cnt);
+			err = build_rdma_write(wqe, wr, &t3_wr_flit_cnt);
 			break;
 		case IB_WR_RDMA_READ:
+		case IB_WR_RDMA_READ_WITH_INV:
 			t3_wr_opcode = T3_WR_READ;
 			t3_wr_flags = 0; /* T3 reads are always signaled */
-			err = iwch_build_rdma_read(wqe, wr, &t3_wr_flit_cnt);
+			err = build_rdma_read(wqe, wr, &t3_wr_flit_cnt);
 			if (err)
 				break;
 			sqp->read_len = wqe->read.local_len;
 			if (!qhp->wq.oldest_read)
 				qhp->wq.oldest_read = sqp;
 			break;
+		case IB_WR_FAST_REG_MR:
+			t3_wr_opcode = T3_WR_FASTREG;
+			err = build_fastreg(wqe, wr, &t3_wr_flit_cnt,
+						 &wr_cnt, &qhp->wq);
+			break;
+		case IB_WR_LOCAL_INV:
+			if (wr->send_flags & IB_SEND_FENCE)
+				t3_wr_flags |= T3_LOCAL_FENCE_FLAG;
+			t3_wr_opcode = T3_WR_INV_STAG;
+			err = build_inv_stag(wqe, wr, &t3_wr_flit_cnt);
+			break;
 		default:
 			PDBG("%s post of type=%d TBD!\n", __func__,
 			     wr->opcode);
@@ -307,14 +443,15 @@ int iwch_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 
 		build_fw_riwrh((void *) wqe, t3_wr_opcode, t3_wr_flags,
 			       Q_GENBIT(qhp->wq.wptr, qhp->wq.size_log2),
-			       0, t3_wr_flit_cnt);
+			       0, t3_wr_flit_cnt,
+			       (wr_cnt == 1) ? T3_SOPEOP : T3_SOP);
 		PDBG("%s cookie 0x%llx wq idx 0x%x swsq idx %ld opcode %d\n",
 		     __func__, (unsigned long long) wr->wr_id, idx,
 		     Q_PTR2IDX(qhp->wq.sq_wptr, qhp->wq.sq_size_log2),
 		     sqp->opcode);
 		wr = wr->next;
 		num_wrs--;
-		++(qhp->wq.wptr);
+		qhp->wq.wptr += wr_cnt;
 		++(qhp->wq.sq_wptr);
 	}
 	spin_unlock_irqrestore(&qhp->lock, flag);
@@ -345,21 +482,27 @@ int iwch_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr,
 		return -EINVAL;
 	}
 	while (wr) {
+		if (wr->num_sge > T3_MAX_SGE) {
+			err = -EINVAL;
+			*bad_wr = wr;
+			break;
+		}
 		idx = Q_PTR2IDX(qhp->wq.wptr, qhp->wq.size_log2);
 		wqe = (union t3_wr *) (qhp->wq.queue + idx);
 		if (num_wrs)
-			err = iwch_build_rdma_recv(qhp->rhp, wqe, wr);
+			if (wr->sg_list[0].lkey)
+				err = build_rdma_recv(qhp, wqe, wr);
+			else
+				err = build_zero_stag_recv(qhp, wqe, wr);
 		else
 			err = -ENOMEM;
 		if (err) {
 			*bad_wr = wr;
 			break;
 		}
-		qhp->wq.rq[Q_PTR2IDX(qhp->wq.rq_wptr, qhp->wq.rq_size_log2)] =
-			wr->wr_id;
 		build_fw_riwrh((void *) wqe, T3_WR_RCV, T3_COMPLETION_FLAG,
 			       Q_GENBIT(qhp->wq.wptr, qhp->wq.size_log2),
-			       0, sizeof(struct t3_receive_wr) >> 3);
+			       0, sizeof(struct t3_receive_wr) >> 3, T3_SOPEOP);
 		PDBG("%s cookie 0x%llx idx 0x%x rq_wptr 0x%x rw_rptr 0x%x "
 		     "wqe %p \n", __func__, (unsigned long long) wr->wr_id,
 		     idx, qhp->wq.rq_wptr, qhp->wq.rq_rptr, wqe);
@@ -419,10 +562,10 @@ int iwch_bind_mw(struct ib_qp *qp,
 	sgl.lkey = mw_bind->mr->lkey;
 	sgl.length = mw_bind->length;
 	wqe->bind.reserved = 0;
-	wqe->bind.type = T3_VA_BASED_TO;
+	wqe->bind.type = TPT_VATO;
 
 	/* TBD: check perms */
-	wqe->bind.perms = iwch_ib_to_mwbind_access(mw_bind->mw_access_flags);
+	wqe->bind.perms = iwch_ib_to_tpt_access(mw_bind->mw_access_flags);
 	wqe->bind.mr_stag = cpu_to_be32(mw_bind->mr->lkey);
 	wqe->bind.mw_stag = cpu_to_be32(mw->rkey);
 	wqe->bind.mw_len = cpu_to_be32(mw_bind->length);
@@ -430,7 +573,7 @@ int iwch_bind_mw(struct ib_qp *qp,
 	err = iwch_sgl2pbl_map(rhp, &sgl, 1, &pbl_addr, &page_size);
 	if (err) {
 		spin_unlock_irqrestore(&qhp->lock, flag);
-	        return err;
+		return err;
 	}
 	wqe->send.wrid.id0.hi = qhp->wq.sq_wptr;
 	sqp = qhp->wq.sq + Q_PTR2IDX(qhp->wq.sq_wptr, qhp->wq.sq_size_log2);
@@ -441,10 +584,9 @@ int iwch_bind_mw(struct ib_qp *qp,
 	sqp->signaled = (mw_bind->send_flags & IB_SEND_SIGNALED);
 	wqe->bind.mr_pbl_addr = cpu_to_be32(pbl_addr);
 	wqe->bind.mr_pagesz = page_size;
-	wqe->flit[T3_SQ_COOKIE_FLIT] = mw_bind->wr_id;
 	build_fw_riwrh((void *)wqe, T3_WR_BIND, t3_wr_flags,
 		       Q_GENBIT(qhp->wq.wptr, qhp->wq.size_log2), 0,
-			        sizeof(struct t3_bind_mw_wr) >> 3);
+		       sizeof(struct t3_bind_mw_wr) >> 3, T3_SOPEOP);
 	++(qhp->wq.wptr);
 	++(qhp->wq.sq_wptr);
 	spin_unlock_irqrestore(&qhp->lock, flag);
@@ -758,7 +900,8 @@ static int rdma_init(struct iwch_dev *rhp, struct iwch_qp *qhp,
 	init_attr.qp_dma_size = (1UL << qhp->wq.size_log2);
 	init_attr.rqe_count = iwch_rqes_posted(qhp);
 	init_attr.flags = qhp->attr.mpa_attr.initiator ? MPA_INITIATOR : 0;
-	init_attr.flags |= capable(CAP_NET_BIND_SERVICE) ? PRIV_QP : 0;
+	if (!qhp->ibqp.uobject)
+		init_attr.flags |= PRIV_QP;
 	if (peer2peer) {
 		init_attr.rtr_type = RTR_READ;
 		if (init_attr.ord == 0 && qhp->attr.mpa_attr.initiator)

drivers/infiniband/hw/ehca/ehca_irq.c

@@ -531,7 +531,7 @@ void ehca_process_eq(struct ehca_shca *shca, int is_irq)
 {
 	struct ehca_eq *eq = &shca->eq;
 	struct ehca_eqe_cache_entry *eqe_cache = eq->eqe_cache;
-	u64 eqe_value;
+	u64 eqe_value, ret;
 	unsigned long flags;
 	int eqe_cnt, i;
 	int eq_empty = 0;
@@ -583,8 +583,13 @@ void ehca_process_eq(struct ehca_shca *shca, int is_irq)
 			ehca_dbg(&shca->ib_device,
 				 "No eqe found for irq event");
 		goto unlock_irq_spinlock;
-	} else if (!is_irq)
+	} else if (!is_irq) {
+		ret = hipz_h_eoi(eq->ist);
+		if (ret != H_SUCCESS)
+			ehca_err(&shca->ib_device,
+				 "bad return code EOI -rc = %ld\n", ret);
 		ehca_dbg(&shca->ib_device, "deadman found %x eqe", eqe_cnt);
+	}
 	if (unlikely(eqe_cnt == EHCA_EQE_CACHE_SIZE))
 		ehca_dbg(&shca->ib_device, "too many eqes for one irq event");
 	/* enable irq for new packets */

drivers/infiniband/hw/ehca/ehca_main.c

@@ -923,6 +923,7 @@ static struct of_device_id ehca_device_table[] =
 	},
 	{},
 };
+MODULE_DEVICE_TABLE(of, ehca_device_table);
 
 static struct of_platform_driver ehca_driver = {
 	.name        = "ehca",

drivers/infiniband/hw/ehca/ehca_reqs.c

@@ -544,8 +544,16 @@ int ehca_post_recv(struct ib_qp *qp,
 		   struct ib_recv_wr *recv_wr,
 		   struct ib_recv_wr **bad_recv_wr)
 {
-	return internal_post_recv(container_of(qp, struct ehca_qp, ib_qp),
+	struct ehca_qp *my_qp = container_of(qp, struct ehca_qp, ib_qp);
-				  qp->device, recv_wr, bad_recv_wr);
+
+	/* Reject WR if QP is in RESET state */
+	if (unlikely(my_qp->state == IB_QPS_RESET)) {
+		ehca_err(qp->device, "Invalid QP state  qp_state=%d qpn=%x",
+			 my_qp->state, qp->qp_num);
+		return -EINVAL;
+	}
+
+	return internal_post_recv(my_qp, qp->device, recv_wr, bad_recv_wr);
 }
 
 int ehca_post_srq_recv(struct ib_srq *srq,
@@ -681,7 +689,7 @@ poll_cq_one_read_cqe:
 	wc->dlid_path_bits = cqe->dlid;
 	wc->src_qp = cqe->remote_qp_number;
 	wc->wc_flags = cqe->w_completion_flags;
-	wc->imm_data = cpu_to_be32(cqe->immediate_data);
+	wc->ex.imm_data = cpu_to_be32(cqe->immediate_data);
 	wc->sl = cqe->service_level;
 
 poll_cq_one_exit0:

drivers/infiniband/hw/ehca/hcp_if.c

@@ -933,3 +933,13 @@ u64 hipz_h_error_data(const struct ipz_adapter_handle adapter_handle,
 				       r_cb,
 				       0, 0, 0, 0);
 }
+
+u64 hipz_h_eoi(int irq)
+{
+	unsigned long xirr;
+
+	iosync();
+	xirr = (0xffULL << 24) | irq;
+
+	return plpar_hcall_norets(H_EOI, xirr);
+}

drivers/infiniband/hw/ehca/hcp_if.h

@@ -260,5 +260,6 @@ u64 hipz_h_error_data(const struct ipz_adapter_handle adapter_handle,
 		      const u64 ressource_handle,
 		      void *rblock,
 		      unsigned long *byte_count);
+u64 hipz_h_eoi(int irq);
 
 #endif /* __HCP_IF_H__ */

drivers/infiniband/hw/ipath/ipath_cq.c

@@ -82,7 +82,7 @@ void ipath_cq_enter(struct ipath_cq *cq, struct ib_wc *entry, int solicited)
 		wc->uqueue[head].opcode = entry->opcode;
 		wc->uqueue[head].vendor_err = entry->vendor_err;
 		wc->uqueue[head].byte_len = entry->byte_len;
-		wc->uqueue[head].imm_data = (__u32 __force)entry->imm_data;
+		wc->uqueue[head].ex.imm_data = (__u32 __force) entry->ex.imm_data;
 		wc->uqueue[head].qp_num = entry->qp->qp_num;
 		wc->uqueue[head].src_qp = entry->src_qp;
 		wc->uqueue[head].wc_flags = entry->wc_flags;

drivers/infiniband/hw/ipath/ipath_iba7220.c

@@ -2228,8 +2228,8 @@ static void ipath_autoneg_send(struct ipath_devdata *dd, int which)
 		0xffffffff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0,
 		0x40000001, 0x1388, 0x15e, /* rest 0's */
 		};
-	dcnt = sizeof(madpayload_start)/sizeof(madpayload_start[0]);
+	dcnt = ARRAY_SIZE(madpayload_start);
-	hcnt = sizeof(hdr)/sizeof(hdr[0]);
+	hcnt = ARRAY_SIZE(hdr);
 	if (!swapped) {
 		/* for maintainability, do it at runtime */
 		for (i = 0; i < hcnt; i++) {

drivers/infiniband/hw/ipath/ipath_mad.c

@@ -111,9 +111,9 @@ static int recv_subn_get_nodeinfo(struct ib_smp *smp,
 	nip->revision = cpu_to_be32((majrev << 16) | minrev);
 	nip->local_port_num = port;
 	vendor = dd->ipath_vendorid;
-	nip->vendor_id[0] = 0;
+	nip->vendor_id[0] = IPATH_SRC_OUI_1;
-	nip->vendor_id[1] = vendor >> 8;
+	nip->vendor_id[1] = IPATH_SRC_OUI_2;
-	nip->vendor_id[2] = vendor;
+	nip->vendor_id[2] = IPATH_SRC_OUI_3;
 
 	return reply(smp);
 }

drivers/infiniband/hw/ipath/ipath_rc.c

@@ -1703,11 +1703,11 @@ void ipath_rc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 	case OP(SEND_LAST_WITH_IMMEDIATE):
 	send_last_imm:
 		if (header_in_data) {
-			wc.imm_data = *(__be32 *) data;
+			wc.ex.imm_data = *(__be32 *) data;
 			data += sizeof(__be32);
 		} else {
 			/* Immediate data comes after BTH */
-			wc.imm_data = ohdr->u.imm_data;
+			wc.ex.imm_data = ohdr->u.imm_data;
 		}
 		hdrsize += 4;
 		wc.wc_flags = IB_WC_WITH_IMM;

drivers/infiniband/hw/ipath/ipath_ruc.c

@@ -331,7 +331,7 @@ again:
 	switch (wqe->wr.opcode) {
 	case IB_WR_SEND_WITH_IMM:
 		wc.wc_flags = IB_WC_WITH_IMM;
-		wc.imm_data = wqe->wr.ex.imm_data;
+		wc.ex.imm_data = wqe->wr.ex.imm_data;
 		/* FALLTHROUGH */
 	case IB_WR_SEND:
 		if (!ipath_get_rwqe(qp, 0))
@@ -342,7 +342,7 @@ again:
 		if (unlikely(!(qp->qp_access_flags & IB_ACCESS_REMOTE_WRITE)))
 			goto inv_err;
 		wc.wc_flags = IB_WC_WITH_IMM;
-		wc.imm_data = wqe->wr.ex.imm_data;
+		wc.ex.imm_data = wqe->wr.ex.imm_data;
 		if (!ipath_get_rwqe(qp, 1))
 			goto rnr_nak;
 		/* FALLTHROUGH */

drivers/infiniband/hw/ipath/ipath_uc.c

@@ -379,11 +379,11 @@ void ipath_uc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 	case OP(SEND_LAST_WITH_IMMEDIATE):
 	send_last_imm:
 		if (header_in_data) {
-			wc.imm_data = *(__be32 *) data;
+			wc.ex.imm_data = *(__be32 *) data;
 			data += sizeof(__be32);
 		} else {
 			/* Immediate data comes after BTH */
-			wc.imm_data = ohdr->u.imm_data;
+			wc.ex.imm_data = ohdr->u.imm_data;
 		}
 		hdrsize += 4;
 		wc.wc_flags = IB_WC_WITH_IMM;
@@ -483,11 +483,11 @@ void ipath_uc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 	case OP(RDMA_WRITE_LAST_WITH_IMMEDIATE):
 	rdma_last_imm:
 		if (header_in_data) {
-			wc.imm_data = *(__be32 *) data;
+			wc.ex.imm_data = *(__be32 *) data;
 			data += sizeof(__be32);
 		} else {
 			/* Immediate data comes after BTH */
-			wc.imm_data = ohdr->u.imm_data;
+			wc.ex.imm_data = ohdr->u.imm_data;
 		}
 		hdrsize += 4;
 		wc.wc_flags = IB_WC_WITH_IMM;

drivers/infiniband/hw/ipath/ipath_ud.c

@@ -96,7 +96,7 @@ static void ipath_ud_loopback(struct ipath_qp *sqp, struct ipath_swqe *swqe)
 
 	if (swqe->wr.opcode == IB_WR_SEND_WITH_IMM) {
 		wc.wc_flags = IB_WC_WITH_IMM;
-		wc.imm_data = swqe->wr.ex.imm_data;
+		wc.ex.imm_data = swqe->wr.ex.imm_data;
 	}
 
 	/*
@@ -492,14 +492,14 @@ void ipath_ud_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr,
 	if (qp->ibqp.qp_num > 1 &&
 	    opcode == IB_OPCODE_UD_SEND_ONLY_WITH_IMMEDIATE) {
 		if (header_in_data) {
-			wc.imm_data = *(__be32 *) data;
+			wc.ex.imm_data = *(__be32 *) data;
 			data += sizeof(__be32);
 		} else
-			wc.imm_data = ohdr->u.ud.imm_data;
+			wc.ex.imm_data = ohdr->u.ud.imm_data;
 		wc.wc_flags = IB_WC_WITH_IMM;
 		hdrsize += sizeof(u32);
 	} else if (opcode == IB_OPCODE_UD_SEND_ONLY) {
-		wc.imm_data = 0;
+		wc.ex.imm_data = 0;
 		wc.wc_flags = 0;
 	} else {
 		dev->n_pkt_drops++;

drivers/infiniband/hw/ipath/ipath_verbs.c

@@ -1497,7 +1497,8 @@ static int ipath_query_device(struct ib_device *ibdev,
 		IB_DEVICE_SYS_IMAGE_GUID | IB_DEVICE_RC_RNR_NAK_GEN |
 		IB_DEVICE_PORT_ACTIVE_EVENT | IB_DEVICE_SRQ_RESIZE;
 	props->page_size_cap = PAGE_SIZE;
-	props->vendor_id = dev->dd->ipath_vendorid;
+	props->vendor_id =
+		IPATH_SRC_OUI_1 << 16 | IPATH_SRC_OUI_2 << 8 | IPATH_SRC_OUI_3;
 	props->vendor_part_id = dev->dd->ipath_deviceid;
 	props->hw_ver = dev->dd->ipath_pcirev;
 

drivers/infiniband/hw/mlx4/cq.c

@@ -663,18 +663,18 @@ repoll:
 
 		switch (cqe->owner_sr_opcode & MLX4_CQE_OPCODE_MASK) {
 		case MLX4_RECV_OPCODE_RDMA_WRITE_IMM:
-			wc->opcode   = IB_WC_RECV_RDMA_WITH_IMM;
+			wc->opcode	= IB_WC_RECV_RDMA_WITH_IMM;
-			wc->wc_flags = IB_WC_WITH_IMM;
+			wc->wc_flags	= IB_WC_WITH_IMM;
-			wc->imm_data = cqe->immed_rss_invalid;
+			wc->ex.imm_data = cqe->immed_rss_invalid;
 			break;
 		case MLX4_RECV_OPCODE_SEND:
 			wc->opcode   = IB_WC_RECV;
 			wc->wc_flags = 0;
 			break;
 		case MLX4_RECV_OPCODE_SEND_IMM:
-			wc->opcode   = IB_WC_RECV;
+			wc->opcode	= IB_WC_RECV;
-			wc->wc_flags = IB_WC_WITH_IMM;
+			wc->wc_flags	= IB_WC_WITH_IMM;
-			wc->imm_data = cqe->immed_rss_invalid;
+			wc->ex.imm_data = cqe->immed_rss_invalid;
 			break;
 		}
 

drivers/infiniband/hw/mlx4/mad.c

@@ -255,7 +255,8 @@ int mlx4_ib_process_mad(struct ib_device *ibdev, int mad_flags,	u8 port_num,
 			return IB_MAD_RESULT_SUCCESS;
 	} else if (in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_PERF_MGMT ||
 		   in_mad->mad_hdr.mgmt_class == MLX4_IB_VENDOR_CLASS1   ||
-		   in_mad->mad_hdr.mgmt_class == MLX4_IB_VENDOR_CLASS2) {
+		   in_mad->mad_hdr.mgmt_class == MLX4_IB_VENDOR_CLASS2   ||
+		   in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_CONG_MGMT) {
 		if (in_mad->mad_hdr.method  != IB_MGMT_METHOD_GET &&
 		    in_mad->mad_hdr.method  != IB_MGMT_METHOD_SET)
 			return IB_MAD_RESULT_SUCCESS;

drivers/infiniband/hw/mlx4/main.c

@@ -90,7 +90,8 @@ static int mlx4_ib_query_device(struct ib_device *ibdev,
 	props->device_cap_flags    = IB_DEVICE_CHANGE_PHY_PORT |
 		IB_DEVICE_PORT_ACTIVE_EVENT		|
 		IB_DEVICE_SYS_IMAGE_GUID		|
-		IB_DEVICE_RC_RNR_NAK_GEN;
+		IB_DEVICE_RC_RNR_NAK_GEN		|
+		IB_DEVICE_BLOCK_MULTICAST_LOOPBACK;
 	if (dev->dev->caps.flags & MLX4_DEV_CAP_FLAG_BAD_PKEY_CNTR)
 		props->device_cap_flags |= IB_DEVICE_BAD_PKEY_CNTR;
 	if (dev->dev->caps.flags & MLX4_DEV_CAP_FLAG_BAD_QKEY_CNTR)
@@ -437,7 +438,9 @@ static int mlx4_ib_dealloc_pd(struct ib_pd *pd)
 static int mlx4_ib_mcg_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
 {
 	return mlx4_multicast_attach(to_mdev(ibqp->device)->dev,
-				     &to_mqp(ibqp)->mqp, gid->raw);
+				     &to_mqp(ibqp)->mqp, gid->raw,
+				     !!(to_mqp(ibqp)->flags &
+					MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK));
 }
 
 static int mlx4_ib_mcg_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)

drivers/infiniband/hw/mlx4/mlx4_ib.h

@@ -101,7 +101,8 @@ struct mlx4_ib_wq {
 };
 
 enum mlx4_ib_qp_flags {
-	MLX4_IB_QP_LSO		= 1 << 0
+	MLX4_IB_QP_LSO				= 1 << 0,
+	MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK	= 1 << 1,
 };
 
 struct mlx4_ib_qp {

drivers/infiniband/hw/mlx4/qp.c

@@ -129,9 +129,10 @@ static void stamp_send_wqe(struct mlx4_ib_qp *qp, int n, int size)
 	int ind;
 	void *buf;
 	__be32 stamp;
+	struct mlx4_wqe_ctrl_seg *ctrl;
 
-	s = roundup(size, 1U << qp->sq.wqe_shift);
 	if (qp->sq_max_wqes_per_wr > 1) {
+		s = roundup(size, 1U << qp->sq.wqe_shift);
 		for (i = 0; i < s; i += 64) {
 			ind = (i >> qp->sq.wqe_shift) + n;
 			stamp = ind & qp->sq.wqe_cnt ? cpu_to_be32(0x7fffffff) :
@@ -141,7 +142,8 @@ static void stamp_send_wqe(struct mlx4_ib_qp *qp, int n, int size)
 			*wqe = stamp;
 		}
 	} else {
-		buf = get_send_wqe(qp, n & (qp->sq.wqe_cnt - 1));
+		ctrl = buf = get_send_wqe(qp, n & (qp->sq.wqe_cnt - 1));
+		s = (ctrl->fence_size & 0x3f) << 4;
 		for (i = 64; i < s; i += 64) {
 			wqe = buf + i;
 			*wqe = cpu_to_be32(0xffffffff);
@@ -452,19 +454,8 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 	spin_lock_init(&qp->rq.lock);
 
 	qp->state	 = IB_QPS_RESET;
-	qp->atomic_rd_en = 0;
-	qp->resp_depth   = 0;
-
-	qp->rq.head	    = 0;
-	qp->rq.tail	    = 0;
-	qp->sq.head	    = 0;
-	qp->sq.tail	    = 0;
-	qp->sq_next_wqe     = 0;
-
 	if (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR)
 		qp->sq_signal_bits = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE);
-	else
-		qp->sq_signal_bits = 0;
 
 	err = set_rq_size(dev, &init_attr->cap, !!pd->uobject, !!init_attr->srq, qp);
 	if (err)
@@ -509,6 +500,9 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 	} else {
 		qp->sq_no_prefetch = 0;
 
+		if (init_attr->create_flags & IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK)
+			qp->flags |= MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK;
+
 		if (init_attr->create_flags & IB_QP_CREATE_IPOIB_UD_LSO)
 			qp->flags |= MLX4_IB_QP_LSO;
 
@@ -682,10 +676,15 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 	struct mlx4_ib_qp *qp;
 	int err;
 
-	/* We only support LSO, and only for kernel UD QPs. */
+	/*
-	if (init_attr->create_flags & ~IB_QP_CREATE_IPOIB_UD_LSO)
+	 * We only support LSO and multicast loopback blocking, and
+	 * only for kernel UD QPs.
+	 */
+	if (init_attr->create_flags & ~(IB_QP_CREATE_IPOIB_UD_LSO |
+					IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK))
 		return ERR_PTR(-EINVAL);
-	if (init_attr->create_flags & IB_QP_CREATE_IPOIB_UD_LSO &&
+
+	if (init_attr->create_flags &&
 	    (pd->uobject || init_attr->qp_type != IB_QPT_UD))
 		return ERR_PTR(-EINVAL);
 
@@ -694,7 +693,7 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 	case IB_QPT_UC:
 	case IB_QPT_UD:
 	{
-		qp = kmalloc(sizeof *qp, GFP_KERNEL);
+		qp = kzalloc(sizeof *qp, GFP_KERNEL);
 		if (!qp)
 			return ERR_PTR(-ENOMEM);
 
@@ -715,7 +714,7 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 		if (pd->uobject)
 			return ERR_PTR(-EINVAL);
 
-		sqp = kmalloc(sizeof *sqp, GFP_KERNEL);
+		sqp = kzalloc(sizeof *sqp, GFP_KERNEL);
 		if (!sqp)
 			return ERR_PTR(-ENOMEM);
 
@@ -906,7 +905,8 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 			       attr->path_mtu);
 			goto out;
 		}
-		context->mtu_msgmax = (attr->path_mtu << 5) | 31;
+		context->mtu_msgmax = (attr->path_mtu << 5) |
+			ilog2(dev->dev->caps.max_msg_sz);
 	}
 
 	if (qp->rq.wqe_cnt)
@@ -1063,6 +1063,8 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 		for (i = 0; i < qp->sq.wqe_cnt; ++i) {
 			ctrl = get_send_wqe(qp, i);
 			ctrl->owner_opcode = cpu_to_be32(1 << 31);
+			if (qp->sq_max_wqes_per_wr == 1)
+				ctrl->fence_size = 1 << (qp->sq.wqe_shift - 4);
 
 			stamp_send_wqe(qp, i, 1 << qp->sq.wqe_shift);
 		}
@@ -1127,23 +1129,6 @@ out:
 	return err;
 }
 
-static const struct ib_qp_attr mlx4_ib_qp_attr = { .port_num = 1 };
-static const int mlx4_ib_qp_attr_mask_table[IB_QPT_UD + 1] = {
-		[IB_QPT_UD]  = (IB_QP_PKEY_INDEX		|
-				IB_QP_PORT			|
-				IB_QP_QKEY),
-		[IB_QPT_UC]  = (IB_QP_PKEY_INDEX		|
-				IB_QP_PORT			|
-				IB_QP_ACCESS_FLAGS),
-		[IB_QPT_RC]  = (IB_QP_PKEY_INDEX		|
-				IB_QP_PORT			|
-				IB_QP_ACCESS_FLAGS),
-		[IB_QPT_SMI] = (IB_QP_PKEY_INDEX		|
-				IB_QP_QKEY),
-		[IB_QPT_GSI] = (IB_QP_PKEY_INDEX		|
-				IB_QP_QKEY),
-};
-
 int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 		      int attr_mask, struct ib_udata *udata)
 {
@@ -1186,15 +1171,6 @@ int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 		goto out;
 	}
 
-	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_ERR) {
-		err = __mlx4_ib_modify_qp(ibqp, &mlx4_ib_qp_attr,
-					  mlx4_ib_qp_attr_mask_table[ibqp->qp_type],
-					  IB_QPS_RESET, IB_QPS_INIT);
-		if (err)
-			goto out;
-		cur_state = IB_QPS_INIT;
-	}
-
 	err = __mlx4_ib_modify_qp(ibqp, attr, attr_mask, cur_state, new_state);
 
 out:
@@ -1865,6 +1841,13 @@ done:
 
 	qp_init_attr->cap	     = qp_attr->cap;
 
+	qp_init_attr->create_flags = 0;
+	if (qp->flags & MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK)
+		qp_init_attr->create_flags |= IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK;
+
+	if (qp->flags & MLX4_IB_QP_LSO)
+		qp_init_attr->create_flags |= IB_QP_CREATE_IPOIB_UD_LSO;
+
 out:
 	mutex_unlock(&qp->mutex);
 	return err;

drivers/infiniband/hw/mthca/mthca_allocator.c

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_allocator.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/errno.h>

drivers/infiniband/hw/mthca/mthca_av.c

@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_av.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/string.h>

drivers/infiniband/hw/mthca/mthca_catas.c

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id$
  */
 
 #include <linux/jiffies.h>
@@ -128,7 +126,6 @@ static void handle_catas(struct mthca_dev *dev)
 static void poll_catas(unsigned long dev_ptr)
 {
 	struct mthca_dev *dev = (struct mthca_dev *) dev_ptr;
-	unsigned long flags;
 	int i;
 
 	for (i = 0; i < dev->catas_err.size; ++i)
@@ -137,13 +134,8 @@ static void poll_catas(unsigned long dev_ptr)
 			return;
 		}
 
-	spin_lock_irqsave(&catas_lock, flags);
+	mod_timer(&dev->catas_err.timer,
-	if (!dev->catas_err.stop)
+		  round_jiffies(jiffies + MTHCA_CATAS_POLL_INTERVAL));
-		mod_timer(&dev->catas_err.timer,
-			  jiffies + MTHCA_CATAS_POLL_INTERVAL);
-	spin_unlock_irqrestore(&catas_lock, flags);
-
-	return;
 }
 
 void mthca_start_catas_poll(struct mthca_dev *dev)
@@ -151,7 +143,6 @@ void mthca_start_catas_poll(struct mthca_dev *dev)
 	unsigned long addr;
 
 	init_timer(&dev->catas_err.timer);
-	dev->catas_err.stop = 0;
 	dev->catas_err.map  = NULL;
 
 	addr = pci_resource_start(dev->pdev, 0) +
@@ -182,10 +173,6 @@ void mthca_start_catas_poll(struct mthca_dev *dev)
 
 void mthca_stop_catas_poll(struct mthca_dev *dev)
 {
-	spin_lock_irq(&catas_lock);
-	dev->catas_err.stop = 1;
-	spin_unlock_irq(&catas_lock);
-
 	del_timer_sync(&dev->catas_err.timer);
 
 	if (dev->catas_err.map) {

drivers/infiniband/hw/mthca/mthca_cmd.c

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_cmd.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/completion.h>

drivers/infiniband/hw/mthca/mthca_cmd.h

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_cmd.h 1349 2004-12-16 21:09:43Z roland $
  */
 
 #ifndef MTHCA_CMD_H

drivers/infiniband/hw/mthca/mthca_config_reg.h

@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_config_reg.h 1349 2004-12-16 21:09:43Z roland $
  */
 
 #ifndef MTHCA_CONFIG_REG_H

drivers/infiniband/hw/mthca/mthca_cq.c

@@ -32,8 +32,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_cq.c 1369 2004-12-20 16:17:07Z roland $
  */
 
 #include <linux/hardirq.h>
@@ -622,13 +620,13 @@ static inline int mthca_poll_one(struct mthca_dev *dev,
 		case IB_OPCODE_SEND_LAST_WITH_IMMEDIATE:
 		case IB_OPCODE_SEND_ONLY_WITH_IMMEDIATE:
 			entry->wc_flags = IB_WC_WITH_IMM;
-			entry->imm_data = cqe->imm_etype_pkey_eec;
+			entry->ex.imm_data = cqe->imm_etype_pkey_eec;
 			entry->opcode = IB_WC_RECV;
 			break;
 		case IB_OPCODE_RDMA_WRITE_LAST_WITH_IMMEDIATE:
 		case IB_OPCODE_RDMA_WRITE_ONLY_WITH_IMMEDIATE:
 			entry->wc_flags = IB_WC_WITH_IMM;
-			entry->imm_data = cqe->imm_etype_pkey_eec;
+			entry->ex.imm_data = cqe->imm_etype_pkey_eec;
 			entry->opcode = IB_WC_RECV_RDMA_WITH_IMM;
 			break;
 		default:

drivers/infiniband/hw/mthca/mthca_dev.h

@@ -32,8 +32,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_dev.h 1349 2004-12-16 21:09:43Z roland $
  */
 
 #ifndef MTHCA_DEV_H
@@ -279,7 +277,6 @@ struct mthca_mcg_table {
 struct mthca_catas_err {
 	u64			addr;
 	u32 __iomem	       *map;
-	unsigned long		stop;
 	u32			size;
 	struct timer_list	timer;
 	struct list_head	list;

drivers/infiniband/hw/mthca/mthca_doorbell.h

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_doorbell.h 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/types.h>

drivers/infiniband/hw/mthca/mthca_eq.c

@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_eq.c 1382 2004-12-24 02:21:02Z roland $
  */
 
 #include <linux/errno.h>

drivers/infiniband/hw/mthca/mthca_mad.c

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_mad.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/string.h>

drivers/infiniband/hw/mthca/mthca_main.c

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_main.c 1396 2004-12-28 04:10:27Z roland $
  */
 
 #include <linux/module.h>

drivers/infiniband/hw/mthca/mthca_mcg.c

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_mcg.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/string.h>

drivers/infiniband/hw/mthca/mthca_memfree.c

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id$
  */
 
 #include <linux/mm.h>

drivers/infiniband/hw/mthca/mthca_memfree.h

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id$
  */
 
 #ifndef MTHCA_MEMFREE_H

drivers/infiniband/hw/mthca/mthca_mr.c

@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_mr.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/slab.h>

drivers/infiniband/hw/mthca/mthca_pd.c

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_pd.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/errno.h>

drivers/infiniband/hw/mthca/mthca_profile.c

@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_profile.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/module.h>

drivers/infiniband/hw/mthca/mthca_profile.h

@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_profile.h 1349 2004-12-16 21:09:43Z roland $
  */
 
 #ifndef MTHCA_PROFILE_H

drivers/infiniband/hw/mthca/mthca_provider.c

@@ -32,8 +32,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_provider.c 4859 2006-01-09 21:55:10Z roland $
  */
 
 #include <rdma/ib_smi.h>

drivers/infiniband/hw/mthca/mthca_provider.h

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_provider.h 1349 2004-12-16 21:09:43Z roland $
  */
 
 #ifndef MTHCA_PROVIDER_H

drivers/infiniband/hw/mthca/mthca_qp.c

@@ -31,8 +31,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_qp.c 1355 2004-12-17 15:23:43Z roland $
  */
 
 #include <linux/string.h>
@@ -850,23 +848,6 @@ out:
 	return err;
 }
 
-static const struct ib_qp_attr dummy_init_attr = { .port_num = 1 };
-static const int dummy_init_attr_mask[] = {
-	[IB_QPT_UD]  = (IB_QP_PKEY_INDEX		|
-			IB_QP_PORT			|
-			IB_QP_QKEY),
-	[IB_QPT_UC]  = (IB_QP_PKEY_INDEX		|
-			IB_QP_PORT			|
-			IB_QP_ACCESS_FLAGS),
-	[IB_QPT_RC]  = (IB_QP_PKEY_INDEX		|
-			IB_QP_PORT			|
-			IB_QP_ACCESS_FLAGS),
-	[IB_QPT_SMI] = (IB_QP_PKEY_INDEX		|
-			IB_QP_QKEY),
-	[IB_QPT_GSI] = (IB_QP_PKEY_INDEX		|
-			IB_QP_QKEY),
-};
-
 int mthca_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask,
 		    struct ib_udata *udata)
 {
@@ -928,15 +909,6 @@ int mthca_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask,
 		goto out;
 	}
 
-	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_ERR) {
-		err = __mthca_modify_qp(ibqp, &dummy_init_attr,
-					dummy_init_attr_mask[ibqp->qp_type],
-					IB_QPS_RESET, IB_QPS_INIT);
-		if (err)
-			goto out;
-		cur_state = IB_QPS_INIT;
-	}
-
 	err = __mthca_modify_qp(ibqp, attr, attr_mask, cur_state, new_state);
 
 out:
@@ -1277,10 +1249,10 @@ static int mthca_set_qp_size(struct mthca_dev *dev, struct ib_qp_cap *cap,
 		return -EINVAL;
 
 	/*
-	 * For MLX transport we need 2 extra S/G entries:
+	 * For MLX transport we need 2 extra send gather entries:
 	 * one for the header and one for the checksum at the end
 	 */
-	if (qp->transport == MLX && cap->max_recv_sge + 2 > dev->limits.max_sg)
+	if (qp->transport == MLX && cap->max_send_sge + 2 > dev->limits.max_sg)
 		return -EINVAL;
 
 	if (mthca_is_memfree(dev)) {

drivers/infiniband/hw/mthca/mthca_reset.c

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_reset.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/init.h>

drivers/infiniband/hw/mthca/mthca_srq.c

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_srq.c 3047 2005-08-10 03:59:35Z roland $
  */
 
 #include <linux/slab.h>

drivers/infiniband/hw/mthca/mthca_uar.c

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id$
  */
 
 #include <asm/page.h>		/* PAGE_SHIFT */

drivers/infiniband/hw/mthca/mthca_user.h

@@ -29,7 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
  */
 
 #ifndef MTHCA_USER_H

drivers/infiniband/hw/mthca/mthca_wqe.h

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: mthca_wqe.h 3047 2005-08-10 03:59:35Z roland $
  */
 
 #ifndef MTHCA_WQE_H

drivers/infiniband/hw/nes/nes.c

@@ -328,7 +328,7 @@ void nes_rem_ref(struct ib_qp *ibqp)
 		set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_WQE_ID_IDX, nesqp->hwqp.qp_id);
 		u64temp = (u64)nesqp->nesqp_context_pbase;
 		set_wqe_64bit_value(cqp_wqe->wqe_words, NES_CQP_QP_WQE_CONTEXT_LOW_IDX, u64temp);
-		nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+		nes_post_cqp_request(nesdev, cqp_request);
 	}
 }
 

drivers/infiniband/hw/nes/nes.h

@@ -94,9 +94,6 @@
 
 #define MAX_DPC_ITERATIONS               128
 
-#define NES_CQP_REQUEST_NO_DOORBELL_RING 0
-#define NES_CQP_REQUEST_RING_DOORBELL    1
-
 #define NES_DRV_OPT_ENABLE_MPA_VER_0     0x00000001
 #define NES_DRV_OPT_DISABLE_MPA_CRC      0x00000002
 #define NES_DRV_OPT_DISABLE_FIRST_WRITE  0x00000004
@@ -538,7 +535,11 @@ void nes_read_1G_phy_reg(struct nes_device *, u8, u8, u16 *);
 void nes_write_10G_phy_reg(struct nes_device *, u16, u8, u16, u16);
 void nes_read_10G_phy_reg(struct nes_device *, u8, u8, u16);
 struct nes_cqp_request *nes_get_cqp_request(struct nes_device *);
-void nes_post_cqp_request(struct nes_device *, struct nes_cqp_request *, int);
+void nes_free_cqp_request(struct nes_device *nesdev,
+			  struct nes_cqp_request *cqp_request);
+void nes_put_cqp_request(struct nes_device *nesdev,
+			 struct nes_cqp_request *cqp_request);
+void nes_post_cqp_request(struct nes_device *, struct nes_cqp_request *);
 int nes_arp_table(struct nes_device *, u32, u8 *, u32);
 void nes_mh_fix(unsigned long);
 void nes_clc(unsigned long);

drivers/infiniband/hw/nes/nes_cm.c

@@ -1603,7 +1603,6 @@ static struct nes_cm_listener *mini_cm_listen(struct nes_cm_core *cm_core,
 			return NULL;
 		}
 
-		memset(listener, 0, sizeof(struct nes_cm_listener));
 		listener->loc_addr = htonl(cm_info->loc_addr);
 		listener->loc_port = htons(cm_info->loc_port);
 		listener->reused_node = 0;

drivers/infiniband/hw/nes/nes_hw.c

@@ -398,7 +398,7 @@ struct nes_adapter *nes_init_adapter(struct nes_device *nesdev, u8 hw_rev) {
 	nesadapter->base_pd = 1;
 
 	nesadapter->device_cap_flags =
-		IB_DEVICE_ZERO_STAG | IB_DEVICE_MEM_WINDOW;
+		IB_DEVICE_LOCAL_DMA_LKEY | IB_DEVICE_MEM_WINDOW;
 
 	nesadapter->allocated_qps = (unsigned long *)&(((unsigned char *)nesadapter)
 			[(sizeof(struct nes_adapter)+(sizeof(unsigned long)-1))&(~(sizeof(unsigned long)-1))]);
@@ -2710,39 +2710,11 @@ static void nes_cqp_ce_handler(struct nes_device *nesdev, struct nes_hw_cq *cq)
 					barrier();
 					cqp_request->request_done = 1;
 					wake_up(&cqp_request->waitq);
-					if (atomic_dec_and_test(&cqp_request->refcount)) {
+					nes_put_cqp_request(nesdev, cqp_request);
-						nes_debug(NES_DBG_CQP, "CQP request %p (opcode 0x%02X) freed.\n",
-								cqp_request,
-								le32_to_cpu(cqp_request->cqp_wqe.wqe_words[NES_CQP_WQE_OPCODE_IDX])&0x3f);
-						if (cqp_request->dynamic) {
-							kfree(cqp_request);
-						} else {
-							spin_lock_irqsave(&nesdev->cqp.lock, flags);
-							list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-							spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-						}
-					}
-				} else if (cqp_request->callback) {
-					/* Envoke the callback routine */
-					cqp_request->cqp_callback(nesdev, cqp_request);
-					if (cqp_request->dynamic) {
-						kfree(cqp_request);
-					} else {
-						spin_lock_irqsave(&nesdev->cqp.lock, flags);
-						list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-						spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-					}
 				} else {
-					nes_debug(NES_DBG_CQP, "CQP request %p (opcode 0x%02X) freed.\n",
+					if (cqp_request->callback)
-							cqp_request,
+						cqp_request->cqp_callback(nesdev, cqp_request);
-							le32_to_cpu(cqp_request->cqp_wqe.wqe_words[NES_CQP_WQE_OPCODE_IDX]) & 0x3f);
+					nes_free_cqp_request(nesdev, cqp_request);
-					if (cqp_request->dynamic) {
-						kfree(cqp_request);
-					} else {
-						spin_lock_irqsave(&nesdev->cqp.lock, flags);
-						list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-						spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-					}
 				}
 			} else {
 				wake_up(&nesdev->cqp.waitq);
@@ -3149,7 +3121,6 @@ int nes_manage_apbvt(struct nes_vnic *nesvnic, u32 accel_local_port,
 {
 	struct nes_device *nesdev = nesvnic->nesdev;
 	struct nes_hw_cqp_wqe *cqp_wqe;
-	unsigned long flags;
 	struct nes_cqp_request *cqp_request;
 	int ret = 0;
 	u16 major_code;
@@ -3176,7 +3147,7 @@ int nes_manage_apbvt(struct nes_vnic *nesvnic, u32 accel_local_port,
 	nes_debug(NES_DBG_QP, "Waiting for CQP completion for APBVT.\n");
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	if (add_port == NES_MANAGE_APBVT_ADD)
 		ret = wait_event_timeout(cqp_request->waitq, (cqp_request->request_done != 0),
@@ -3184,15 +3155,9 @@ int nes_manage_apbvt(struct nes_vnic *nesvnic, u32 accel_local_port,
 	nes_debug(NES_DBG_QP, "Completed, ret=%u,  CQP Major:Minor codes = 0x%04X:0x%04X\n",
 			ret, cqp_request->major_code, cqp_request->minor_code);
 	major_code = cqp_request->major_code;
-	if (atomic_dec_and_test(&cqp_request->refcount)) {
+
-		if (cqp_request->dynamic) {
+	nes_put_cqp_request(nesdev, cqp_request);
-			kfree(cqp_request);
+
-		} else {
-			spin_lock_irqsave(&nesdev->cqp.lock, flags);
-			list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-			spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-		}
-	}
 	if (!ret)
 		return -ETIME;
 	else if (major_code)
@@ -3252,7 +3217,7 @@ void nes_manage_arp_cache(struct net_device *netdev, unsigned char *mac_addr,
 			nesdev->cqp.sq_head, nesdev->cqp.sq_tail);
 
 	atomic_set(&cqp_request->refcount, 1);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 }
 
 
@@ -3262,7 +3227,6 @@ void nes_manage_arp_cache(struct net_device *netdev, unsigned char *mac_addr,
 void flush_wqes(struct nes_device *nesdev, struct nes_qp *nesqp,
 		u32 which_wq, u32 wait_completion)
 {
-	unsigned long flags;
 	struct nes_cqp_request *cqp_request;
 	struct nes_hw_cqp_wqe *cqp_wqe;
 	int ret;
@@ -3285,7 +3249,7 @@ void flush_wqes(struct nes_device *nesdev, struct nes_qp *nesqp,
 			cpu_to_le32(NES_CQP_FLUSH_WQES | which_wq);
 	cqp_wqe->wqe_words[NES_CQP_WQE_ID_IDX] = cpu_to_le32(nesqp->hwqp.qp_id);
 
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	if (wait_completion) {
 		/* Wait for CQP */
@@ -3294,14 +3258,6 @@ void flush_wqes(struct nes_device *nesdev, struct nes_qp *nesqp,
 		nes_debug(NES_DBG_QP, "Flush SQ QP WQEs completed, ret=%u,"
 				" CQP Major:Minor codes = 0x%04X:0x%04X\n",
 				ret, cqp_request->major_code, cqp_request->minor_code);
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
+		nes_put_cqp_request(nesdev, cqp_request);
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
 	}
 }

drivers/infiniband/hw/nes/nes_hw.h

@@ -1172,7 +1172,7 @@ struct nes_vnic {
 	u32    mcrq_qp_id;
 	struct nes_ucontext *mcrq_ucontext;
 	struct nes_cqp_request* (*get_cqp_request)(struct nes_device *nesdev);
-	void (*post_cqp_request)(struct nes_device*, struct nes_cqp_request *, int);
+	void (*post_cqp_request)(struct nes_device*, struct nes_cqp_request *);
 	int (*mcrq_mcast_filter)( struct nes_vnic* nesvnic, __u8* dmi_addr );
 	struct net_device_stats netstats;
 	/* used to put the netdev on the adapters logical port list */

drivers/infiniband/hw/nes/nes_utils.c

@@ -567,12 +567,36 @@ struct nes_cqp_request *nes_get_cqp_request(struct nes_device *nesdev)
 	return cqp_request;
 }
 
+void nes_free_cqp_request(struct nes_device *nesdev,
+			  struct nes_cqp_request *cqp_request)
+{
+	unsigned long flags;
+
+	nes_debug(NES_DBG_CQP, "CQP request %p (opcode 0x%02X) freed.\n",
+		  cqp_request,
+		  le32_to_cpu(cqp_request->cqp_wqe.wqe_words[NES_CQP_WQE_OPCODE_IDX]) & 0x3f);
+
+	if (cqp_request->dynamic) {
+		kfree(cqp_request);
+	} else {
+		spin_lock_irqsave(&nesdev->cqp.lock, flags);
+		list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
+		spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
+	}
+}
+
+void nes_put_cqp_request(struct nes_device *nesdev,
+			 struct nes_cqp_request *cqp_request)
+{
+	if (atomic_dec_and_test(&cqp_request->refcount))
+		nes_free_cqp_request(nesdev, cqp_request);
+}
 
 /**
  * nes_post_cqp_request
  */
 void nes_post_cqp_request(struct nes_device *nesdev,
-		struct nes_cqp_request *cqp_request, int ring_doorbell)
+			  struct nes_cqp_request *cqp_request)
 {
 	struct nes_hw_cqp_wqe *cqp_wqe;
 	unsigned long flags;
@@ -600,10 +624,9 @@ void nes_post_cqp_request(struct nes_device *nesdev,
 				nesdev->cqp.sq_head, nesdev->cqp.sq_tail, nesdev->cqp.sq_size,
 				cqp_request->waiting, atomic_read(&cqp_request->refcount));
 		barrier();
-		if (ring_doorbell) {
+
-			/* Ring doorbell (1 WQEs) */
+		/* Ring doorbell (1 WQEs) */
-			nes_write32(nesdev->regs+NES_WQE_ALLOC, 0x01800000 | nesdev->cqp.qp_id);
+		nes_write32(nesdev->regs+NES_WQE_ALLOC, 0x01800000 | nesdev->cqp.qp_id);
-		}
 
 		barrier();
 	} else {

drivers/infiniband/hw/nes/nes_verbs.c

@@ -55,7 +55,6 @@ static void nes_unregister_ofa_device(struct nes_ib_device *nesibdev);
  * nes_alloc_mw
  */
 static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd) {
-	unsigned long flags;
 	struct nes_pd *nespd = to_nespd(ibpd);
 	struct nes_vnic *nesvnic = to_nesvnic(ibpd->device);
 	struct nes_device *nesdev = nesvnic->nesdev;
@@ -119,7 +118,7 @@ static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd) {
 	set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_STAG_IDX, stag);
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	ret = wait_event_timeout(cqp_request->waitq, (cqp_request->request_done != 0),
@@ -128,15 +127,7 @@ static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd) {
 			" CQP Major:Minor codes = 0x%04X:0x%04X.\n",
 			stag, ret, cqp_request->major_code, cqp_request->minor_code);
 	if ((!ret) || (cqp_request->major_code)) {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
+		nes_put_cqp_request(nesdev, cqp_request);
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
 		kfree(nesmr);
 		nes_free_resource(nesadapter, nesadapter->allocated_mrs, stag_index);
 		if (!ret) {
@@ -144,17 +135,8 @@ static struct ib_mw *nes_alloc_mw(struct ib_pd *ibpd) {
 		} else {
 			return ERR_PTR(-ENOMEM);
 		}
-	} else {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
 	}
+	nes_put_cqp_request(nesdev, cqp_request);
 
 	nesmr->ibmw.rkey = stag;
 	nesmr->mode = IWNES_MEMREG_TYPE_MW;
@@ -178,7 +160,6 @@ static int nes_dealloc_mw(struct ib_mw *ibmw)
 	struct nes_hw_cqp_wqe *cqp_wqe;
 	struct nes_cqp_request *cqp_request;
 	int err = 0;
-	unsigned long flags;
 	int ret;
 
 	/* Deallocate the window with the adapter */
@@ -194,7 +175,7 @@ static int nes_dealloc_mw(struct ib_mw *ibmw)
 	set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_STAG_IDX, ibmw->rkey);
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	nes_debug(NES_DBG_MR, "Waiting for deallocate STag 0x%08X to complete.\n",
@@ -204,32 +185,12 @@ static int nes_dealloc_mw(struct ib_mw *ibmw)
 	nes_debug(NES_DBG_MR, "Deallocate STag completed, wait_event_timeout ret = %u,"
 			" CQP Major:Minor codes = 0x%04X:0x%04X.\n",
 			ret, cqp_request->major_code, cqp_request->minor_code);
-	if ((!ret) || (cqp_request->major_code)) {
+	if (!ret)
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
+		err = -ETIME;
-			if (cqp_request->dynamic) {
+	else if (cqp_request->major_code)
-				kfree(cqp_request);
+		err = -EIO;
-			} else {
+
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
+	nes_put_cqp_request(nesdev, cqp_request);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
-		if (!ret) {
-			err = -ETIME;
-		} else {
-			err = -EIO;
-		}
-	} else {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
-	}
 
 	nes_free_resource(nesadapter, nesadapter->allocated_mrs,
 			(ibmw->rkey & 0x0fffff00) >> 8);
@@ -516,7 +477,7 @@ static struct ib_fmr *nes_alloc_fmr(struct ib_pd *ibpd,
 			(nesfmr->nesmr.pbls_used-1) : nesfmr->nesmr.pbls_used);
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	ret = wait_event_timeout(cqp_request->waitq, (cqp_request->request_done != 0),
@@ -526,29 +487,11 @@ static struct ib_fmr *nes_alloc_fmr(struct ib_pd *ibpd,
 			stag, ret, cqp_request->major_code, cqp_request->minor_code);
 
 	if ((!ret) || (cqp_request->major_code)) {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
+		nes_put_cqp_request(nesdev, cqp_request);
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
 		ret = (!ret) ? -ETIME : -EIO;
 		goto failed_leaf_vpbl_pages_alloc;
-	} else {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
 	}
-
+	nes_put_cqp_request(nesdev, cqp_request);
 	nesfmr->nesmr.ibfmr.lkey = stag;
 	nesfmr->nesmr.ibfmr.rkey = stag;
 	nesfmr->attr = *ibfmr_attr;
@@ -1474,7 +1417,7 @@ static struct ib_qp *nes_create_qp(struct ib_pd *ibpd,
 			set_wqe_64bit_value(cqp_wqe->wqe_words, NES_CQP_QP_WQE_CONTEXT_LOW_IDX, u64temp);
 
 			atomic_set(&cqp_request->refcount, 2);
-			nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+			nes_post_cqp_request(nesdev, cqp_request);
 
 			/* Wait for CQP */
 			nes_debug(NES_DBG_QP, "Waiting for create iWARP QP%u to complete.\n",
@@ -1487,15 +1430,7 @@ static struct ib_qp *nes_create_qp(struct ib_pd *ibpd,
 					nesqp->hwqp.qp_id, ret, nesdev->cqp.sq_head, nesdev->cqp.sq_tail,
 					cqp_request->major_code, cqp_request->minor_code);
 			if ((!ret) || (cqp_request->major_code)) {
-				if (atomic_dec_and_test(&cqp_request->refcount)) {
+				nes_put_cqp_request(nesdev, cqp_request);
-					if (cqp_request->dynamic) {
-						kfree(cqp_request);
-					} else {
-						spin_lock_irqsave(&nesdev->cqp.lock, flags);
-						list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-						spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-					}
-				}
 				nes_free_resource(nesadapter, nesadapter->allocated_qps, qp_num);
 				nes_free_qp_mem(nesdev, nesqp,virt_wqs);
 				kfree(nesqp->allocated_buffer);
@@ -1504,18 +1439,10 @@ static struct ib_qp *nes_create_qp(struct ib_pd *ibpd,
 				} else {
 					return ERR_PTR(-EIO);
 				}
-			} else {
-				if (atomic_dec_and_test(&cqp_request->refcount)) {
-					if (cqp_request->dynamic) {
-						kfree(cqp_request);
-					} else {
-						spin_lock_irqsave(&nesdev->cqp.lock, flags);
-						list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-						spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-					}
-				}
 			}
 
+			nes_put_cqp_request(nesdev, cqp_request);
+
 			if (ibpd->uobject) {
 				uresp.mmap_sq_db_index = nesqp->mmap_sq_db_index;
 				uresp.actual_sq_size = sq_size;
@@ -1817,7 +1744,7 @@ static struct ib_cq *nes_create_cq(struct ib_device *ibdev, int entries,
 			cpu_to_le32(((u32)((u64temp) >> 33)) & 0x7FFFFFFF);
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	nes_debug(NES_DBG_CQ, "Waiting for create iWARP CQ%u to complete.\n",
@@ -1827,32 +1754,15 @@ static struct ib_cq *nes_create_cq(struct ib_device *ibdev, int entries,
 	nes_debug(NES_DBG_CQ, "Create iWARP CQ%u completed, wait_event_timeout ret = %d.\n",
 			nescq->hw_cq.cq_number, ret);
 	if ((!ret) || (cqp_request->major_code)) {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
+		nes_put_cqp_request(nesdev, cqp_request);
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
 		if (!context)
 			pci_free_consistent(nesdev->pcidev, nescq->cq_mem_size, mem,
 					nescq->hw_cq.cq_pbase);
 		nes_free_resource(nesadapter, nesadapter->allocated_cqs, cq_num);
 		kfree(nescq);
 		return ERR_PTR(-EIO);
-	} else {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
 	}
+	nes_put_cqp_request(nesdev, cqp_request);
 
 	if (context) {
 		/* free the nespbl */
@@ -1931,7 +1841,7 @@ static int nes_destroy_cq(struct ib_cq *ib_cq)
 		(nescq->hw_cq.cq_number | ((u32)PCI_FUNC(nesdev->pcidev->devfn) << 16)));
 	nes_free_resource(nesadapter, nesadapter->allocated_cqs, nescq->hw_cq.cq_number);
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	nes_debug(NES_DBG_CQ, "Waiting for destroy iWARP CQ%u to complete.\n",
@@ -1942,37 +1852,18 @@ static int nes_destroy_cq(struct ib_cq *ib_cq)
 			" CQP Major:Minor codes = 0x%04X:0x%04X.\n",
 			nescq->hw_cq.cq_number, ret, cqp_request->major_code,
 			cqp_request->minor_code);
-	if ((!ret) || (cqp_request->major_code)) {
+	if (!ret) {
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
+		nes_debug(NES_DBG_CQ, "iWARP CQ%u destroy timeout expired\n",
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
-		if (!ret) {
-			nes_debug(NES_DBG_CQ, "iWARP CQ%u destroy timeout expired\n",
 					nescq->hw_cq.cq_number);
-			ret = -ETIME;
+		ret = -ETIME;
-		} else {
+	} else if (cqp_request->major_code) {
-			nes_debug(NES_DBG_CQ, "iWARP CQ%u destroy failed\n",
+		nes_debug(NES_DBG_CQ, "iWARP CQ%u destroy failed\n",
 					nescq->hw_cq.cq_number);
-			ret = -EIO;
+		ret = -EIO;
-		}
 	} else {
 		ret = 0;
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
-			if (cqp_request->dynamic) {
-				kfree(cqp_request);
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
 	}
+	nes_put_cqp_request(nesdev, cqp_request);
 
 	if (nescq->cq_mem_size)
 		pci_free_consistent(nesdev->pcidev, nescq->cq_mem_size,
@@ -2096,7 +1987,7 @@ static int nes_reg_mr(struct nes_device *nesdev, struct nes_pd *nespd,
 	barrier();
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	ret = wait_event_timeout(cqp_request->waitq, (0 != cqp_request->request_done),
@@ -2105,15 +1996,8 @@ static int nes_reg_mr(struct nes_device *nesdev, struct nes_pd *nespd,
 			" CQP Major:Minor codes = 0x%04X:0x%04X.\n",
 			stag, ret, cqp_request->major_code, cqp_request->minor_code);
 	major_code = cqp_request->major_code;
-	if (atomic_dec_and_test(&cqp_request->refcount)) {
+	nes_put_cqp_request(nesdev, cqp_request);
-		if (cqp_request->dynamic) {
+
-			kfree(cqp_request);
-		} else {
-			spin_lock_irqsave(&nesdev->cqp.lock, flags);
-			list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-			spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-		}
-	}
 	if (!ret)
 		return -ETIME;
 	else if (major_code)
@@ -2754,7 +2638,7 @@ static int nes_dereg_mr(struct ib_mr *ib_mr)
 	set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_STAG_IDX, ib_mr->rkey);
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	nes_debug(NES_DBG_MR, "Waiting for deallocate STag 0x%08X completed\n", ib_mr->rkey);
@@ -2771,15 +2655,9 @@ static int nes_dereg_mr(struct ib_mr *ib_mr)
 
 	major_code = cqp_request->major_code;
 	minor_code = cqp_request->minor_code;
-	if (atomic_dec_and_test(&cqp_request->refcount)) {
+
-		if (cqp_request->dynamic) {
+	nes_put_cqp_request(nesdev, cqp_request);
-			kfree(cqp_request);
+
-		} else {
-			spin_lock_irqsave(&nesdev->cqp.lock, flags);
-			list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-			spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-		}
-	}
 	if (!ret) {
 		nes_debug(NES_DBG_MR, "Timeout waiting to destroy STag,"
 				" ib_mr=%p, rkey = 0x%08X\n",
@@ -2904,7 +2782,6 @@ int nes_hw_modify_qp(struct nes_device *nesdev, struct nes_qp *nesqp,
 	/* struct iw_cm_id *cm_id = nesqp->cm_id; */
 	/* struct iw_cm_event cm_event; */
 	struct nes_cqp_request *cqp_request;
-	unsigned long flags;
 	int ret;
 	u16 major_code;
 
@@ -2932,7 +2809,7 @@ int nes_hw_modify_qp(struct nes_device *nesdev, struct nes_qp *nesqp,
 	set_wqe_64bit_value(cqp_wqe->wqe_words, NES_CQP_QP_WQE_CONTEXT_LOW_IDX, (u64)nesqp->nesqp_context_pbase);
 
 	atomic_set(&cqp_request->refcount, 2);
-	nes_post_cqp_request(nesdev, cqp_request, NES_CQP_REQUEST_RING_DOORBELL);
+	nes_post_cqp_request(nesdev, cqp_request);
 
 	/* Wait for CQP */
 	if (wait_completion) {
@@ -2950,15 +2827,9 @@ int nes_hw_modify_qp(struct nes_device *nesdev, struct nes_qp *nesqp,
 					nesqp->hwqp.qp_id, cqp_request->major_code,
 					cqp_request->minor_code, next_iwarp_state);
 		}
-		if (atomic_dec_and_test(&cqp_request->refcount)) {
+
-			if (cqp_request->dynamic) {
+		nes_put_cqp_request(nesdev, cqp_request);
-				kfree(cqp_request);
+
-			} else {
-				spin_lock_irqsave(&nesdev->cqp.lock, flags);
-				list_add_tail(&cqp_request->list, &nesdev->cqp_avail_reqs);
-				spin_unlock_irqrestore(&nesdev->cqp.lock, flags);
-			}
-		}
 		if (!ret)
 			return -ETIME;
 		else if (major_code)

drivers/infiniband/ulp/ipoib/Kconfig

@@ -1,6 +1,7 @@
 config INFINIBAND_IPOIB
 	tristate "IP-over-InfiniBand"
 	depends on NETDEVICES && INET && (IPV6 || IPV6=n)
+	select INET_LRO
 	---help---
 	  Support for the IP-over-InfiniBand protocol (IPoIB). This
 	  transports IP packets over InfiniBand so you can use your IB

drivers/infiniband/ulp/ipoib/ipoib.h

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: ipoib.h 1358 2004-12-17 22:00:11Z roland $
  */
 
 #ifndef _IPOIB_H
@@ -52,9 +50,16 @@
 #include <rdma/ib_verbs.h>
 #include <rdma/ib_pack.h>
 #include <rdma/ib_sa.h>
+#include <linux/inet_lro.h>
 
 /* constants */
 
+enum ipoib_flush_level {
+	IPOIB_FLUSH_LIGHT,
+	IPOIB_FLUSH_NORMAL,
+	IPOIB_FLUSH_HEAVY
+};
+
 enum {
 	IPOIB_ENCAP_LEN		  = 4,
 
@@ -65,8 +70,8 @@ enum {
 	IPOIB_CM_BUF_SIZE	  = IPOIB_CM_MTU  + IPOIB_ENCAP_LEN,
 	IPOIB_CM_HEAD_SIZE	  = IPOIB_CM_BUF_SIZE % PAGE_SIZE,
 	IPOIB_CM_RX_SG		  = ALIGN(IPOIB_CM_BUF_SIZE, PAGE_SIZE) / PAGE_SIZE,
-	IPOIB_RX_RING_SIZE	  = 128,
+	IPOIB_RX_RING_SIZE	  = 256,
-	IPOIB_TX_RING_SIZE	  = 64,
+	IPOIB_TX_RING_SIZE	  = 128,
 	IPOIB_MAX_QUEUE_SIZE	  = 8192,
 	IPOIB_MIN_QUEUE_SIZE	  = 2,
 	IPOIB_CM_MAX_CONN_QP	  = 4096,
@@ -84,7 +89,6 @@ enum {
 	IPOIB_FLAG_SUBINTERFACE	  = 5,
 	IPOIB_MCAST_RUN		  = 6,
 	IPOIB_STOP_REAPER	  = 7,
-	IPOIB_MCAST_STARTED	  = 8,
 	IPOIB_FLAG_ADMIN_CM	  = 9,
 	IPOIB_FLAG_UMCAST	  = 10,
 	IPOIB_FLAG_CSUM		  = 11,
@@ -96,7 +100,11 @@ enum {
 	IPOIB_MCAST_FLAG_BUSY	  = 2,	/* joining or already joined */
 	IPOIB_MCAST_FLAG_ATTACHED = 3,
 
+	IPOIB_MAX_LRO_DESCRIPTORS = 8,
+	IPOIB_LRO_MAX_AGGR 	  = 64,
+
 	MAX_SEND_CQE		  = 16,
+	IPOIB_CM_COPYBREAK	  = 256,
 };
 
 #define	IPOIB_OP_RECV   (1ul << 31)
@@ -149,6 +157,11 @@ struct ipoib_tx_buf {
 	u64		mapping[MAX_SKB_FRAGS + 1];
 };
 
+struct ipoib_cm_tx_buf {
+	struct sk_buff *skb;
+	u64		mapping;
+};
+
 struct ib_cm_id;
 
 struct ipoib_cm_data {
@@ -207,7 +220,7 @@ struct ipoib_cm_tx {
 	struct net_device   *dev;
 	struct ipoib_neigh  *neigh;
 	struct ipoib_path   *path;
-	struct ipoib_tx_buf *tx_ring;
+	struct ipoib_cm_tx_buf *tx_ring;
 	unsigned	     tx_head;
 	unsigned	     tx_tail;
 	unsigned long	     flags;
@@ -249,6 +262,11 @@ struct ipoib_ethtool_st {
 	u16     max_coalesced_frames;
 };
 
+struct ipoib_lro {
+	struct net_lro_mgr lro_mgr;
+	struct net_lro_desc lro_desc[IPOIB_MAX_LRO_DESCRIPTORS];
+};
+
 /*
  * Device private locking: tx_lock protects members used in TX fast
  * path (and we use LLTX so upper layers don't do extra locking).
@@ -264,7 +282,6 @@ struct ipoib_dev_priv {
 
 	unsigned long flags;
 
-	struct mutex mcast_mutex;
 	struct mutex vlan_mutex;
 
 	struct rb_root  path_tree;
@@ -276,10 +293,11 @@ struct ipoib_dev_priv {
 
 	struct delayed_work pkey_poll_task;
 	struct delayed_work mcast_task;
-	struct work_struct flush_task;
+	struct work_struct flush_light;
+	struct work_struct flush_normal;
+	struct work_struct flush_heavy;
 	struct work_struct restart_task;
 	struct delayed_work ah_reap_task;
-	struct work_struct pkey_event_task;
 
 	struct ib_device *ca;
 	u8		  port;
@@ -335,6 +353,8 @@ struct ipoib_dev_priv {
 	int	hca_caps;
 	struct ipoib_ethtool_st ethtool;
 	struct timer_list poll_timer;
+
+	struct ipoib_lro lro;
 };
 
 struct ipoib_ah {
@@ -359,6 +379,7 @@ struct ipoib_path {
 
 	struct rb_node	      rb_node;
 	struct list_head      list;
+	int  		      valid;
 };
 
 struct ipoib_neigh {
@@ -423,11 +444,14 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb,
 		struct ipoib_ah *address, u32 qpn);
 void ipoib_reap_ah(struct work_struct *work);
 
+void ipoib_mark_paths_invalid(struct net_device *dev);
 void ipoib_flush_paths(struct net_device *dev);
 struct ipoib_dev_priv *ipoib_intf_alloc(const char *format);
 
 int ipoib_ib_dev_init(struct net_device *dev, struct ib_device *ca, int port);
-void ipoib_ib_dev_flush(struct work_struct *work);
+void ipoib_ib_dev_flush_light(struct work_struct *work);
+void ipoib_ib_dev_flush_normal(struct work_struct *work);
+void ipoib_ib_dev_flush_heavy(struct work_struct *work);
 void ipoib_pkey_event(struct work_struct *work);
 void ipoib_ib_dev_cleanup(struct net_device *dev);
 
@@ -466,9 +490,7 @@ void ipoib_path_iter_read(struct ipoib_path_iter *iter,
 #endif
 
 int ipoib_mcast_attach(struct net_device *dev, u16 mlid,
-		       union ib_gid *mgid);
+		       union ib_gid *mgid, int set_qkey);
-int ipoib_mcast_detach(struct net_device *dev, u16 mlid,
-		       union ib_gid *mgid);
 
 int ipoib_init_qp(struct net_device *dev);
 int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca);

drivers/infiniband/ulp/ipoib/ipoib_cm.c

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id$
  */
 
 #include <rdma/ib_cm.h>
@@ -113,18 +111,20 @@ static int ipoib_cm_post_receive_srq(struct net_device *dev, int id)
 }
 
 static int ipoib_cm_post_receive_nonsrq(struct net_device *dev,
-					struct ipoib_cm_rx *rx, int id)
+					struct ipoib_cm_rx *rx,
+					struct ib_recv_wr *wr,
+					struct ib_sge *sge, int id)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 	struct ib_recv_wr *bad_wr;
 	int i, ret;
 
-	priv->cm.rx_wr.wr_id = id | IPOIB_OP_CM | IPOIB_OP_RECV;
+	wr->wr_id = id | IPOIB_OP_CM | IPOIB_OP_RECV;
 
 	for (i = 0; i < IPOIB_CM_RX_SG; ++i)
-		priv->cm.rx_sge[i].addr = rx->rx_ring[id].mapping[i];
+		sge[i].addr = rx->rx_ring[id].mapping[i];
 
-	ret = ib_post_recv(rx->qp, &priv->cm.rx_wr, &bad_wr);
+	ret = ib_post_recv(rx->qp, wr, &bad_wr);
 	if (unlikely(ret)) {
 		ipoib_warn(priv, "post recv failed for buf %d (%d)\n", id, ret);
 		ipoib_cm_dma_unmap_rx(priv, IPOIB_CM_RX_SG - 1,
@@ -322,10 +322,33 @@ static int ipoib_cm_modify_rx_qp(struct net_device *dev,
 	return 0;
 }
 
+static void ipoib_cm_init_rx_wr(struct net_device *dev,
+				struct ib_recv_wr *wr,
+				struct ib_sge *sge)
+{
+	struct ipoib_dev_priv *priv = netdev_priv(dev);
+	int i;
+
+	for (i = 0; i < priv->cm.num_frags; ++i)
+		sge[i].lkey = priv->mr->lkey;
+
+	sge[0].length = IPOIB_CM_HEAD_SIZE;
+	for (i = 1; i < priv->cm.num_frags; ++i)
+		sge[i].length = PAGE_SIZE;
+
+	wr->next    = NULL;
+	wr->sg_list = priv->cm.rx_sge;
+	wr->num_sge = priv->cm.num_frags;
+}
+
 static int ipoib_cm_nonsrq_init_rx(struct net_device *dev, struct ib_cm_id *cm_id,
 				   struct ipoib_cm_rx *rx)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
+	struct {
+		struct ib_recv_wr wr;
+		struct ib_sge sge[IPOIB_CM_RX_SG];
+	} *t;
 	int ret;
 	int i;
 
@@ -333,6 +356,14 @@ static int ipoib_cm_nonsrq_init_rx(struct net_device *dev, struct ib_cm_id *cm_i
 	if (!rx->rx_ring)
 		return -ENOMEM;
 
+	t = kmalloc(sizeof *t, GFP_KERNEL);
+	if (!t) {
+		ret = -ENOMEM;
+		goto err_free;
+	}
+
+	ipoib_cm_init_rx_wr(dev, &t->wr, t->sge);
+
 	spin_lock_irq(&priv->lock);
 
 	if (priv->cm.nonsrq_conn_qp >= ipoib_max_conn_qp) {
@@ -351,8 +382,8 @@ static int ipoib_cm_nonsrq_init_rx(struct net_device *dev, struct ib_cm_id *cm_i
 			ipoib_warn(priv, "failed to allocate receive buffer %d\n", i);
 				ret = -ENOMEM;
 				goto err_count;
-			}
+		}
-		ret = ipoib_cm_post_receive_nonsrq(dev, rx, i);
+		ret = ipoib_cm_post_receive_nonsrq(dev, rx, &t->wr, t->sge, i);
 		if (ret) {
 			ipoib_warn(priv, "ipoib_cm_post_receive_nonsrq "
 				   "failed for buf %d\n", i);
@@ -363,6 +394,8 @@ static int ipoib_cm_nonsrq_init_rx(struct net_device *dev, struct ib_cm_id *cm_i
 
 	rx->recv_count = ipoib_recvq_size;
 
+	kfree(t);
+
 	return 0;
 
 err_count:
@@ -371,6 +404,7 @@ err_count:
 	spin_unlock_irq(&priv->lock);
 
 err_free:
+	kfree(t);
 	ipoib_cm_free_rx_ring(dev, rx->rx_ring);
 
 	return ret;
@@ -525,6 +559,7 @@ void ipoib_cm_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
 	u64 mapping[IPOIB_CM_RX_SG];
 	int frags;
 	int has_srq;
+	struct sk_buff *small_skb;
 
 	ipoib_dbg_data(priv, "cm recv completion: id %d, status: %d\n",
 		       wr_id, wc->status);
@@ -579,6 +614,23 @@ void ipoib_cm_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
 		}
 	}
 
+	if (wc->byte_len < IPOIB_CM_COPYBREAK) {
+		int dlen = wc->byte_len;
+
+		small_skb = dev_alloc_skb(dlen + 12);
+		if (small_skb) {
+			skb_reserve(small_skb, 12);
+			ib_dma_sync_single_for_cpu(priv->ca, rx_ring[wr_id].mapping[0],
+						   dlen, DMA_FROM_DEVICE);
+			skb_copy_from_linear_data(skb, small_skb->data, dlen);
+			ib_dma_sync_single_for_device(priv->ca, rx_ring[wr_id].mapping[0],
+						      dlen, DMA_FROM_DEVICE);
+			skb_put(small_skb, dlen);
+			skb = small_skb;
+			goto copied;
+		}
+	}
+
 	frags = PAGE_ALIGN(wc->byte_len - min(wc->byte_len,
 					      (unsigned)IPOIB_CM_HEAD_SIZE)) / PAGE_SIZE;
 
@@ -601,6 +653,7 @@ void ipoib_cm_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
 
 	skb_put_frags(skb, IPOIB_CM_HEAD_SIZE, wc->byte_len, newskb);
 
+copied:
 	skb->protocol = ((struct ipoib_header *) skb->data)->proto;
 	skb_reset_mac_header(skb);
 	skb_pull(skb, IPOIB_ENCAP_LEN);
@@ -620,7 +673,10 @@ repost:
 			ipoib_warn(priv, "ipoib_cm_post_receive_srq failed "
 				   "for buf %d\n", wr_id);
 	} else {
-		if (unlikely(ipoib_cm_post_receive_nonsrq(dev, p, wr_id))) {
+		if (unlikely(ipoib_cm_post_receive_nonsrq(dev, p,
+							  &priv->cm.rx_wr,
+							  priv->cm.rx_sge,
+							  wr_id))) {
 			--p->recv_count;
 			ipoib_warn(priv, "ipoib_cm_post_receive_nonsrq failed "
 				   "for buf %d\n", wr_id);
@@ -647,7 +703,7 @@ static inline int post_send(struct ipoib_dev_priv *priv,
 void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_tx *tx)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
-	struct ipoib_tx_buf *tx_req;
+	struct ipoib_cm_tx_buf *tx_req;
 	u64 addr;
 
 	if (unlikely(skb->len > tx->mtu)) {
@@ -678,7 +734,7 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_
 		return;
 	}
 
-	tx_req->mapping[0] = addr;
+	tx_req->mapping = addr;
 
 	if (unlikely(post_send(priv, tx, tx->tx_head & (ipoib_sendq_size - 1),
 			       addr, skb->len))) {
@@ -703,7 +759,7 @@ void ipoib_cm_handle_tx_wc(struct net_device *dev, struct ib_wc *wc)
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 	struct ipoib_cm_tx *tx = wc->qp->qp_context;
 	unsigned int wr_id = wc->wr_id & ~IPOIB_OP_CM;
-	struct ipoib_tx_buf *tx_req;
+	struct ipoib_cm_tx_buf *tx_req;
 	unsigned long flags;
 
 	ipoib_dbg_data(priv, "cm send completion: id %d, status: %d\n",
@@ -717,7 +773,7 @@ void ipoib_cm_handle_tx_wc(struct net_device *dev, struct ib_wc *wc)
 
 	tx_req = &tx->tx_ring[wr_id];
 
-	ib_dma_unmap_single(priv->ca, tx_req->mapping[0], tx_req->skb->len, DMA_TO_DEVICE);
+	ib_dma_unmap_single(priv->ca, tx_req->mapping, tx_req->skb->len, DMA_TO_DEVICE);
 
 	/* FIXME: is this right? Shouldn't we only increment on success? */
 	++dev->stats.tx_packets;
@@ -1087,7 +1143,7 @@ err_tx:
 static void ipoib_cm_tx_destroy(struct ipoib_cm_tx *p)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(p->dev);
-	struct ipoib_tx_buf *tx_req;
+	struct ipoib_cm_tx_buf *tx_req;
 	unsigned long flags;
 	unsigned long begin;
 
@@ -1115,7 +1171,7 @@ timeout:
 
 	while ((int) p->tx_tail - (int) p->tx_head < 0) {
 		tx_req = &p->tx_ring[p->tx_tail & (ipoib_sendq_size - 1)];
-		ib_dma_unmap_single(priv->ca, tx_req->mapping[0], tx_req->skb->len,
+		ib_dma_unmap_single(priv->ca, tx_req->mapping, tx_req->skb->len,
 				    DMA_TO_DEVICE);
 		dev_kfree_skb_any(tx_req->skb);
 		++p->tx_tail;
@@ -1384,7 +1440,9 @@ static ssize_t set_mode(struct device *d, struct device_attribute *attr,
 		ipoib_warn(priv, "enabling connected mode "
 			   "will cause multicast packet drops\n");
 
+		rtnl_lock();
 		dev->features &= ~(NETIF_F_IP_CSUM | NETIF_F_SG | NETIF_F_TSO);
+		rtnl_unlock();
 		priv->tx_wr.send_flags &= ~IB_SEND_IP_CSUM;
 
 		ipoib_flush_paths(dev);
@@ -1393,14 +1451,16 @@ static ssize_t set_mode(struct device *d, struct device_attribute *attr,
 
 	if (!strcmp(buf, "datagram\n")) {
 		clear_bit(IPOIB_FLAG_ADMIN_CM, &priv->flags);
-		dev->mtu = min(priv->mcast_mtu, dev->mtu);
-		ipoib_flush_paths(dev);
 
+		rtnl_lock();
 		if (test_bit(IPOIB_FLAG_CSUM, &priv->flags)) {
 			dev->features |= NETIF_F_IP_CSUM | NETIF_F_SG;
 			if (priv->hca_caps & IB_DEVICE_UD_TSO)
 				dev->features |= NETIF_F_TSO;
 		}
+		dev_set_mtu(dev, min(priv->mcast_mtu, dev->mtu));
+		rtnl_unlock();
+		ipoib_flush_paths(dev);
 
 		return count;
 	}
@@ -1485,15 +1545,7 @@ int ipoib_cm_dev_init(struct net_device *dev)
 		priv->cm.num_frags  = IPOIB_CM_RX_SG;
 	}
 
-	for (i = 0; i < priv->cm.num_frags; ++i)
+	ipoib_cm_init_rx_wr(dev, &priv->cm.rx_wr, priv->cm.rx_sge);
-		priv->cm.rx_sge[i].lkey	= priv->mr->lkey;
-
-	priv->cm.rx_sge[0].length = IPOIB_CM_HEAD_SIZE;
-	for (i = 1; i < priv->cm.num_frags; ++i)
-		priv->cm.rx_sge[i].length = PAGE_SIZE;
-	priv->cm.rx_wr.next = NULL;
-	priv->cm.rx_wr.sg_list = priv->cm.rx_sge;
-	priv->cm.rx_wr.num_sge = priv->cm.num_frags;
 
 	if (ipoib_cm_has_srq(dev)) {
 		for (i = 0; i < ipoib_recvq_size; ++i) {

drivers/infiniband/ulp/ipoib/ipoib_ethtool.c

@@ -86,11 +86,57 @@ static int ipoib_set_coalesce(struct net_device *dev,
 	return 0;
 }
 
+static const char ipoib_stats_keys[][ETH_GSTRING_LEN] = {
+	"LRO aggregated", "LRO flushed",
+	"LRO avg aggr", "LRO no desc"
+};
+
+static void ipoib_get_strings(struct net_device *netdev, u32 stringset, u8 *data)
+{
+	switch (stringset) {
+	case ETH_SS_STATS:
+		memcpy(data, *ipoib_stats_keys,	sizeof(ipoib_stats_keys));
+		break;
+	}
+}
+
+static int ipoib_get_sset_count(struct net_device *dev, int sset)
+{
+	switch (sset) {
+	case ETH_SS_STATS:
+		return ARRAY_SIZE(ipoib_stats_keys);
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
+static void ipoib_get_ethtool_stats(struct net_device *dev,
+				struct ethtool_stats *stats, uint64_t *data)
+{
+	struct ipoib_dev_priv *priv = netdev_priv(dev);
+	int index = 0;
+
+	/* Get LRO statistics */
+	data[index++] = priv->lro.lro_mgr.stats.aggregated;
+	data[index++] = priv->lro.lro_mgr.stats.flushed;
+	if (priv->lro.lro_mgr.stats.flushed)
+		data[index++] = priv->lro.lro_mgr.stats.aggregated /
+				priv->lro.lro_mgr.stats.flushed;
+	else
+		data[index++] = 0;
+	data[index++] = priv->lro.lro_mgr.stats.no_desc;
+}
+
 static const struct ethtool_ops ipoib_ethtool_ops = {
 	.get_drvinfo		= ipoib_get_drvinfo,
 	.get_tso		= ethtool_op_get_tso,
 	.get_coalesce		= ipoib_get_coalesce,
 	.set_coalesce		= ipoib_set_coalesce,
+	.get_flags		= ethtool_op_get_flags,
+	.set_flags		= ethtool_op_set_flags,
+	.get_strings		= ipoib_get_strings,
+	.get_sset_count		= ipoib_get_sset_count,
+	.get_ethtool_stats	= ipoib_get_ethtool_stats,
 };
 
 void ipoib_set_ethtool_ops(struct net_device *dev)

drivers/infiniband/ulp/ipoib/ipoib_fs.c

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: ipoib_fs.c 1389 2004-12-27 22:56:47Z roland $
  */
 
 #include <linux/err.h>

drivers/infiniband/ulp/ipoib/ipoib_ib.c

@@ -31,8 +31,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: ipoib_ib.c 1386 2004-12-27 16:23:17Z roland $
  */
 
 #include <linux/delay.h>
@@ -290,7 +288,10 @@ static void ipoib_ib_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
 	if (test_bit(IPOIB_FLAG_CSUM, &priv->flags) && likely(wc->csum_ok))
 		skb->ip_summed = CHECKSUM_UNNECESSARY;
 
-	netif_receive_skb(skb);
+	if (dev->features & NETIF_F_LRO)
+		lro_receive_skb(&priv->lro.lro_mgr, skb, NULL);
+	else
+		netif_receive_skb(skb);
 
 repost:
 	if (unlikely(ipoib_ib_post_receive(dev, wr_id)))
@@ -442,6 +443,9 @@ poll_more:
 	}
 
 	if (done < budget) {
+		if (dev->features & NETIF_F_LRO)
+			lro_flush_all(&priv->lro.lro_mgr);
+
 		netif_rx_complete(dev, napi);
 		if (unlikely(ib_req_notify_cq(priv->recv_cq,
 					      IB_CQ_NEXT_COMP |
@@ -898,7 +902,8 @@ int ipoib_ib_dev_init(struct net_device *dev, struct ib_device *ca, int port)
 	return 0;
 }
 
-static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv, int pkey_event)
+static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv,
+				enum ipoib_flush_level level)
 {
 	struct ipoib_dev_priv *cpriv;
 	struct net_device *dev = priv->dev;
@@ -911,7 +916,7 @@ static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv, int pkey_event)
 	 * the parent is down.
 	 */
 	list_for_each_entry(cpriv, &priv->child_intfs, list)
-		__ipoib_ib_dev_flush(cpriv, pkey_event);
+		__ipoib_ib_dev_flush(cpriv, level);
 
 	mutex_unlock(&priv->vlan_mutex);
 
@@ -925,7 +930,7 @@ static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv, int pkey_event)
 		return;
 	}
 
-	if (pkey_event) {
+	if (level == IPOIB_FLUSH_HEAVY) {
 		if (ib_find_pkey(priv->ca, priv->port, priv->pkey, &new_index)) {
 			clear_bit(IPOIB_PKEY_ASSIGNED, &priv->flags);
 			ipoib_ib_dev_down(dev, 0);
@@ -943,11 +948,15 @@ static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv, int pkey_event)
 		priv->pkey_index = new_index;
 	}
 
-	ipoib_dbg(priv, "flushing\n");
+	if (level == IPOIB_FLUSH_LIGHT) {
+		ipoib_mark_paths_invalid(dev);
+		ipoib_mcast_dev_flush(dev);
+	}
 
-	ipoib_ib_dev_down(dev, 0);
+	if (level >= IPOIB_FLUSH_NORMAL)
+		ipoib_ib_dev_down(dev, 0);
 
-	if (pkey_event) {
+	if (level == IPOIB_FLUSH_HEAVY) {
 		ipoib_ib_dev_stop(dev, 0);
 		ipoib_ib_dev_open(dev);
 	}
@@ -957,27 +966,34 @@ static void __ipoib_ib_dev_flush(struct ipoib_dev_priv *priv, int pkey_event)
 	 * we get here, don't bring it back up if it's not configured up
 	 */
 	if (test_bit(IPOIB_FLAG_ADMIN_UP, &priv->flags)) {
-		ipoib_ib_dev_up(dev);
+		if (level >= IPOIB_FLUSH_NORMAL)
+			ipoib_ib_dev_up(dev);
 		ipoib_mcast_restart_task(&priv->restart_task);
 	}
 }
 
-void ipoib_ib_dev_flush(struct work_struct *work)
+void ipoib_ib_dev_flush_light(struct work_struct *work)
 {
 	struct ipoib_dev_priv *priv =
-		container_of(work, struct ipoib_dev_priv, flush_task);
+		container_of(work, struct ipoib_dev_priv, flush_light);
 
-	ipoib_dbg(priv, "Flushing %s\n", priv->dev->name);
+	__ipoib_ib_dev_flush(priv, IPOIB_FLUSH_LIGHT);
-	__ipoib_ib_dev_flush(priv, 0);
 }
 
-void ipoib_pkey_event(struct work_struct *work)
+void ipoib_ib_dev_flush_normal(struct work_struct *work)
 {
 	struct ipoib_dev_priv *priv =
-		container_of(work, struct ipoib_dev_priv, pkey_event_task);
+		container_of(work, struct ipoib_dev_priv, flush_normal);
 
-	ipoib_dbg(priv, "Flushing %s and restarting its QP\n", priv->dev->name);
+	__ipoib_ib_dev_flush(priv, IPOIB_FLUSH_NORMAL);
-	__ipoib_ib_dev_flush(priv, 1);
+}
+
+void ipoib_ib_dev_flush_heavy(struct work_struct *work)
+{
+	struct ipoib_dev_priv *priv =
+		container_of(work, struct ipoib_dev_priv, flush_heavy);
+
+	__ipoib_ib_dev_flush(priv, IPOIB_FLUSH_HEAVY);
 }
 
 void ipoib_ib_dev_cleanup(struct net_device *dev)

drivers/infiniband/ulp/ipoib/ipoib_main.c

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: ipoib_main.c 1377 2004-12-23 19:57:12Z roland $
  */
 
 #include "ipoib.h"
@@ -62,6 +60,15 @@ MODULE_PARM_DESC(send_queue_size, "Number of descriptors in send queue");
 module_param_named(recv_queue_size, ipoib_recvq_size, int, 0444);
 MODULE_PARM_DESC(recv_queue_size, "Number of descriptors in receive queue");
 
+static int lro;
+module_param(lro, bool, 0444);
+MODULE_PARM_DESC(lro,  "Enable LRO (Large Receive Offload)");
+
+static int lro_max_aggr = IPOIB_LRO_MAX_AGGR;
+module_param(lro_max_aggr, int, 0644);
+MODULE_PARM_DESC(lro_max_aggr, "LRO: Max packets to be aggregated "
+		"(default = 64)");
+
 #ifdef CONFIG_INFINIBAND_IPOIB_DEBUG
 int ipoib_debug_level;
 
@@ -350,6 +357,23 @@ void ipoib_path_iter_read(struct ipoib_path_iter *iter,
 
 #endif /* CONFIG_INFINIBAND_IPOIB_DEBUG */
 
+void ipoib_mark_paths_invalid(struct net_device *dev)
+{
+	struct ipoib_dev_priv *priv = netdev_priv(dev);
+	struct ipoib_path *path, *tp;
+
+	spin_lock_irq(&priv->lock);
+
+	list_for_each_entry_safe(path, tp, &priv->path_list, list) {
+		ipoib_dbg(priv, "mark path LID 0x%04x GID " IPOIB_GID_FMT " invalid\n",
+			be16_to_cpu(path->pathrec.dlid),
+			IPOIB_GID_ARG(path->pathrec.dgid));
+		path->valid =  0;
+	}
+
+	spin_unlock_irq(&priv->lock);
+}
+
 void ipoib_flush_paths(struct net_device *dev)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
@@ -386,6 +410,7 @@ static void path_rec_completion(int status,
 	struct net_device *dev = path->dev;
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 	struct ipoib_ah *ah = NULL;
+	struct ipoib_ah *old_ah;
 	struct ipoib_neigh *neigh, *tn;
 	struct sk_buff_head skqueue;
 	struct sk_buff *skb;
@@ -409,6 +434,7 @@ static void path_rec_completion(int status,
 
 	spin_lock_irqsave(&priv->lock, flags);
 
+	old_ah   = path->ah;
 	path->ah = ah;
 
 	if (ah) {
@@ -421,6 +447,17 @@ static void path_rec_completion(int status,
 			__skb_queue_tail(&skqueue, skb);
 
 		list_for_each_entry_safe(neigh, tn, &path->neigh_list, list) {
+			if (neigh->ah) {
+				WARN_ON(neigh->ah != old_ah);
+				/*
+				 * Dropping the ah reference inside
+				 * priv->lock is safe here, because we
+				 * will hold one more reference from
+				 * the original value of path->ah (ie
+				 * old_ah).
+				 */
+				ipoib_put_ah(neigh->ah);
+			}
 			kref_get(&path->ah->ref);
 			neigh->ah = path->ah;
 			memcpy(&neigh->dgid.raw, &path->pathrec.dgid.raw,
@@ -443,6 +480,7 @@ static void path_rec_completion(int status,
 			while ((skb = __skb_dequeue(&neigh->queue)))
 				__skb_queue_tail(&skqueue, skb);
 		}
+		path->valid = 1;
 	}
 
 	path->query = NULL;
@@ -450,6 +488,9 @@ static void path_rec_completion(int status,
 
 	spin_unlock_irqrestore(&priv->lock, flags);
 
+	if (old_ah)
+		ipoib_put_ah(old_ah);
+
 	while ((skb = __skb_dequeue(&skqueue))) {
 		skb->dev = dev;
 		if (dev_queue_xmit(skb))
@@ -623,8 +664,9 @@ static void unicast_arp_send(struct sk_buff *skb, struct net_device *dev,
 	spin_lock(&priv->lock);
 
 	path = __path_find(dev, phdr->hwaddr + 4);
-	if (!path) {
+	if (!path || !path->valid) {
-		path = path_rec_create(dev, phdr->hwaddr + 4);
+		if (!path)
+			path = path_rec_create(dev, phdr->hwaddr + 4);
 		if (path) {
 			/* put pseudoheader back on for next time */
 			skb_push(skb, sizeof *phdr);
@@ -938,6 +980,54 @@ static const struct header_ops ipoib_header_ops = {
 	.create	= ipoib_hard_header,
 };
 
+static int get_skb_hdr(struct sk_buff *skb, void **iphdr,
+		       void **tcph, u64 *hdr_flags, void *priv)
+{
+	unsigned int ip_len;
+	struct iphdr *iph;
+
+	if (unlikely(skb->protocol != htons(ETH_P_IP)))
+		return -1;
+
+	/*
+	 * In the future we may add an else clause that verifies the
+	 * checksum and allows devices which do not calculate checksum
+	 * to use LRO.
+	 */
+	if (unlikely(skb->ip_summed != CHECKSUM_UNNECESSARY))
+		return -1;
+
+	/* Check for non-TCP packet */
+	skb_reset_network_header(skb);
+	iph = ip_hdr(skb);
+	if (iph->protocol != IPPROTO_TCP)
+		return -1;
+
+	ip_len = ip_hdrlen(skb);
+	skb_set_transport_header(skb, ip_len);
+	*tcph = tcp_hdr(skb);
+
+	/* check if IP header and TCP header are complete */
+	if (ntohs(iph->tot_len) < ip_len + tcp_hdrlen(skb))
+		return -1;
+
+	*hdr_flags = LRO_IPV4 | LRO_TCP;
+	*iphdr = iph;
+
+	return 0;
+}
+
+static void ipoib_lro_setup(struct ipoib_dev_priv *priv)
+{
+	priv->lro.lro_mgr.max_aggr	 = lro_max_aggr;
+	priv->lro.lro_mgr.max_desc	 = IPOIB_MAX_LRO_DESCRIPTORS;
+	priv->lro.lro_mgr.lro_arr	 = priv->lro.lro_desc;
+	priv->lro.lro_mgr.get_skb_header = get_skb_hdr;
+	priv->lro.lro_mgr.features	 = LRO_F_NAPI;
+	priv->lro.lro_mgr.dev		 = priv->dev;
+	priv->lro.lro_mgr.ip_summed_aggr = CHECKSUM_UNNECESSARY;
+}
+
 static void ipoib_setup(struct net_device *dev)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
@@ -977,10 +1067,11 @@ static void ipoib_setup(struct net_device *dev)
 
 	priv->dev = dev;
 
+	ipoib_lro_setup(priv);
+
 	spin_lock_init(&priv->lock);
 	spin_lock_init(&priv->tx_lock);
 
-	mutex_init(&priv->mcast_mutex);
 	mutex_init(&priv->vlan_mutex);
 
 	INIT_LIST_HEAD(&priv->path_list);
@@ -989,9 +1080,10 @@ static void ipoib_setup(struct net_device *dev)
 	INIT_LIST_HEAD(&priv->multicast_list);
 
 	INIT_DELAYED_WORK(&priv->pkey_poll_task, ipoib_pkey_poll);
-	INIT_WORK(&priv->pkey_event_task, ipoib_pkey_event);
 	INIT_DELAYED_WORK(&priv->mcast_task,   ipoib_mcast_join_task);
-	INIT_WORK(&priv->flush_task,   ipoib_ib_dev_flush);
+	INIT_WORK(&priv->flush_light,   ipoib_ib_dev_flush_light);
+	INIT_WORK(&priv->flush_normal,   ipoib_ib_dev_flush_normal);
+	INIT_WORK(&priv->flush_heavy,   ipoib_ib_dev_flush_heavy);
 	INIT_WORK(&priv->restart_task, ipoib_mcast_restart_task);
 	INIT_DELAYED_WORK(&priv->ah_reap_task, ipoib_reap_ah);
 }
@@ -1154,6 +1246,9 @@ static struct net_device *ipoib_add_port(const char *format,
 		priv->dev->features |= NETIF_F_SG | NETIF_F_IP_CSUM;
 	}
 
+	if (lro)
+		priv->dev->features |= NETIF_F_LRO;
+
 	/*
 	 * Set the full membership bit, so that we join the right
 	 * broadcast group, etc.
@@ -1304,6 +1399,12 @@ static int __init ipoib_init_module(void)
 	ipoib_max_conn_qp = min(ipoib_max_conn_qp, IPOIB_CM_MAX_CONN_QP);
 #endif
 
+	/*
+	 * When copying small received packets, we only copy from the
+	 * linear data part of the SKB, so we rely on this condition.
+	 */
+	BUILD_BUG_ON(IPOIB_CM_COPYBREAK > IPOIB_CM_HEAD_SIZE);
+
 	ret = ipoib_register_debugfs();
 	if (ret)
 		return ret;

drivers/infiniband/ulp/ipoib/ipoib_multicast.c

@@ -30,8 +30,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: ipoib_multicast.c 1362 2004-12-18 15:56:29Z roland $
  */
 
 #include <linux/skbuff.h>
@@ -188,6 +186,7 @@ static int ipoib_mcast_join_finish(struct ipoib_mcast *mcast,
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 	struct ipoib_ah *ah;
 	int ret;
+	int set_qkey = 0;
 
 	mcast->mcmember = *mcmember;
 
@@ -202,6 +201,7 @@ static int ipoib_mcast_join_finish(struct ipoib_mcast *mcast,
 		priv->qkey = be32_to_cpu(priv->broadcast->mcmember.qkey);
 		spin_unlock_irq(&priv->lock);
 		priv->tx_wr.wr.ud.remote_qkey = priv->qkey;
+		set_qkey = 1;
 	}
 
 	if (!test_bit(IPOIB_MCAST_FLAG_SENDONLY, &mcast->flags)) {
@@ -214,7 +214,7 @@ static int ipoib_mcast_join_finish(struct ipoib_mcast *mcast,
 		}
 
 		ret = ipoib_mcast_attach(dev, be16_to_cpu(mcast->mcmember.mlid),
-					 &mcast->mcmember.mgid);
+					 &mcast->mcmember.mgid, set_qkey);
 		if (ret < 0) {
 			ipoib_warn(priv, "couldn't attach QP to multicast group "
 				   IPOIB_GID_FMT "\n",
@@ -575,8 +575,11 @@ void ipoib_mcast_join_task(struct work_struct *work)
 
 	priv->mcast_mtu = IPOIB_UD_MTU(ib_mtu_enum_to_int(priv->broadcast->mcmember.mtu));
 
-	if (!ipoib_cm_admin_enabled(dev))
+	if (!ipoib_cm_admin_enabled(dev)) {
-		dev->mtu = min(priv->mcast_mtu, priv->admin_mtu);
+		rtnl_lock();
+		dev_set_mtu(dev, min(priv->mcast_mtu, priv->admin_mtu));
+		rtnl_unlock();
+	}
 
 	ipoib_dbg_mcast(priv, "successfully joined all multicast groups\n");
 
@@ -594,10 +597,6 @@ int ipoib_mcast_start_thread(struct net_device *dev)
 		queue_delayed_work(ipoib_workqueue, &priv->mcast_task, 0);
 	mutex_unlock(&mcast_mutex);
 
-	spin_lock_irq(&priv->lock);
-	set_bit(IPOIB_MCAST_STARTED, &priv->flags);
-	spin_unlock_irq(&priv->lock);
-
 	return 0;
 }
 
@@ -607,10 +606,6 @@ int ipoib_mcast_stop_thread(struct net_device *dev, int flush)
 
 	ipoib_dbg_mcast(priv, "stopping multicast thread\n");
 
-	spin_lock_irq(&priv->lock);
-	clear_bit(IPOIB_MCAST_STARTED, &priv->flags);
-	spin_unlock_irq(&priv->lock);
-
 	mutex_lock(&mcast_mutex);
 	clear_bit(IPOIB_MCAST_RUN, &priv->flags);
 	cancel_delayed_work(&priv->mcast_task);
@@ -635,10 +630,10 @@ static int ipoib_mcast_leave(struct net_device *dev, struct ipoib_mcast *mcast)
 				IPOIB_GID_ARG(mcast->mcmember.mgid));
 
 		/* Remove ourselves from the multicast group */
-		ret = ipoib_mcast_detach(dev, be16_to_cpu(mcast->mcmember.mlid),
+		ret = ib_detach_mcast(priv->qp, &mcast->mcmember.mgid,
-					 &mcast->mcmember.mgid);
+				      be16_to_cpu(mcast->mcmember.mlid));
 		if (ret)
-			ipoib_warn(priv, "ipoib_mcast_detach failed (result = %d)\n", ret);
+			ipoib_warn(priv, "ib_detach_mcast failed (result = %d)\n", ret);
 	}
 
 	return 0;

drivers/infiniband/ulp/ipoib/ipoib_verbs.c

@@ -29,24 +29,17 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: ipoib_verbs.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include "ipoib.h"
 
-int ipoib_mcast_attach(struct net_device *dev, u16 mlid, union ib_gid *mgid)
+int ipoib_mcast_attach(struct net_device *dev, u16 mlid, union ib_gid *mgid, int set_qkey)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
-	struct ib_qp_attr *qp_attr;
+	struct ib_qp_attr *qp_attr = NULL;
 	int ret;
 	u16 pkey_index;
 
-	ret = -ENOMEM;
-	qp_attr = kmalloc(sizeof *qp_attr, GFP_KERNEL);
-	if (!qp_attr)
-		goto out;
-
 	if (ib_find_pkey(priv->ca, priv->port, priv->pkey, &pkey_index)) {
 		clear_bit(IPOIB_PKEY_ASSIGNED, &priv->flags);
 		ret = -ENXIO;
@@ -54,18 +47,23 @@ int ipoib_mcast_attach(struct net_device *dev, u16 mlid, union ib_gid *mgid)
 	}
 	set_bit(IPOIB_PKEY_ASSIGNED, &priv->flags);
 
-	/* set correct QKey for QP */
+	if (set_qkey) {
-	qp_attr->qkey = priv->qkey;
+		ret = -ENOMEM;
-	ret = ib_modify_qp(priv->qp, qp_attr, IB_QP_QKEY);
+		qp_attr = kmalloc(sizeof *qp_attr, GFP_KERNEL);
-	if (ret) {
+		if (!qp_attr)
-		ipoib_warn(priv, "failed to modify QP, ret = %d\n", ret);
+			goto out;
-		goto out;
+
+		/* set correct QKey for QP */
+		qp_attr->qkey = priv->qkey;
+		ret = ib_modify_qp(priv->qp, qp_attr, IB_QP_QKEY);
+		if (ret) {
+			ipoib_warn(priv, "failed to modify QP, ret = %d\n", ret);
+			goto out;
+		}
 	}
 
 	/* attach QP to multicast group */
-	mutex_lock(&priv->mcast_mutex);
 	ret = ib_attach_mcast(priv->qp, mgid, mlid);
-	mutex_unlock(&priv->mcast_mutex);
 	if (ret)
 		ipoib_warn(priv, "failed to attach to multicast group, ret = %d\n", ret);
 
@@ -74,20 +72,6 @@ out:
 	return ret;
 }
 
-int ipoib_mcast_detach(struct net_device *dev, u16 mlid, union ib_gid *mgid)
-{
-	struct ipoib_dev_priv *priv = netdev_priv(dev);
-	int ret;
-
-	mutex_lock(&priv->mcast_mutex);
-	ret = ib_detach_mcast(priv->qp, mgid, mlid);
-	mutex_unlock(&priv->mcast_mutex);
-	if (ret)
-		ipoib_warn(priv, "ib_detach_mcast failed (result = %d)\n", ret);
-
-	return ret;
-}
-
 int ipoib_init_qp(struct net_device *dev)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
@@ -201,7 +185,10 @@ int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca)
 	init_attr.recv_cq = priv->recv_cq;
 
 	if (priv->hca_caps & IB_DEVICE_UD_TSO)
-		init_attr.create_flags = IB_QP_CREATE_IPOIB_UD_LSO;
+		init_attr.create_flags |= IB_QP_CREATE_IPOIB_UD_LSO;
+
+	if (priv->hca_caps & IB_DEVICE_BLOCK_MULTICAST_LOOPBACK)
+		init_attr.create_flags |= IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK;
 
 	if (dev->features & NETIF_F_SG)
 		init_attr.cap.max_send_sge = MAX_SKB_FRAGS + 1;
@@ -289,15 +276,17 @@ void ipoib_event(struct ib_event_handler *handler,
 	if (record->element.port_num != priv->port)
 		return;
 
-	if (record->event == IB_EVENT_PORT_ERR    ||
+	ipoib_dbg(priv, "Event %d on device %s port %d\n", record->event,
-	    record->event == IB_EVENT_PORT_ACTIVE ||
+		  record->device->name, record->element.port_num);
-	    record->event == IB_EVENT_LID_CHANGE  ||
+
-	    record->event == IB_EVENT_SM_CHANGE   ||
+	if (record->event == IB_EVENT_SM_CHANGE ||
 	    record->event == IB_EVENT_CLIENT_REREGISTER) {
-		ipoib_dbg(priv, "Port state change event\n");
+		queue_work(ipoib_workqueue, &priv->flush_light);
-		queue_work(ipoib_workqueue, &priv->flush_task);
+	} else if (record->event == IB_EVENT_PORT_ERR ||
+		   record->event == IB_EVENT_PORT_ACTIVE ||
+		   record->event == IB_EVENT_LID_CHANGE) {
+		queue_work(ipoib_workqueue, &priv->flush_normal);
 	} else if (record->event == IB_EVENT_PKEY_CHANGE) {
-		ipoib_dbg(priv, "P_Key change event on port:%d\n", priv->port);
+		queue_work(ipoib_workqueue, &priv->flush_heavy);
-		queue_work(ipoib_workqueue, &priv->pkey_event_task);
 	}
 }

drivers/infiniband/ulp/ipoib/ipoib_vlan.c

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: ipoib_vlan.c 1349 2004-12-16 21:09:43Z roland $
  */
 
 #include <linux/module.h>

drivers/infiniband/ulp/iser/iscsi_iser.c

@@ -42,9 +42,6 @@
  *	Zhenyu Wang
  * Modified by:
  *      Erez Zilber
- *
- *
- * $Id: iscsi_iser.c 6965 2006-05-07 11:36:20Z ogerlitz $
  */
 
 #include <linux/types.h>

drivers/infiniband/ulp/iser/iscsi_iser.h

@@ -36,8 +36,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: iscsi_iser.h 7051 2006-05-10 12:29:11Z ogerlitz $
  */
 #ifndef __ISCSI_ISER_H__
 #define __ISCSI_ISER_H__

drivers/infiniband/ulp/iser/iser_initiator.c

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: iser_initiator.c 6964 2006-05-07 11:11:43Z ogerlitz $
  */
 #include <linux/kernel.h>
 #include <linux/slab.h>

drivers/infiniband/ulp/iser/iser_memory.c

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: iser_memory.c 6964 2006-05-07 11:11:43Z ogerlitz $
  */
 #include <linux/module.h>
 #include <linux/kernel.h>

drivers/infiniband/ulp/iser/iser_verbs.c

@@ -29,8 +29,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: iser_verbs.c 7051 2006-05-10 12:29:11Z ogerlitz $
  */
 #include <linux/kernel.h>
 #include <linux/module.h>

drivers/infiniband/ulp/srp/ib_srp.c

@@ -28,8 +28,6 @@
  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
- *
- * $Id: ib_srp.c 3932 2005-11-01 17:19:29Z roland $
  */
 
 #include <linux/module.h>
@@ -49,8 +47,6 @@
 #include <scsi/srp.h>
 #include <scsi/scsi_transport_srp.h>
 
-#include <rdma/ib_cache.h>
-
 #include "ib_srp.h"
 
 #define DRV_NAME	"ib_srp"
@@ -183,10 +179,10 @@ static int srp_init_qp(struct srp_target_port *target,
 	if (!attr)
 		return -ENOMEM;
 
-	ret = ib_find_cached_pkey(target->srp_host->srp_dev->dev,
+	ret = ib_find_pkey(target->srp_host->srp_dev->dev,
-				  target->srp_host->port,
+			   target->srp_host->port,
-				  be16_to_cpu(target->path.pkey),
+			   be16_to_cpu(target->path.pkey),
-				  &attr->pkey_index);
+			   &attr->pkey_index);
 	if (ret)
 		goto out;
 
@@ -1883,8 +1879,7 @@ static ssize_t srp_create_target(struct device *dev,
 	if (ret)
 		goto err;
 
-	ib_get_cached_gid(host->srp_dev->dev, host->port, 0,
+	ib_query_gid(host->srp_dev->dev, host->port, 0, &target->path.sgid);
-			  &target->path.sgid);
 
 	shost_printk(KERN_DEBUG, target->scsi_host, PFX
 		     "new target: id_ext %016llx ioc_guid %016llx pkey %04x "