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

* 'for-linus' of master.kernel.org:/pub/scm/linux/kernel/git/roland/infiniband:
  IPoIB: Make sure struct ipoib_neigh.queue is always initialized
  IB/iser: Use the new verbs DMA mapping functions
  IB/srp: Use new verbs IB DMA mapping functions
  IPoIB: Use the new verbs DMA mapping functions
  IB/core: Use the new verbs DMA mapping functions
  IB/ipath: Implement new verbs DMA mapping functions
  IB: Add DMA mapping functions to allow device drivers to interpose
  RDMA/cma: Export rdma cm interface to userspace
  RDMA/cma: Add support for RDMA_PS_UDP
  RDMA/cma: Allow early transition to RTS to handle lost CM messages
  RDMA/cma: Report connect info with connect events
  RDMA/cma: Remove unneeded qp_type parameter from rdma_cm
  IB/ipath: Fix IRQ for PCI Express HCAs
  RDMA/amso1100: Fix memory leak in c2_qp_modify()
  IB/iser: Remove unused "write-only" variables
  IB/ipath: Remove unused "write-only" variables
  IB/fmr: ib_flush_fmr_pool() may wait too long

drivers/infiniband/core/Makefile

@@ -1,9 +1,11 @@
 infiniband-$(CONFIG_INFINIBAND_ADDR_TRANS)	:= ib_addr.o rdma_cm.o
+user_access-$(CONFIG_INFINIBAND_ADDR_TRANS)	:= rdma_ucm.o
 
 obj-$(CONFIG_INFINIBAND) +=		ib_core.o ib_mad.o ib_sa.o \
 					ib_cm.o iw_cm.o $(infiniband-y)
 obj-$(CONFIG_INFINIBAND_USER_MAD) +=	ib_umad.o
-obj-$(CONFIG_INFINIBAND_USER_ACCESS) +=	ib_uverbs.o ib_ucm.o
+obj-$(CONFIG_INFINIBAND_USER_ACCESS) +=	ib_uverbs.o ib_ucm.o \
+					$(user_access-y)
 
 ib_core-y :=			packer.o ud_header.o verbs.o sysfs.o \
 				device.o fmr_pool.o cache.o
@@ -18,6 +20,8 @@ iw_cm-y :=			iwcm.o
 
 rdma_cm-y :=			cma.o
 
+rdma_ucm-y :=			ucma.o
+
 ib_addr-y :=			addr.o
 
 ib_umad-y :=			user_mad.o

drivers/infiniband/core/cm.c

@@ -3289,6 +3289,10 @@ static int cm_init_qp_rts_attr(struct cm_id_private *cm_id_priv,
 
 	spin_lock_irqsave(&cm_id_priv->lock, flags);
 	switch (cm_id_priv->id.state) {
+	/* Allow transition to RTS before sending REP */
+	case IB_CM_REQ_RCVD:
+	case IB_CM_MRA_REQ_SENT:
+
 	case IB_CM_REP_RCVD:
 	case IB_CM_MRA_REP_SENT:
 	case IB_CM_REP_SENT:

drivers/infiniband/core/cma.c

@@ -70,6 +70,7 @@ static DEFINE_MUTEX(lock);
 static struct workqueue_struct *cma_wq;
 static DEFINE_IDR(sdp_ps);
 static DEFINE_IDR(tcp_ps);
+static DEFINE_IDR(udp_ps);
 
 struct cma_device {
 	struct list_head	list;
@@ -133,7 +134,6 @@ struct rdma_id_private {
 
 	u32			seq_num;
 	u32			qp_num;
-	enum ib_qp_type		qp_type;
 	u8			srq;
 };
 
@@ -392,7 +392,6 @@ int rdma_create_qp(struct rdma_cm_id *id, struct ib_pd *pd,
 
 	id->qp = qp;
 	id_priv->qp_num = qp->qp_num;
-	id_priv->qp_type = qp->qp_type;
 	id_priv->srq = (qp->srq != NULL);
 	return 0;
 err:
@@ -510,9 +509,17 @@ static inline int cma_any_addr(struct sockaddr *addr)
 	return cma_zero_addr(addr) || cma_loopback_addr(addr);
 }
 
+static inline __be16 cma_port(struct sockaddr *addr)
+{
+	if (addr->sa_family == AF_INET)
+		return ((struct sockaddr_in *) addr)->sin_port;
+	else
+		return ((struct sockaddr_in6 *) addr)->sin6_port;
+}
+
 static inline int cma_any_port(struct sockaddr *addr)
 {
-	return !((struct sockaddr_in *) addr)->sin_port;
+	return !cma_port(addr);
 }
 
 static int cma_get_net_info(void *hdr, enum rdma_port_space ps,
@@ -594,20 +601,6 @@ static inline int cma_user_data_offset(enum rdma_port_space ps)
 	}
 }
 
-static int cma_notify_user(struct rdma_id_private *id_priv,
-			   enum rdma_cm_event_type type, int status,
-			   void *data, u8 data_len)
-{
-	struct rdma_cm_event event;
-
-	event.event = type;
-	event.status = status;
-	event.private_data = data;
-	event.private_data_len = data_len;
-
-	return id_priv->id.event_handler(&id_priv->id, &event);
-}
-
 static void cma_cancel_route(struct rdma_id_private *id_priv)
 {
 	switch (rdma_node_get_transport(id_priv->id.device->node_type)) {
@@ -776,63 +769,61 @@ static int cma_verify_rep(struct rdma_id_private *id_priv, void *data)
 	return 0;
 }
 
-static int cma_rtu_recv(struct rdma_id_private *id_priv)
+static void cma_set_rep_event_data(struct rdma_cm_event *event,
+				   struct ib_cm_rep_event_param *rep_data,
+				   void *private_data)
 {
-	int ret;
-
-	ret = cma_modify_qp_rts(&id_priv->id);
-	if (ret)
-		goto reject;
-
-	return 0;
-reject:
-	cma_modify_qp_err(&id_priv->id);
-	ib_send_cm_rej(id_priv->cm_id.ib, IB_CM_REJ_CONSUMER_DEFINED,
-		       NULL, 0, NULL, 0);
-	return ret;
+	event->param.conn.private_data = private_data;
+	event->param.conn.private_data_len = IB_CM_REP_PRIVATE_DATA_SIZE;
+	event->param.conn.responder_resources = rep_data->responder_resources;
+	event->param.conn.initiator_depth = rep_data->initiator_depth;
+	event->param.conn.flow_control = rep_data->flow_control;
+	event->param.conn.rnr_retry_count = rep_data->rnr_retry_count;
+	event->param.conn.srq = rep_data->srq;
+	event->param.conn.qp_num = rep_data->remote_qpn;
 }
 
 static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 {
 	struct rdma_id_private *id_priv = cm_id->context;
-	enum rdma_cm_event_type event;
-	u8 private_data_len = 0;
-	int ret = 0, status = 0;
+	struct rdma_cm_event event;
+	int ret = 0;
 
 	atomic_inc(&id_priv->dev_remove);
 	if (!cma_comp(id_priv, CMA_CONNECT))
 		goto out;
 
+	memset(&event, 0, sizeof event);
 	switch (ib_event->event) {
 	case IB_CM_REQ_ERROR:
 	case IB_CM_REP_ERROR:
-		event = RDMA_CM_EVENT_UNREACHABLE;
-		status = -ETIMEDOUT;
+		event.event = RDMA_CM_EVENT_UNREACHABLE;
+		event.status = -ETIMEDOUT;
 		break;
 	case IB_CM_REP_RECEIVED:
-		status = cma_verify_rep(id_priv, ib_event->private_data);
-		if (status)
-			event = RDMA_CM_EVENT_CONNECT_ERROR;
+		event.status = cma_verify_rep(id_priv, ib_event->private_data);
+		if (event.status)
+			event.event = RDMA_CM_EVENT_CONNECT_ERROR;
 		else if (id_priv->id.qp && id_priv->id.ps != RDMA_PS_SDP) {
-			status = cma_rep_recv(id_priv);
-			event = status ? RDMA_CM_EVENT_CONNECT_ERROR :
-					 RDMA_CM_EVENT_ESTABLISHED;
+			event.status = cma_rep_recv(id_priv);
+			event.event = event.status ? RDMA_CM_EVENT_CONNECT_ERROR :
+						     RDMA_CM_EVENT_ESTABLISHED;
 		} else
-			event = RDMA_CM_EVENT_CONNECT_RESPONSE;
-		private_data_len = IB_CM_REP_PRIVATE_DATA_SIZE;
+			event.event = RDMA_CM_EVENT_CONNECT_RESPONSE;
+		cma_set_rep_event_data(&event, &ib_event->param.rep_rcvd,
+				       ib_event->private_data);
 		break;
 	case IB_CM_RTU_RECEIVED:
-		status = cma_rtu_recv(id_priv);
-		event = status ? RDMA_CM_EVENT_CONNECT_ERROR :
-				 RDMA_CM_EVENT_ESTABLISHED;
+	case IB_CM_USER_ESTABLISHED:
+		event.event = RDMA_CM_EVENT_ESTABLISHED;
 		break;
 	case IB_CM_DREQ_ERROR:
-		status = -ETIMEDOUT; /* fall through */
+		event.status = -ETIMEDOUT; /* fall through */
 	case IB_CM_DREQ_RECEIVED:
 	case IB_CM_DREP_RECEIVED:
 		if (!cma_comp_exch(id_priv, CMA_CONNECT, CMA_DISCONNECT))
 			goto out;
-		event = RDMA_CM_EVENT_DISCONNECTED;
+		event.event = RDMA_CM_EVENT_DISCONNECTED;
 		break;
 	case IB_CM_TIMEWAIT_EXIT:
 	case IB_CM_MRA_RECEIVED:
@@ -840,9 +831,10 @@ static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 		goto out;
 	case IB_CM_REJ_RECEIVED:
 		cma_modify_qp_err(&id_priv->id);
-		status = ib_event->param.rej_rcvd.reason;
-		event = RDMA_CM_EVENT_REJECTED;
-		private_data_len = IB_CM_REJ_PRIVATE_DATA_SIZE;
+		event.status = ib_event->param.rej_rcvd.reason;
+		event.event = RDMA_CM_EVENT_REJECTED;
+		event.param.conn.private_data = ib_event->private_data;
+		event.param.conn.private_data_len = IB_CM_REJ_PRIVATE_DATA_SIZE;
 		break;
 	default:
 		printk(KERN_ERR "RDMA CMA: unexpected IB CM event: %d",
@@ -850,8 +842,7 @@ static int cma_ib_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 		goto out;
 	}
 
-	ret = cma_notify_user(id_priv, event, status, ib_event->private_data,
-			      private_data_len);
+	ret = id_priv->id.event_handler(&id_priv->id, &event);
 	if (ret) {
 		/* Destroy the CM ID by returning a non-zero value. */
 		id_priv->cm_id.ib = NULL;
@@ -865,8 +856,8 @@ out:
 	return ret;
 }
 
-static struct rdma_id_private *cma_new_id(struct rdma_cm_id *listen_id,
-					  struct ib_cm_event *ib_event)
+static struct rdma_id_private *cma_new_conn_id(struct rdma_cm_id *listen_id,
+					       struct ib_cm_event *ib_event)
 {
 	struct rdma_id_private *id_priv;
 	struct rdma_cm_id *id;
@@ -913,9 +904,61 @@ err:
 	return NULL;
 }
 
+static struct rdma_id_private *cma_new_udp_id(struct rdma_cm_id *listen_id,
+					      struct ib_cm_event *ib_event)
+{
+	struct rdma_id_private *id_priv;
+	struct rdma_cm_id *id;
+	union cma_ip_addr *src, *dst;
+	__u16 port;
+	u8 ip_ver;
+	int ret;
+
+	id = rdma_create_id(listen_id->event_handler, listen_id->context,
+			    listen_id->ps);
+	if (IS_ERR(id))
+		return NULL;
+
+
+	if (cma_get_net_info(ib_event->private_data, listen_id->ps,
+			     &ip_ver, &port, &src, &dst))
+		goto err;
+
+	cma_save_net_info(&id->route.addr, &listen_id->route.addr,
+			  ip_ver, port, src, dst);
+
+	ret = rdma_translate_ip(&id->route.addr.src_addr,
+				&id->route.addr.dev_addr);
+	if (ret)
+		goto err;
+
+	id_priv = container_of(id, struct rdma_id_private, id);
+	id_priv->state = CMA_CONNECT;
+	return id_priv;
+err:
+	rdma_destroy_id(id);
+	return NULL;
+}
+
+static void cma_set_req_event_data(struct rdma_cm_event *event,
+				   struct ib_cm_req_event_param *req_data,
+				   void *private_data, int offset)
+{
+	event->param.conn.private_data = private_data + offset;
+	event->param.conn.private_data_len = IB_CM_REQ_PRIVATE_DATA_SIZE - offset;
+	event->param.conn.responder_resources = req_data->responder_resources;
+	event->param.conn.initiator_depth = req_data->initiator_depth;
+	event->param.conn.flow_control = req_data->flow_control;
+	event->param.conn.retry_count = req_data->retry_count;
+	event->param.conn.rnr_retry_count = req_data->rnr_retry_count;
+	event->param.conn.srq = req_data->srq;
+	event->param.conn.qp_num = req_data->remote_qpn;
+}
+
 static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 {
 	struct rdma_id_private *listen_id, *conn_id;
+	struct rdma_cm_event event;
 	int offset, ret;
 
 	listen_id = cm_id->context;
@@ -925,7 +968,19 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 		goto out;
 	}
 
-	conn_id = cma_new_id(&listen_id->id, ib_event);
+	memset(&event, 0, sizeof event);
+	offset = cma_user_data_offset(listen_id->id.ps);
+	event.event = RDMA_CM_EVENT_CONNECT_REQUEST;
+	if (listen_id->id.ps == RDMA_PS_UDP) {
+		conn_id = cma_new_udp_id(&listen_id->id, ib_event);
+		event.param.ud.private_data = ib_event->private_data + offset;
+		event.param.ud.private_data_len =
+				IB_CM_SIDR_REQ_PRIVATE_DATA_SIZE - offset;
+	} else {
+		conn_id = cma_new_conn_id(&listen_id->id, ib_event);
+		cma_set_req_event_data(&event, &ib_event->param.req_rcvd,
+				       ib_event->private_data, offset);
+	}
 	if (!conn_id) {
 		ret = -ENOMEM;
 		goto out;
@@ -942,10 +997,7 @@ static int cma_req_handler(struct ib_cm_id *cm_id, struct ib_cm_event *ib_event)
 	cm_id->context = conn_id;
 	cm_id->cm_handler = cma_ib_handler;
 
-	offset = cma_user_data_offset(listen_id->id.ps);
-	ret = cma_notify_user(conn_id, RDMA_CM_EVENT_CONNECT_REQUEST, 0,
-			      ib_event->private_data + offset,
-			      IB_CM_REQ_PRIVATE_DATA_SIZE - offset);
+	ret = conn_id->id.event_handler(&conn_id->id, &event);
 	if (!ret)
 		goto out;
 
@@ -964,8 +1016,7 @@ out:
 
 static __be64 cma_get_service_id(enum rdma_port_space ps, struct sockaddr *addr)
 {
-	return cpu_to_be64(((u64)ps << 16) +
-	       be16_to_cpu(((struct sockaddr_in *) addr)->sin_port));
+	return cpu_to_be64(((u64)ps << 16) + be16_to_cpu(cma_port(addr)));
 }
 
 static void cma_set_compare_data(enum rdma_port_space ps, struct sockaddr *addr,
@@ -1021,15 +1072,16 @@ static void cma_set_compare_data(enum rdma_port_space ps, struct sockaddr *addr,
 static int cma_iw_handler(struct iw_cm_id *iw_id, struct iw_cm_event *iw_event)
 {
 	struct rdma_id_private *id_priv = iw_id->context;
-	enum rdma_cm_event_type event = 0;
+	struct rdma_cm_event event;
 	struct sockaddr_in *sin;
 	int ret = 0;
 
+	memset(&event, 0, sizeof event);
 	atomic_inc(&id_priv->dev_remove);
 
 	switch (iw_event->event) {
 	case IW_CM_EVENT_CLOSE:
-		event = RDMA_CM_EVENT_DISCONNECTED;
+		event.event = RDMA_CM_EVENT_DISCONNECTED;
 		break;
 	case IW_CM_EVENT_CONNECT_REPLY:
 		sin = (struct sockaddr_in *) &id_priv->id.route.addr.src_addr;
@@ -1037,20 +1089,21 @@ static int cma_iw_handler(struct iw_cm_id *iw_id, struct iw_cm_event *iw_event)
 		sin = (struct sockaddr_in *) &id_priv->id.route.addr.dst_addr;
 		*sin = iw_event->remote_addr;
 		if (iw_event->status)
-			event = RDMA_CM_EVENT_REJECTED;
+			event.event = RDMA_CM_EVENT_REJECTED;
 		else
-			event = RDMA_CM_EVENT_ESTABLISHED;
+			event.event = RDMA_CM_EVENT_ESTABLISHED;
 		break;
 	case IW_CM_EVENT_ESTABLISHED:
-		event = RDMA_CM_EVENT_ESTABLISHED;
+		event.event = RDMA_CM_EVENT_ESTABLISHED;
 		break;
 	default:
 		BUG_ON(1);
 	}
 
-	ret = cma_notify_user(id_priv, event, iw_event->status,
-			      iw_event->private_data,
-			      iw_event->private_data_len);
+	event.status = iw_event->status;
+	event.param.conn.private_data = iw_event->private_data;
+	event.param.conn.private_data_len = iw_event->private_data_len;
+	ret = id_priv->id.event_handler(&id_priv->id, &event);
 	if (ret) {
 		/* Destroy the CM ID by returning a non-zero value. */
 		id_priv->cm_id.iw = NULL;
@@ -1071,6 +1124,7 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id,
 	struct rdma_id_private *listen_id, *conn_id;
 	struct sockaddr_in *sin;
 	struct net_device *dev = NULL;
+	struct rdma_cm_event event;
 	int ret;
 
 	listen_id = cm_id->context;
@@ -1124,9 +1178,11 @@ static int iw_conn_req_handler(struct iw_cm_id *cm_id,
 	sin = (struct sockaddr_in *) &new_cm_id->route.addr.dst_addr;
 	*sin = iw_event->remote_addr;
 
-	ret = cma_notify_user(conn_id, RDMA_CM_EVENT_CONNECT_REQUEST, 0,
-			      iw_event->private_data,
-			      iw_event->private_data_len);
+	memset(&event, 0, sizeof event);
+	event.event = RDMA_CM_EVENT_CONNECT_REQUEST;
+	event.param.conn.private_data = iw_event->private_data;
+	event.param.conn.private_data_len = iw_event->private_data_len;
+	ret = conn_id->id.event_handler(&conn_id->id, &event);
 	if (ret) {
 		/* User wants to destroy the CM ID */
 		conn_id->cm_id.iw = NULL;
@@ -1515,8 +1571,9 @@ static void addr_handler(int status, struct sockaddr *src_addr,
 			 struct rdma_dev_addr *dev_addr, void *context)
 {
 	struct rdma_id_private *id_priv = context;
-	enum rdma_cm_event_type event;
+	struct rdma_cm_event event;
 
+	memset(&event, 0, sizeof event);
 	atomic_inc(&id_priv->dev_remove);
 
 	/*
@@ -1536,14 +1593,15 @@ static void addr_handler(int status, struct sockaddr *src_addr,
 	if (status) {
 		if (!cma_comp_exch(id_priv, CMA_ADDR_RESOLVED, CMA_ADDR_BOUND))
 			goto out;
-		event = RDMA_CM_EVENT_ADDR_ERROR;
+		event.event = RDMA_CM_EVENT_ADDR_ERROR;
+		event.status = status;
 	} else {
 		memcpy(&id_priv->id.route.addr.src_addr, src_addr,
 		       ip_addr_size(src_addr));
-		event = RDMA_CM_EVENT_ADDR_RESOLVED;
+		event.event = RDMA_CM_EVENT_ADDR_RESOLVED;
 	}
 
-	if (cma_notify_user(id_priv, event, status, NULL, 0)) {
+	if (id_priv->id.event_handler(&id_priv->id, &event)) {
 		cma_exch(id_priv, CMA_DESTROYING);
 		cma_release_remove(id_priv);
 		cma_deref_id(id_priv);
@@ -1733,6 +1791,9 @@ static int cma_get_port(struct rdma_id_private *id_priv)
 	case RDMA_PS_TCP:
 		ps = &tcp_ps;
 		break;
+	case RDMA_PS_UDP:
+		ps = &udp_ps;
+		break;
 	default:
 		return -EPROTONOSUPPORT;
 	}
@@ -1821,6 +1882,110 @@ static int cma_format_hdr(void *hdr, enum rdma_port_space ps,
 	return 0;
 }
 
+static int cma_sidr_rep_handler(struct ib_cm_id *cm_id,
+				struct ib_cm_event *ib_event)
+{
+	struct rdma_id_private *id_priv = cm_id->context;
+	struct rdma_cm_event event;
+	struct ib_cm_sidr_rep_event_param *rep = &ib_event->param.sidr_rep_rcvd;
+	int ret = 0;
+
+	memset(&event, 0, sizeof event);
+	atomic_inc(&id_priv->dev_remove);
+	if (!cma_comp(id_priv, CMA_CONNECT))
+		goto out;
+
+	switch (ib_event->event) {
+	case IB_CM_SIDR_REQ_ERROR:
+		event.event = RDMA_CM_EVENT_UNREACHABLE;
+		event.status = -ETIMEDOUT;
+		break;
+	case IB_CM_SIDR_REP_RECEIVED:
+		event.param.ud.private_data = ib_event->private_data;
+		event.param.ud.private_data_len = IB_CM_SIDR_REP_PRIVATE_DATA_SIZE;
+		if (rep->status != IB_SIDR_SUCCESS) {
+			event.event = RDMA_CM_EVENT_UNREACHABLE;
+			event.status = ib_event->param.sidr_rep_rcvd.status;
+			break;
+		}
+		if (rep->qkey != RDMA_UD_QKEY) {
+			event.event = RDMA_CM_EVENT_UNREACHABLE;
+			event.status = -EINVAL;
+			break;
+		}
+		ib_init_ah_from_path(id_priv->id.device, id_priv->id.port_num,
+				     id_priv->id.route.path_rec,
+				     &event.param.ud.ah_attr);
+		event.param.ud.qp_num = rep->qpn;
+		event.param.ud.qkey = rep->qkey;
+		event.event = RDMA_CM_EVENT_ESTABLISHED;
+		event.status = 0;
+		break;
+	default:
+		printk(KERN_ERR "RDMA CMA: unexpected IB CM event: %d",
+		       ib_event->event);
+		goto out;
+	}
+
+	ret = id_priv->id.event_handler(&id_priv->id, &event);
+	if (ret) {
+		/* Destroy the CM ID by returning a non-zero value. */
+		id_priv->cm_id.ib = NULL;
+		cma_exch(id_priv, CMA_DESTROYING);
+		cma_release_remove(id_priv);
+		rdma_destroy_id(&id_priv->id);
+		return ret;
+	}
+out:
+	cma_release_remove(id_priv);
+	return ret;
+}
+
+static int cma_resolve_ib_udp(struct rdma_id_private *id_priv,
+			      struct rdma_conn_param *conn_param)
+{
+	struct ib_cm_sidr_req_param req;
+	struct rdma_route *route;
+	int ret;
+
+	req.private_data_len = sizeof(struct cma_hdr) +
+			       conn_param->private_data_len;
+	req.private_data = kzalloc(req.private_data_len, GFP_ATOMIC);
+	if (!req.private_data)
+		return -ENOMEM;
+
+	if (conn_param->private_data && conn_param->private_data_len)
+		memcpy((void *) req.private_data + sizeof(struct cma_hdr),
+		       conn_param->private_data, conn_param->private_data_len);
+
+	route = &id_priv->id.route;
+	ret = cma_format_hdr((void *) req.private_data, id_priv->id.ps, route);
+	if (ret)
+		goto out;
+
+	id_priv->cm_id.ib = ib_create_cm_id(id_priv->id.device,
+					    cma_sidr_rep_handler, id_priv);
+	if (IS_ERR(id_priv->cm_id.ib)) {
+		ret = PTR_ERR(id_priv->cm_id.ib);
+		goto out;
+	}
+
+	req.path = route->path_rec;
+	req.service_id = cma_get_service_id(id_priv->id.ps,
+					    &route->addr.dst_addr);
+	req.timeout_ms = 1 << (CMA_CM_RESPONSE_TIMEOUT - 8);
+	req.max_cm_retries = CMA_MAX_CM_RETRIES;
+
+	ret = ib_send_cm_sidr_req(id_priv->cm_id.ib, &req);
+	if (ret) {
+		ib_destroy_cm_id(id_priv->cm_id.ib);
+		id_priv->cm_id.ib = NULL;
+	}
+out:
+	kfree(req.private_data);
+	return ret;
+}
+
 static int cma_connect_ib(struct rdma_id_private *id_priv,
 			  struct rdma_conn_param *conn_param)
 {
@@ -1860,7 +2025,7 @@ static int cma_connect_ib(struct rdma_id_private *id_priv,
 	req.service_id = cma_get_service_id(id_priv->id.ps,
 					    &route->addr.dst_addr);
 	req.qp_num = id_priv->qp_num;
-	req.qp_type = id_priv->qp_type;
+	req.qp_type = IB_QPT_RC;
 	req.starting_psn = id_priv->seq_num;
 	req.responder_resources = conn_param->responder_resources;
 	req.initiator_depth = conn_param->initiator_depth;
@@ -1937,13 +2102,15 @@ int rdma_connect(struct rdma_cm_id *id, struct rdma_conn_param *conn_param)
 
 	if (!id->qp) {
 		id_priv->qp_num = conn_param->qp_num;
-		id_priv->qp_type = conn_param->qp_type;
 		id_priv->srq = conn_param->srq;
 	}
 
 	switch (rdma_node_get_transport(id->device->node_type)) {
 	case RDMA_TRANSPORT_IB:
-		ret = cma_connect_ib(id_priv, conn_param);
+		if (id->ps == RDMA_PS_UDP)
+			ret = cma_resolve_ib_udp(id_priv, conn_param);
+		else
+			ret = cma_connect_ib(id_priv, conn_param);
 		break;
 	case RDMA_TRANSPORT_IWARP:
 		ret = cma_connect_iw(id_priv, conn_param);
@@ -1966,11 +2133,25 @@ static int cma_accept_ib(struct rdma_id_private *id_priv,
 			 struct rdma_conn_param *conn_param)
 {
 	struct ib_cm_rep_param rep;
-	int ret;
+	struct ib_qp_attr qp_attr;
+	int qp_attr_mask, ret;
 
-	ret = cma_modify_qp_rtr(&id_priv->id);
-	if (ret)
-		return ret;
+	if (id_priv->id.qp) {
+		ret = cma_modify_qp_rtr(&id_priv->id);
+		if (ret)
+			goto out;
+
+		qp_attr.qp_state = IB_QPS_RTS;
+		ret = ib_cm_init_qp_attr(id_priv->cm_id.ib, &qp_attr,
+					 &qp_attr_mask);
+		if (ret)
+			goto out;
+
+		qp_attr.max_rd_atomic = conn_param->initiator_depth;
+		ret = ib_modify_qp(id_priv->id.qp, &qp_attr, qp_attr_mask);
+		if (ret)
+			goto out;
+	}
 
 	memset(&rep, 0, sizeof rep);
 	rep.qp_num = id_priv->qp_num;
@@ -1985,7 +2166,9 @@ static int cma_accept_ib(struct rdma_id_private *id_priv,
 	rep.rnr_retry_count = conn_param->rnr_retry_count;
 	rep.srq = id_priv->srq ? 1 : 0;
 
-	return ib_send_cm_rep(id_priv->cm_id.ib, &rep);
+	ret = ib_send_cm_rep(id_priv->cm_id.ib, &rep);
+out:
+	return ret;
 }
 
 static int cma_accept_iw(struct rdma_id_private *id_priv,
@@ -2010,6 +2193,24 @@ static int cma_accept_iw(struct rdma_id_private *id_priv,
 	return iw_cm_accept(id_priv->cm_id.iw, &iw_param);
 }
 
+static int cma_send_sidr_rep(struct rdma_id_private *id_priv,
+			     enum ib_cm_sidr_status status,
+			     const void *private_data, int private_data_len)
+{
+	struct ib_cm_sidr_rep_param rep;
+
+	memset(&rep, 0, sizeof rep);
+	rep.status = status;
+	if (status == IB_SIDR_SUCCESS) {
+		rep.qp_num = id_priv->qp_num;
+		rep.qkey = RDMA_UD_QKEY;
+	}
+	rep.private_data = private_data;
+	rep.private_data_len = private_data_len;
+
+	return ib_send_cm_sidr_rep(id_priv->cm_id.ib, &rep);
+}
+
 int rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param)
 {
 	struct rdma_id_private *id_priv;
@@ -2021,13 +2222,16 @@ int rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param)
 
 	if (!id->qp && conn_param) {
 		id_priv->qp_num = conn_param->qp_num;
-		id_priv->qp_type = conn_param->qp_type;
 		id_priv->srq = conn_param->srq;
 	}
 
 	switch (rdma_node_get_transport(id->device->node_type)) {
 	case RDMA_TRANSPORT_IB:
-		if (conn_param)
+		if (id->ps == RDMA_PS_UDP)
+			ret = cma_send_sidr_rep(id_priv, IB_SIDR_SUCCESS,
+						conn_param->private_data,
+						conn_param->private_data_len);
+		else if (conn_param)
 			ret = cma_accept_ib(id_priv, conn_param);
 		else
 			ret = cma_rep_recv(id_priv);
@@ -2051,6 +2255,27 @@ reject:
 }
 EXPORT_SYMBOL(rdma_accept);
 
+int rdma_notify(struct rdma_cm_id *id, enum ib_event_type event)
+{
+	struct rdma_id_private *id_priv;
+	int ret;
+
+	id_priv = container_of(id, struct rdma_id_private, id);
+	if (!cma_comp(id_priv, CMA_CONNECT))
+		return -EINVAL;
+
+	switch (id->device->node_type) {
+	case RDMA_NODE_IB_CA:
+		ret = ib_cm_notify(id_priv->cm_id.ib, event);
+		break;
+	default:
+		ret = 0;
+		break;
+	}
+	return ret;
+}
+EXPORT_SYMBOL(rdma_notify);
+
 int rdma_reject(struct rdma_cm_id *id, const void *private_data,
 		u8 private_data_len)
 {
@@ -2063,9 +2288,13 @@ int rdma_reject(struct rdma_cm_id *id, const void *private_data,
 
 	switch (rdma_node_get_transport(id->device->node_type)) {
 	case RDMA_TRANSPORT_IB:
-		ret = ib_send_cm_rej(id_priv->cm_id.ib,
-				     IB_CM_REJ_CONSUMER_DEFINED, NULL, 0,
-				     private_data, private_data_len);
+		if (id->ps == RDMA_PS_UDP)
+			ret = cma_send_sidr_rep(id_priv, IB_SIDR_REJECT,
+						private_data, private_data_len);
+		else
+			ret = ib_send_cm_rej(id_priv->cm_id.ib,
+					     IB_CM_REJ_CONSUMER_DEFINED, NULL,
+					     0, private_data, private_data_len);
 		break;
 	case RDMA_TRANSPORT_IWARP:
 		ret = iw_cm_reject(id_priv->cm_id.iw,
@@ -2136,6 +2365,7 @@ static void cma_add_one(struct ib_device *device)
 
 static int cma_remove_id_dev(struct rdma_id_private *id_priv)
 {
+	struct rdma_cm_event event;
 	enum cma_state state;
 
 	/* Record that we want to remove the device */
@@ -2150,8 +2380,9 @@ static int cma_remove_id_dev(struct rdma_id_private *id_priv)
 	if (!cma_comp(id_priv, CMA_DEVICE_REMOVAL))
 		return 0;
 
-	return cma_notify_user(id_priv, RDMA_CM_EVENT_DEVICE_REMOVAL,
-			       0, NULL, 0);
+	memset(&event, 0, sizeof event);
+	event.event = RDMA_CM_EVENT_DEVICE_REMOVAL;
+	return id_priv->id.event_handler(&id_priv->id, &event);
 }
 
 static void cma_process_remove(struct cma_device *cma_dev)
@@ -2233,6 +2464,7 @@ static void cma_cleanup(void)
 	destroy_workqueue(cma_wq);
 	idr_destroy(&sdp_ps);
 	idr_destroy(&tcp_ps);
+	idr_destroy(&udp_ps);
 }
 
 module_init(cma_init);

drivers/infiniband/core/fmr_pool.c

@@ -394,20 +394,12 @@ EXPORT_SYMBOL(ib_destroy_fmr_pool);
  */
 int ib_flush_fmr_pool(struct ib_fmr_pool *pool)
 {
-	int serial;
-
-	atomic_inc(&pool->req_ser);
-	/*
-	 * It's OK if someone else bumps req_ser again here -- we'll
-	 * just wait a little longer.
-	 */
-	serial = atomic_read(&pool->req_ser);
+	int serial = atomic_inc_return(&pool->req_ser);
 
 	wake_up_process(pool->thread);
 
 	if (wait_event_interruptible(pool->force_wait,
-				     atomic_read(&pool->flush_ser) -
-				     atomic_read(&pool->req_ser) >= 0))
+				     atomic_read(&pool->flush_ser) - serial >= 0))
 		return -EINTR;
 
 	return 0;

drivers/infiniband/core/mad.c

@@ -998,17 +998,17 @@ int ib_send_mad(struct ib_mad_send_wr_private *mad_send_wr)
 
 	mad_agent = mad_send_wr->send_buf.mad_agent;
 	sge = mad_send_wr->sg_list;
-	sge[0].addr = dma_map_single(mad_agent->device->dma_device,
-				     mad_send_wr->send_buf.mad,
-				     sge[0].length,
-				     DMA_TO_DEVICE);
-	pci_unmap_addr_set(mad_send_wr, header_mapping, sge[0].addr);
-
-	sge[1].addr = dma_map_single(mad_agent->device->dma_device,
-				     ib_get_payload(mad_send_wr),
-				     sge[1].length,
-				     DMA_TO_DEVICE);
-	pci_unmap_addr_set(mad_send_wr, payload_mapping, sge[1].addr);
+	sge[0].addr = ib_dma_map_single(mad_agent->device,
+					mad_send_wr->send_buf.mad,
+					sge[0].length,
+					DMA_TO_DEVICE);
+	mad_send_wr->header_mapping = sge[0].addr;
+
+	sge[1].addr = ib_dma_map_single(mad_agent->device,
+					ib_get_payload(mad_send_wr),
+					sge[1].length,
+					DMA_TO_DEVICE);
+	mad_send_wr->payload_mapping = sge[1].addr;
 
 	spin_lock_irqsave(&qp_info->send_queue.lock, flags);
 	if (qp_info->send_queue.count < qp_info->send_queue.max_active) {
@@ -1026,12 +1026,12 @@ int ib_send_mad(struct ib_mad_send_wr_private *mad_send_wr)
 	}
 	spin_unlock_irqrestore(&qp_info->send_queue.lock, flags);
 	if (ret) {
-		dma_unmap_single(mad_agent->device->dma_device,
-				 pci_unmap_addr(mad_send_wr, header_mapping),
-				 sge[0].length, DMA_TO_DEVICE);
-		dma_unmap_single(mad_agent->device->dma_device,
-				 pci_unmap_addr(mad_send_wr, payload_mapping),
-				 sge[1].length, DMA_TO_DEVICE);
+		ib_dma_unmap_single(mad_agent->device,
+				    mad_send_wr->header_mapping,
+				    sge[0].length, DMA_TO_DEVICE);
+		ib_dma_unmap_single(mad_agent->device,
+				    mad_send_wr->payload_mapping,
+				    sge[1].length, DMA_TO_DEVICE);
 	}
 	return ret;
 }
@@ -1850,11 +1850,11 @@ static void ib_mad_recv_done_handler(struct ib_mad_port_private *port_priv,
 	mad_priv_hdr = container_of(mad_list, struct ib_mad_private_header,
 				    mad_list);
 	recv = container_of(mad_priv_hdr, struct ib_mad_private, header);
-	dma_unmap_single(port_priv->device->dma_device,
-			 pci_unmap_addr(&recv->header, mapping),
-			 sizeof(struct ib_mad_private) -
-			 sizeof(struct ib_mad_private_header),
-			 DMA_FROM_DEVICE);
+	ib_dma_unmap_single(port_priv->device,
+			    recv->header.mapping,
+			    sizeof(struct ib_mad_private) -
+			      sizeof(struct ib_mad_private_header),
+			    DMA_FROM_DEVICE);
 
 	/* Setup MAD receive work completion from "normal" work completion */
 	recv->header.wc = *wc;
@@ -2080,12 +2080,12 @@ static void ib_mad_send_done_handler(struct ib_mad_port_private *port_priv,
 	qp_info = send_queue->qp_info;
 
 retry:
-	dma_unmap_single(mad_send_wr->send_buf.mad_agent->device->dma_device,
-			 pci_unmap_addr(mad_send_wr, header_mapping),
-			 mad_send_wr->sg_list[0].length, DMA_TO_DEVICE);
-	dma_unmap_single(mad_send_wr->send_buf.mad_agent->device->dma_device,
-			 pci_unmap_addr(mad_send_wr, payload_mapping),
-			 mad_send_wr->sg_list[1].length, DMA_TO_DEVICE);
+	ib_dma_unmap_single(mad_send_wr->send_buf.mad_agent->device,
+			    mad_send_wr->header_mapping,
+			    mad_send_wr->sg_list[0].length, DMA_TO_DEVICE);
+	ib_dma_unmap_single(mad_send_wr->send_buf.mad_agent->device,
+			    mad_send_wr->payload_mapping,
+			    mad_send_wr->sg_list[1].length, DMA_TO_DEVICE);
 	queued_send_wr = NULL;
 	spin_lock_irqsave(&send_queue->lock, flags);
 	list_del(&mad_list->list);
@@ -2528,13 +2528,12 @@ static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info,
 				break;
 			}
 		}
-		sg_list.addr = dma_map_single(qp_info->port_priv->
-					        device->dma_device,
-					      &mad_priv->grh,
-					      sizeof *mad_priv -
-					        sizeof mad_priv->header,
-					      DMA_FROM_DEVICE);
-		pci_unmap_addr_set(&mad_priv->header, mapping, sg_list.addr);
+		sg_list.addr = ib_dma_map_single(qp_info->port_priv->device,
+						 &mad_priv->grh,
+						 sizeof *mad_priv -
+						   sizeof mad_priv->header,
+						 DMA_FROM_DEVICE);
+		mad_priv->header.mapping = sg_list.addr;
 		recv_wr.wr_id = (unsigned long)&mad_priv->header.mad_list;
 		mad_priv->header.mad_list.mad_queue = recv_queue;
 
@@ -2549,12 +2548,11 @@ static int ib_mad_post_receive_mads(struct ib_mad_qp_info *qp_info,
 			list_del(&mad_priv->header.mad_list.list);
 			recv_queue->count--;
 			spin_unlock_irqrestore(&recv_queue->lock, flags);
-			dma_unmap_single(qp_info->port_priv->device->dma_device,
-					 pci_unmap_addr(&mad_priv->header,
-							mapping),
-					 sizeof *mad_priv -
-					   sizeof mad_priv->header,
-					 DMA_FROM_DEVICE);
+			ib_dma_unmap_single(qp_info->port_priv->device,
+					    mad_priv->header.mapping,
+					    sizeof *mad_priv -
+					      sizeof mad_priv->header,
+					    DMA_FROM_DEVICE);
 			kmem_cache_free(ib_mad_cache, mad_priv);
 			printk(KERN_ERR PFX "ib_post_recv failed: %d\n", ret);
 			break;
@@ -2586,11 +2584,11 @@ static void cleanup_recv_queue(struct ib_mad_qp_info *qp_info)
 		/* Remove from posted receive MAD list */
 		list_del(&mad_list->list);
 
-		dma_unmap_single(qp_info->port_priv->device->dma_device,
-				 pci_unmap_addr(&recv->header, mapping),
-				 sizeof(struct ib_mad_private) -
-				 sizeof(struct ib_mad_private_header),
-				 DMA_FROM_DEVICE);
+		ib_dma_unmap_single(qp_info->port_priv->device,
+				    recv->header.mapping,
+				    sizeof(struct ib_mad_private) -
+				      sizeof(struct ib_mad_private_header),
+				    DMA_FROM_DEVICE);
 		kmem_cache_free(ib_mad_cache, recv);
 	}
 

drivers/infiniband/core/mad_priv.h

@@ -73,7 +73,7 @@ struct ib_mad_private_header {
 	struct ib_mad_list_head mad_list;
 	struct ib_mad_recv_wc recv_wc;
 	struct ib_wc wc;
-	DECLARE_PCI_UNMAP_ADDR(mapping)
+	u64 mapping;
 } __attribute__ ((packed));
 
 struct ib_mad_private {
@@ -126,8 +126,8 @@ struct ib_mad_send_wr_private {
 	struct list_head agent_list;
 	struct ib_mad_agent_private *mad_agent_priv;
 	struct ib_mad_send_buf send_buf;
-	DECLARE_PCI_UNMAP_ADDR(header_mapping)
-	DECLARE_PCI_UNMAP_ADDR(payload_mapping)
+	u64 header_mapping;
+	u64 payload_mapping;
 	struct ib_send_wr send_wr;
 	struct ib_sge sg_list[IB_MAD_SEND_REQ_MAX_SG];
 	__be64 tid;

drivers/infiniband/core/ucma.c

@@ -0,0 +1,874 @@
+/*
+ * Copyright (c) 2005-2006 Intel Corporation.  All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * 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:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *	copyright notice, this list of conditions and the following
+ *	disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *	copyright notice, this list of conditions and the following
+ *	disclaimer 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>
+#include <linux/mutex.h>
+#include <linux/poll.h>
+#include <linux/idr.h>
+#include <linux/in.h>
+#include <linux/in6.h>
+#include <linux/miscdevice.h>
+
+#include <rdma/rdma_user_cm.h>
+#include <rdma/ib_marshall.h>
+#include <rdma/rdma_cm.h>
+
+MODULE_AUTHOR("Sean Hefty");
+MODULE_DESCRIPTION("RDMA Userspace Connection Manager Access");
+MODULE_LICENSE("Dual BSD/GPL");
+
+enum {
+	UCMA_MAX_BACKLOG	= 128
+};
+
+struct ucma_file {
+	struct mutex		mut;
+	struct file		*filp;
+	struct list_head	ctx_list;
+	struct list_head	event_list;
+	wait_queue_head_t	poll_wait;
+};
+
+struct ucma_context {
+	int			id;
+	struct completion	comp;
+	atomic_t		ref;
+	int			events_reported;
+	int			backlog;
+
+	struct ucma_file	*file;
+	struct rdma_cm_id	*cm_id;
+	u64			uid;
+
+	struct list_head	list;
+};
+
+struct ucma_event {
+	struct ucma_context	*ctx;
+	struct list_head	list;
+	struct rdma_cm_id	*cm_id;
+	struct rdma_ucm_event_resp resp;
+};
+
+static DEFINE_MUTEX(mut);
+static DEFINE_IDR(ctx_idr);
+
+static inline struct ucma_context *_ucma_find_context(int id,
+						      struct ucma_file *file)
+{
+	struct ucma_context *ctx;
+
+	ctx = idr_find(&ctx_idr, id);
+	if (!ctx)
+		ctx = ERR_PTR(-ENOENT);
+	else if (ctx->file != file)
+		ctx = ERR_PTR(-EINVAL);
+	return ctx;
+}
+
+static struct ucma_context *ucma_get_ctx(struct ucma_file *file, int id)
+{
+	struct ucma_context *ctx;
+
+	mutex_lock(&mut);
+	ctx = _ucma_find_context(id, file);
+	if (!IS_ERR(ctx))
+		atomic_inc(&ctx->ref);
+	mutex_unlock(&mut);
+	return ctx;
+}
+
+static void ucma_put_ctx(struct ucma_context *ctx)
+{
+	if (atomic_dec_and_test(&ctx->ref))
+		complete(&ctx->comp);
+}
+
+static struct ucma_context *ucma_alloc_ctx(struct ucma_file *file)
+{
+	struct ucma_context *ctx;
+	int ret;
+
+	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
+	if (!ctx)
+		return NULL;
+
+	atomic_set(&ctx->ref, 1);
+	init_completion(&ctx->comp);
+	ctx->file = file;
+
+	do {
+		ret = idr_pre_get(&ctx_idr, GFP_KERNEL);
+		if (!ret)
+			goto error;
+
+		mutex_lock(&mut);
+		ret = idr_get_new(&ctx_idr, ctx, &ctx->id);
+		mutex_unlock(&mut);
+	} while (ret == -EAGAIN);
+
+	if (ret)
+		goto error;
+
+	list_add_tail(&ctx->list, &file->ctx_list);
+	return ctx;
+
+error:
+	kfree(ctx);
+	return NULL;
+}
+
+static void ucma_copy_conn_event(struct rdma_ucm_conn_param *dst,
+				 struct rdma_conn_param *src)
+{
+	if (src->private_data_len)
+		memcpy(dst->private_data, src->private_data,
+		       src->private_data_len);
+	dst->private_data_len = src->private_data_len;
+	dst->responder_resources =src->responder_resources;
+	dst->initiator_depth = src->initiator_depth;
+	dst->flow_control = src->flow_control;
+	dst->retry_count = src->retry_count;
+	dst->rnr_retry_count = src->rnr_retry_count;
+	dst->srq = src->srq;
+	dst->qp_num = src->qp_num;
+}
+
+static void ucma_copy_ud_event(struct rdma_ucm_ud_param *dst,
+			       struct rdma_ud_param *src)
+{
+	if (src->private_data_len)
+		memcpy(dst->private_data, src->private_data,
+		       src->private_data_len);
+	dst->private_data_len = src->private_data_len;
+	ib_copy_ah_attr_to_user(&dst->ah_attr, &src->ah_attr);
+	dst->qp_num = src->qp_num;
+	dst->qkey = src->qkey;
+}
+
+static void ucma_set_event_context(struct ucma_context *ctx,
+				   struct rdma_cm_event *event,
+				   struct ucma_event *uevent)
+{
+	uevent->ctx = ctx;
+	uevent->resp.uid = ctx->uid;
+	uevent->resp.id = ctx->id;
+}
+
+static int ucma_event_handler(struct rdma_cm_id *cm_id,
+			      struct rdma_cm_event *event)
+{
+	struct ucma_event *uevent;
+	struct ucma_context *ctx = cm_id->context;
+	int ret = 0;
+
+	uevent = kzalloc(sizeof(*uevent), GFP_KERNEL);
+	if (!uevent)
+		return event->event == RDMA_CM_EVENT_CONNECT_REQUEST;
+
+	uevent->cm_id = cm_id;
+	ucma_set_event_context(ctx, event, uevent);
+	uevent->resp.event = event->event;
+	uevent->resp.status = event->status;
+	if (cm_id->ps == RDMA_PS_UDP)
+		ucma_copy_ud_event(&uevent->resp.param.ud, &event->param.ud);
+	else
+		ucma_copy_conn_event(&uevent->resp.param.conn,
+				     &event->param.conn);
+
+	mutex_lock(&ctx->file->mut);
+	if (event->event == RDMA_CM_EVENT_CONNECT_REQUEST) {
+		if (!ctx->backlog) {
+			ret = -EDQUOT;
+			goto out;
+		}
+		ctx->backlog--;
+	}
+	list_add_tail(&uevent->list, &ctx->file->event_list);
+	wake_up_interruptible(&ctx->file->poll_wait);
+out:
+	mutex_unlock(&ctx->file->mut);
+	return ret;
+}
+
+static ssize_t ucma_get_event(struct ucma_file *file, const char __user *inbuf,
+			      int in_len, int out_len)
+{
+	struct ucma_context *ctx;
+	struct rdma_ucm_get_event cmd;
+	struct ucma_event *uevent;
+	int ret = 0;
+	DEFINE_WAIT(wait);
+
+	if (out_len < sizeof uevent->resp)
+		return -ENOSPC;
+
+	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+		return -EFAULT;
+
+	mutex_lock(&file->mut);
+	while (list_empty(&file->event_list)) {
+		if (file->filp->f_flags & O_NONBLOCK) {
+			ret = -EAGAIN;
+			break;
+		}
+
+		if (signal_pending(current)) {
+			ret = -ERESTARTSYS;
+			break;
+		}
+
+		prepare_to_wait(&file->poll_wait, &wait, TASK_INTERRUPTIBLE);
+		mutex_unlock(&file->mut);
+		schedule();
+		mutex_lock(&file->mut);
+		finish_wait(&file->poll_wait, &wait);
+	}
+
+	if (ret)
+		goto done;
+
+	uevent = list_entry(file->event_list.next, struct ucma_event, list);
+
+	if (uevent->resp.event == RDMA_CM_EVENT_CONNECT_REQUEST) {
+		ctx = ucma_alloc_ctx(file);
+		if (!ctx) {
+			ret = -ENOMEM;
+			goto done;
+		}
+		uevent->ctx->backlog++;
+		ctx->cm_id = uevent->cm_id;
+		ctx->cm_id->context = ctx;
+		uevent->resp.id = ctx->id;
+	}
+
+	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+			 &uevent->resp, sizeof uevent->resp)) {
+		ret = -EFAULT;
+		goto done;
+	}
+
+	list_del(&uevent->list);
+	uevent->ctx->events_reported++;
+	kfree(uevent);
+done:
+	mutex_unlock(&file->mut);
+	return ret;
+}
+
+static ssize_t ucma_create_id(struct ucma_file *file,
+				const char __user *inbuf,
+				int in_len, int out_len)
+{
+	struct rdma_ucm_create_id cmd;
+	struct rdma_ucm_create_id_resp resp;
+	struct ucma_context *ctx;
+	int ret;
+
+	if (out_len < sizeof(resp))
+		return -ENOSPC;
+
+	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+		return -EFAULT;
+
+	mutex_lock(&file->mut);
+	ctx = ucma_alloc_ctx(file);
+	mutex_unlock(&file->mut);
+	if (!ctx)
+		return -ENOMEM;
+
+	ctx->uid = cmd.uid;
+	ctx->cm_id = rdma_create_id(ucma_event_handler, ctx, cmd.ps);
+	if (IS_ERR(ctx->cm_id)) {
+		ret = PTR_ERR(ctx->cm_id);
+		goto err1;
+	}
+
+	resp.id = ctx->id;
+	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+			 &resp, sizeof(resp))) {
+		ret = -EFAULT;
+		goto err2;
+	}
+	return 0;
+
+err2:
+	rdma_destroy_id(ctx->cm_id);
+err1:
+	mutex_lock(&mut);
+	idr_remove(&ctx_idr, ctx->id);
+	mutex_unlock(&mut);
+	kfree(ctx);
+	return ret;
+}
+
+static void ucma_cleanup_events(struct ucma_context *ctx)
+{
+	struct ucma_event *uevent, *tmp;
+
+	list_for_each_entry_safe(uevent, tmp, &ctx->file->event_list, list) {
+		if (uevent->ctx != ctx)
+			continue;
+
+		list_del(&uevent->list);
+
+		/* clear incoming connections. */
+		if (uevent->resp.event == RDMA_CM_EVENT_CONNECT_REQUEST)
+			rdma_destroy_id(uevent->cm_id);
+
+		kfree(uevent);
+	}
+}
+
+static int ucma_free_ctx(struct ucma_context *ctx)
+{
+	int events_reported;
+
+	/* No new events will be generated after destroying the id. */
+	rdma_destroy_id(ctx->cm_id);
+
+	/* Cleanup events not yet reported to the user. */
+	mutex_lock(&ctx->file->mut);
+	ucma_cleanup_events(ctx);
+	list_del(&ctx->list);
+	mutex_unlock(&ctx->file->mut);
+
+	events_reported = ctx->events_reported;
+	kfree(ctx);
+	return events_reported;
+}
+
+static ssize_t ucma_destroy_id(struct ucma_file *file, const char __user *inbuf,
+			       int in_len, int out_len)
+{
+	struct rdma_ucm_destroy_id cmd;
+	struct rdma_ucm_destroy_id_resp resp;
+	struct ucma_context *ctx;
+	int ret = 0;
+
+	if (out_len < sizeof(resp))
+		return -ENOSPC;
+
+	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+		return -EFAULT;
+
+	mutex_lock(&mut);
+	ctx = _ucma_find_context(cmd.id, file);
+	if (!IS_ERR(ctx))
+		idr_remove(&ctx_idr, ctx->id);
+	mutex_unlock(&mut);
+
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
+
+	ucma_put_ctx(ctx);
+	wait_for_completion(&ctx->comp);
+	resp.events_reported = ucma_free_ctx(ctx);
+
+	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+			 &resp, sizeof(resp)))
+		ret = -EFAULT;
+
+	return ret;
+}
+
+static ssize_t ucma_bind_addr(struct ucma_file *file, const char __user *inbuf,
+			      int in_len, int out_len)
+{
+	struct rdma_ucm_bind_addr cmd;
+	struct ucma_context *ctx;
+	int ret;
+
+	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+		return -EFAULT;
+
+	ctx = ucma_get_ctx(file, cmd.id);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
+
+	ret = rdma_bind_addr(ctx->cm_id, (struct sockaddr *) &cmd.addr);
+	ucma_put_ctx(ctx);
+	return ret;
+}
+
+static ssize_t ucma_resolve_addr(struct ucma_file *file,
+				 const char __user *inbuf,
+				 int in_len, int out_len)
+{
+	struct rdma_ucm_resolve_addr cmd;
+	struct ucma_context *ctx;
+	int ret;
+
+	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+		return -EFAULT;
+
+	ctx = ucma_get_ctx(file, cmd.id);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
+
+	ret = rdma_resolve_addr(ctx->cm_id, (struct sockaddr *) &cmd.src_addr,
+				(struct sockaddr *) &cmd.dst_addr,
+				cmd.timeout_ms);
+	ucma_put_ctx(ctx);
+	return ret;
+}
+
+static ssize_t ucma_resolve_route(struct ucma_file *file,
+				  const char __user *inbuf,
+				  int in_len, int out_len)
+{
+	struct rdma_ucm_resolve_route cmd;
+	struct ucma_context *ctx;
+	int ret;
+
+	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+		return -EFAULT;
+
+	ctx = ucma_get_ctx(file, cmd.id);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
+
+	ret = rdma_resolve_route(ctx->cm_id, cmd.timeout_ms);
+	ucma_put_ctx(ctx);
+	return ret;
+}
+
+static void ucma_copy_ib_route(struct rdma_ucm_query_route_resp *resp,
+			       struct rdma_route *route)
+{
+	struct rdma_dev_addr *dev_addr;
+
+	resp->num_paths = route->num_paths;
+	switch (route->num_paths) {
+	case 0:
+		dev_addr = &route->addr.dev_addr;
+		ib_addr_get_dgid(dev_addr,
+				 (union ib_gid *) &resp->ib_route[0].dgid);
+		ib_addr_get_sgid(dev_addr,
+				 (union ib_gid *) &resp->ib_route[0].sgid);
+		resp->ib_route[0].pkey = cpu_to_be16(ib_addr_get_pkey(dev_addr));
+		break;
+	case 2:
+		ib_copy_path_rec_to_user(&resp->ib_route[1],
+					 &route->path_rec[1]);
+		/* fall through */
+	case 1:
+		ib_copy_path_rec_to_user(&resp->ib_route[0],
+					 &route->path_rec[0]);
+		break;
+	default:
+		break;
+	}
+}
+
+static ssize_t ucma_query_route(struct ucma_file *file,
+				const char __user *inbuf,
+				int in_len, int out_len)
+{
+	struct rdma_ucm_query_route cmd;
+	struct rdma_ucm_query_route_resp resp;
+	struct ucma_context *ctx;
+	struct sockaddr *addr;
+	int ret = 0;
+
+	if (out_len < sizeof(resp))
+		return -ENOSPC;
+
+	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+		return -EFAULT;
+
+	ctx = ucma_get_ctx(file, cmd.id);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
+
+	memset(&resp, 0, sizeof resp);
+	addr = &ctx->cm_id->route.addr.src_addr;
+	memcpy(&resp.src_addr, addr, addr->sa_family == AF_INET ?
+				     sizeof(struct sockaddr_in) :
+				     sizeof(struct sockaddr_in6));
+	addr = &ctx->cm_id->route.addr.dst_addr;
+	memcpy(&resp.dst_addr, addr, addr->sa_family == AF_INET ?
+				     sizeof(struct sockaddr_in) :
+				     sizeof(struct sockaddr_in6));
+	if (!ctx->cm_id->device)
+		goto out;
+
+	resp.node_guid = ctx->cm_id->device->node_guid;
+	resp.port_num = ctx->cm_id->port_num;
+	switch (rdma_node_get_transport(ctx->cm_id->device->node_type)) {
+	case RDMA_TRANSPORT_IB:
+		ucma_copy_ib_route(&resp, &ctx->cm_id->route);
+		break;
+	default:
+		break;
+	}
+
+out:
+	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+			 &resp, sizeof(resp)))
+		ret = -EFAULT;
+
+	ucma_put_ctx(ctx);
+	return ret;
+}
+
+static void ucma_copy_conn_param(struct rdma_conn_param *dst,
+				 struct rdma_ucm_conn_param *src)
+{
+	dst->private_data = src->private_data;
+	dst->private_data_len = src->private_data_len;
+	dst->responder_resources =src->responder_resources;
+	dst->initiator_depth = src->initiator_depth;
+	dst->flow_control = src->flow_control;
+	dst->retry_count = src->retry_count;
+	dst->rnr_retry_count = src->rnr_retry_count;
+	dst->srq = src->srq;
+	dst->qp_num = src->qp_num;
+}
+
+static ssize_t ucma_connect(struct ucma_file *file, const char __user *inbuf,
+			    int in_len, int out_len)
+{
+	struct rdma_ucm_connect cmd;
+	struct rdma_conn_param conn_param;
+	struct ucma_context *ctx;
+	int ret;
+
+	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+		return -EFAULT;
+
+	if (!cmd.conn_param.valid)
+		return -EINVAL;
+
+	ctx = ucma_get_ctx(file, cmd.id);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
+
+	ucma_copy_conn_param(&conn_param, &cmd.conn_param);
+	ret = rdma_connect(ctx->cm_id, &conn_param);
+	ucma_put_ctx(ctx);
+	return ret;
+}
+
+static ssize_t ucma_listen(struct ucma_file *file, const char __user *inbuf,
+			   int in_len, int out_len)
+{
+	struct rdma_ucm_listen cmd;
+	struct ucma_context *ctx;
+	int ret;
+
+	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+		return -EFAULT;
+
+	ctx = ucma_get_ctx(file, cmd.id);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
+
+	ctx->backlog = cmd.backlog > 0 && cmd.backlog < UCMA_MAX_BACKLOG ?
+		       cmd.backlog : UCMA_MAX_BACKLOG;
+	ret = rdma_listen(ctx->cm_id, ctx->backlog);
+	ucma_put_ctx(ctx);
+	return ret;
+}
+
+static ssize_t ucma_accept(struct ucma_file *file, const char __user *inbuf,
+			   int in_len, int out_len)
+{
+	struct rdma_ucm_accept cmd;
+	struct rdma_conn_param conn_param;
+	struct ucma_context *ctx;
+	int ret;
+
+	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+		return -EFAULT;
+
+	ctx = ucma_get_ctx(file, cmd.id);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
+
+	if (cmd.conn_param.valid) {
+		ctx->uid = cmd.uid;
+		ucma_copy_conn_param(&conn_param, &cmd.conn_param);
+		ret = rdma_accept(ctx->cm_id, &conn_param);
+	} else
+		ret = rdma_accept(ctx->cm_id, NULL);
+
+	ucma_put_ctx(ctx);
+	return ret;
+}
+
+static ssize_t ucma_reject(struct ucma_file *file, const char __user *inbuf,
+			   int in_len, int out_len)
+{
+	struct rdma_ucm_reject cmd;
+	struct ucma_context *ctx;
+	int ret;
+
+	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+		return -EFAULT;
+
+	ctx = ucma_get_ctx(file, cmd.id);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
+
+	ret = rdma_reject(ctx->cm_id, cmd.private_data, cmd.private_data_len);
+	ucma_put_ctx(ctx);
+	return ret;
+}
+
+static ssize_t ucma_disconnect(struct ucma_file *file, const char __user *inbuf,
+			       int in_len, int out_len)
+{
+	struct rdma_ucm_disconnect cmd;
+	struct ucma_context *ctx;
+	int ret;
+
+	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+		return -EFAULT;
+
+	ctx = ucma_get_ctx(file, cmd.id);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
+
+	ret = rdma_disconnect(ctx->cm_id);
+	ucma_put_ctx(ctx);
+	return ret;
+}
+
+static ssize_t ucma_init_qp_attr(struct ucma_file *file,
+				 const char __user *inbuf,
+				 int in_len, int out_len)
+{
+	struct rdma_ucm_init_qp_attr cmd;
+	struct ib_uverbs_qp_attr resp;
+	struct ucma_context *ctx;
+	struct ib_qp_attr qp_attr;
+	int ret;
+
+	if (out_len < sizeof(resp))
+		return -ENOSPC;
+
+	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+		return -EFAULT;
+
+	ctx = ucma_get_ctx(file, cmd.id);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
+
+	resp.qp_attr_mask = 0;
+	memset(&qp_attr, 0, sizeof qp_attr);
+	qp_attr.qp_state = cmd.qp_state;
+	ret = rdma_init_qp_attr(ctx->cm_id, &qp_attr, &resp.qp_attr_mask);
+	if (ret)
+		goto out;
+
+	ib_copy_qp_attr_to_user(&resp, &qp_attr);
+	if (copy_to_user((void __user *)(unsigned long)cmd.response,
+			 &resp, sizeof(resp)))
+		ret = -EFAULT;
+
+out:
+	ucma_put_ctx(ctx);
+	return ret;
+}
+
+static ssize_t ucma_notify(struct ucma_file *file, const char __user *inbuf,
+			   int in_len, int out_len)
+{
+	struct rdma_ucm_notify cmd;
+	struct ucma_context *ctx;
+	int ret;
+
+	if (copy_from_user(&cmd, inbuf, sizeof(cmd)))
+		return -EFAULT;
+
+	ctx = ucma_get_ctx(file, cmd.id);
+	if (IS_ERR(ctx))
+		return PTR_ERR(ctx);
+
+	ret = rdma_notify(ctx->cm_id, (enum ib_event_type) cmd.event);
+	ucma_put_ctx(ctx);
+	return ret;
+}
+
+static ssize_t (*ucma_cmd_table[])(struct ucma_file *file,
+				   const char __user *inbuf,
+				   int in_len, int out_len) = {
+	[RDMA_USER_CM_CMD_CREATE_ID]	= ucma_create_id,
+	[RDMA_USER_CM_CMD_DESTROY_ID]	= ucma_destroy_id,
+	[RDMA_USER_CM_CMD_BIND_ADDR]	= ucma_bind_addr,
+	[RDMA_USER_CM_CMD_RESOLVE_ADDR]	= ucma_resolve_addr,
+	[RDMA_USER_CM_CMD_RESOLVE_ROUTE]= ucma_resolve_route,
+	[RDMA_USER_CM_CMD_QUERY_ROUTE]	= ucma_query_route,
+	[RDMA_USER_CM_CMD_CONNECT]	= ucma_connect,
+	[RDMA_USER_CM_CMD_LISTEN]	= ucma_listen,
+	[RDMA_USER_CM_CMD_ACCEPT]	= ucma_accept,
+	[RDMA_USER_CM_CMD_REJECT]	= ucma_reject,
+	[RDMA_USER_CM_CMD_DISCONNECT]	= ucma_disconnect,
+	[RDMA_USER_CM_CMD_INIT_QP_ATTR]	= ucma_init_qp_attr,
+	[RDMA_USER_CM_CMD_GET_EVENT]	= ucma_get_event,
+	[RDMA_USER_CM_CMD_GET_OPTION]	= NULL,
+	[RDMA_USER_CM_CMD_SET_OPTION]	= NULL,
+	[RDMA_USER_CM_CMD_NOTIFY]	= ucma_notify,
+};
+
+static ssize_t ucma_write(struct file *filp, const char __user *buf,
+			  size_t len, loff_t *pos)
+{
+	struct ucma_file *file = filp->private_data;
+	struct rdma_ucm_cmd_hdr hdr;
+	ssize_t ret;
+
+	if (len < sizeof(hdr))
+		return -EINVAL;
+
+	if (copy_from_user(&hdr, buf, sizeof(hdr)))
+		return -EFAULT;
+
+	if (hdr.cmd < 0 || hdr.cmd >= ARRAY_SIZE(ucma_cmd_table))
+		return -EINVAL;
+
+	if (hdr.in + sizeof(hdr) > len)
+		return -EINVAL;
+
+	if (!ucma_cmd_table[hdr.cmd])
+		return -ENOSYS;
+
+	ret = ucma_cmd_table[hdr.cmd](file, buf + sizeof(hdr), hdr.in, hdr.out);
+	if (!ret)
+		ret = len;
+
+	return ret;
+}
+
+static unsigned int ucma_poll(struct file *filp, struct poll_table_struct *wait)
+{
+	struct ucma_file *file = filp->private_data;
+	unsigned int mask = 0;
+
+	poll_wait(filp, &file->poll_wait, wait);
+
+	if (!list_empty(&file->event_list))
+		mask = POLLIN | POLLRDNORM;
+
+	return mask;
+}
+
+static int ucma_open(struct inode *inode, struct file *filp)
+{
+	struct ucma_file *file;
+
+	file = kmalloc(sizeof *file, GFP_KERNEL);
+	if (!file)
+		return -ENOMEM;
+
+	INIT_LIST_HEAD(&file->event_list);
+	INIT_LIST_HEAD(&file->ctx_list);
+	init_waitqueue_head(&file->poll_wait);
+	mutex_init(&file->mut);
+
+	filp->private_data = file;
+	file->filp = filp;
+	return 0;
+}
+
+static int ucma_close(struct inode *inode, struct file *filp)
+{
+	struct ucma_file *file = filp->private_data;
+	struct ucma_context *ctx, *tmp;
+
+	mutex_lock(&file->mut);
+	list_for_each_entry_safe(ctx, tmp, &file->ctx_list, list) {
+		mutex_unlock(&file->mut);
+
+		mutex_lock(&mut);
+		idr_remove(&ctx_idr, ctx->id);
+		mutex_unlock(&mut);
+
+		ucma_free_ctx(ctx);
+		mutex_lock(&file->mut);
+	}
+	mutex_unlock(&file->mut);
+	kfree(file);
+	return 0;
+}
+
+static struct file_operations ucma_fops = {
+	.owner 	 = THIS_MODULE,
+	.open 	 = ucma_open,
+	.release = ucma_close,
+	.write	 = ucma_write,
+	.poll    = ucma_poll,
+};
+
+static struct miscdevice ucma_misc = {
+	.minor	= MISC_DYNAMIC_MINOR,
+	.name	= "rdma_cm",
+	.fops	= &ucma_fops,
+};
+
+static ssize_t show_abi_version(struct device *dev,
+				struct device_attribute *attr,
+				char *buf)
+{
+	return sprintf(buf, "%d\n", RDMA_USER_CM_ABI_VERSION);
+}
+static DEVICE_ATTR(abi_version, S_IRUGO, show_abi_version, NULL);
+
+static int __init ucma_init(void)
+{
+	int ret;
+
+	ret = misc_register(&ucma_misc);
+	if (ret)
+		return ret;
+
+	ret = device_create_file(ucma_misc.this_device, &dev_attr_abi_version);
+	if (ret) {
+		printk(KERN_ERR "rdma_ucm: couldn't create abi_version attr\n");
+		goto err;
+	}
+	return 0;
+err:
+	misc_deregister(&ucma_misc);
+	return ret;
+}
+
+static void __exit ucma_cleanup(void)
+{
+	device_remove_file(ucma_misc.this_device, &dev_attr_abi_version);
+	misc_deregister(&ucma_misc);
+	idr_destroy(&ctx_idr);
+}
+
+module_init(ucma_init);
+module_exit(ucma_cleanup);

drivers/infiniband/core/uverbs_marshall.c

@@ -32,8 +32,8 @@
 
 #include <rdma/ib_marshall.h>
 
-static void ib_copy_ah_attr_to_user(struct ib_uverbs_ah_attr *dst,
-				    struct ib_ah_attr *src)
+void ib_copy_ah_attr_to_user(struct ib_uverbs_ah_attr *dst,
+			     struct ib_ah_attr *src)
 {
 	memcpy(dst->grh.dgid, src->grh.dgid.raw, sizeof src->grh.dgid);
 	dst->grh.flow_label        = src->grh.flow_label;
@@ -47,6 +47,7 @@ static void ib_copy_ah_attr_to_user(struct ib_uverbs_ah_attr *dst,
 	dst->is_global             = src->ah_flags & IB_AH_GRH ? 1 : 0;
 	dst->port_num 	    	   = src->port_num;
 }
+EXPORT_SYMBOL(ib_copy_ah_attr_to_user);
 
 void ib_copy_qp_attr_to_user(struct ib_uverbs_qp_attr *dst,
 			     struct ib_qp_attr *src)

drivers/infiniband/core/uverbs_mem.c

@@ -52,8 +52,8 @@ static void __ib_umem_release(struct ib_device *dev, struct ib_umem *umem, int d
 	int i;
 
 	list_for_each_entry_safe(chunk, tmp, &umem->chunk_list, list) {
-		dma_unmap_sg(dev->dma_device, chunk->page_list,
-			     chunk->nents, DMA_BIDIRECTIONAL);
+		ib_dma_unmap_sg(dev, chunk->page_list,
+				chunk->nents, DMA_BIDIRECTIONAL);
 		for (i = 0; i < chunk->nents; ++i) {
 			if (umem->writable && dirty)
 				set_page_dirty_lock(chunk->page_list[i].page);
@@ -136,10 +136,10 @@ int ib_umem_get(struct ib_device *dev, struct ib_umem *mem,
 				chunk->page_list[i].length = PAGE_SIZE;
 			}
 
-			chunk->nmap = dma_map_sg(dev->dma_device,
-						 &chunk->page_list[0],
-						 chunk->nents,
-						 DMA_BIDIRECTIONAL);
+			chunk->nmap = ib_dma_map_sg(dev,
+						    &chunk->page_list[0],
+						    chunk->nents,
+						    DMA_BIDIRECTIONAL);
 			if (chunk->nmap <= 0) {
 				for (i = 0; i < chunk->nents; ++i)
 					put_page(chunk->page_list[i].page);

drivers/infiniband/hw/amso1100/c2_qp.c

@@ -161,8 +161,10 @@ int c2_qp_modify(struct c2_dev *c2dev, struct c2_qp *qp,
 
 	if (attr_mask & IB_QP_STATE) {
 		/* Ensure the state is valid */
-		if (attr->qp_state < 0 || attr->qp_state > IB_QPS_ERR)
-			return -EINVAL;
+		if (attr->qp_state < 0 || attr->qp_state > IB_QPS_ERR) {
+			err = -EINVAL;
+			goto bail0;
+		}
 
 		wr.next_qp_state = cpu_to_be32(to_c2_state(attr->qp_state));
 
@@ -184,9 +186,10 @@ int c2_qp_modify(struct c2_dev *c2dev, struct c2_qp *qp,
 		if (attr->cur_qp_state != IB_QPS_RTR &&
 		    attr->cur_qp_state != IB_QPS_RTS &&
 		    attr->cur_qp_state != IB_QPS_SQD &&
-		    attr->cur_qp_state != IB_QPS_SQE)
-			return -EINVAL;
-		else
+		    attr->cur_qp_state != IB_QPS_SQE) {
+			err = -EINVAL;
+			goto bail0;
+		} else
 			wr.next_qp_state =
 			    cpu_to_be32(to_c2_state(attr->cur_qp_state));
 

drivers/infiniband/hw/ipath/Makefile

@@ -6,6 +6,7 @@ obj-$(CONFIG_INFINIBAND_IPATH) += ib_ipath.o
 ib_ipath-y := \
 	ipath_cq.o \
 	ipath_diag.o \
+	ipath_dma.o \
 	ipath_driver.o \
 	ipath_eeprom.o \
 	ipath_file_ops.o \

drivers/infiniband/hw/ipath/ipath_dma.c

@@ -0,0 +1,189 @@
+/*
+ * Copyright (c) 2006 QLogic, Corporation. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * 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:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer 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 <rdma/ib_verbs.h>
+
+#include "ipath_verbs.h"
+
+#define BAD_DMA_ADDRESS ((u64) 0)
+
+/*
+ * The following functions implement driver specific replacements
+ * for the ib_dma_*() functions.
+ *
+ * These functions return kernel virtual addresses instead of
+ * device bus addresses since the driver uses the CPU to copy
+ * data instead of using hardware DMA.
+ */
+
+static int ipath_mapping_error(struct ib_device *dev, u64 dma_addr)
+{
+	return dma_addr == BAD_DMA_ADDRESS;
+}
+
+static u64 ipath_dma_map_single(struct ib_device *dev,
+			        void *cpu_addr, size_t size,
+			        enum dma_data_direction direction)
+{
+	BUG_ON(!valid_dma_direction(direction));
+	return (u64) cpu_addr;
+}
+
+static void ipath_dma_unmap_single(struct ib_device *dev,
+				   u64 addr, size_t size,
+				   enum dma_data_direction direction)
+{
+	BUG_ON(!valid_dma_direction(direction));
+}
+
+static u64 ipath_dma_map_page(struct ib_device *dev,
+			      struct page *page,
+			      unsigned long offset,
+			      size_t size,
+			      enum dma_data_direction direction)
+{
+	u64 addr;
+
+	BUG_ON(!valid_dma_direction(direction));
+
+	if (offset + size > PAGE_SIZE) {
+		addr = BAD_DMA_ADDRESS;
+		goto done;
+	}
+
+	addr = (u64) page_address(page);
+	if (addr)
+		addr += offset;
+	/* TODO: handle highmem pages */
+
+done:
+	return addr;
+}
+
+static void ipath_dma_unmap_page(struct ib_device *dev,
+				 u64 addr, size_t size,
+				 enum dma_data_direction direction)
+{
+	BUG_ON(!valid_dma_direction(direction));
+}
+
+int ipath_map_sg(struct ib_device *dev, struct scatterlist *sg, int nents,
+		 enum dma_data_direction direction)
+{
+	u64 addr;
+	int i;
+	int ret = nents;
+
+	BUG_ON(!valid_dma_direction(direction));
+
+	for (i = 0; i < nents; i++) {
+		addr = (u64) page_address(sg[i].page);
+		/* TODO: handle highmem pages */
+		if (!addr) {
+			ret = 0;
+			break;
+		}
+	}
+	return ret;
+}
+
+static void ipath_unmap_sg(struct ib_device *dev,
+			   struct scatterlist *sg, int nents,
+			   enum dma_data_direction direction)
+{
+	BUG_ON(!valid_dma_direction(direction));
+}
+
+static u64 ipath_sg_dma_address(struct ib_device *dev, struct scatterlist *sg)
+{
+	u64 addr = (u64) page_address(sg->page);
+
+	if (addr)
+		addr += sg->offset;
+	return addr;
+}
+
+static unsigned int ipath_sg_dma_len(struct ib_device *dev,
+				     struct scatterlist *sg)
+{
+	return sg->length;
+}
+
+static void ipath_sync_single_for_cpu(struct ib_device *dev,
+				      u64 addr,
+				      size_t size,
+				      enum dma_data_direction dir)
+{
+}
+
+static void ipath_sync_single_for_device(struct ib_device *dev,
+					 u64 addr,
+					 size_t size,
+					 enum dma_data_direction dir)
+{
+}
+
+static void *ipath_dma_alloc_coherent(struct ib_device *dev, size_t size,
+				      u64 *dma_handle, gfp_t flag)
+{
+	struct page *p;
+	void *addr = NULL;
+
+	p = alloc_pages(flag, get_order(size));
+	if (p)
+		addr = page_address(p);
+	if (dma_handle)
+		*dma_handle = (u64) addr;
+	return addr;
+}
+
+static void ipath_dma_free_coherent(struct ib_device *dev, size_t size,
+				    void *cpu_addr, dma_addr_t dma_handle)
+{
+	free_pages((unsigned long) cpu_addr, get_order(size));
+}
+
+struct ib_dma_mapping_ops ipath_dma_mapping_ops = {
+	ipath_mapping_error,
+	ipath_dma_map_single,
+	ipath_dma_unmap_single,
+	ipath_dma_map_page,
+	ipath_dma_unmap_page,
+	ipath_map_sg,
+	ipath_unmap_sg,
+	ipath_sg_dma_address,
+	ipath_sg_dma_len,
+	ipath_sync_single_for_cpu,
+	ipath_sync_single_for_device,
+	ipath_dma_alloc_coherent,
+	ipath_dma_free_coherent
+};

drivers/infiniband/hw/ipath/ipath_driver.c

@@ -1825,8 +1825,6 @@ void ipath_write_kreg_port(const struct ipath_devdata *dd, ipath_kreg regno,
  */
 void ipath_shutdown_device(struct ipath_devdata *dd)
 {
-	u64 val;
-
 	ipath_dbg("Shutting down the device\n");
 
 	dd->ipath_flags |= IPATH_LINKUNK;
@@ -1849,7 +1847,7 @@ void ipath_shutdown_device(struct ipath_devdata *dd)
 	 */
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl, 0ULL);
 	/* flush it */
-	val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
+	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
 	/*
 	 * enough for anything that's going to trickle out to have actually
 	 * done so.

drivers/infiniband/hw/ipath/ipath_file_ops.c

@@ -699,7 +699,6 @@ static int ipath_manage_rcvq(struct ipath_portdata *pd, unsigned subport,
 			     int start_stop)
 {
 	struct ipath_devdata *dd = pd->port_dd;
-	u64 tval;
 
 	ipath_cdbg(PROC, "%sabling rcv for unit %u port %u:%u\n",
 		   start_stop ? "en" : "dis", dd->ipath_unit,
@@ -729,7 +728,7 @@ static int ipath_manage_rcvq(struct ipath_portdata *pd, unsigned subport,
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvctrl,
 			 dd->ipath_rcvctrl);
 	/* now be sure chip saw it before we return */
-	tval = ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
+	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
 	if (start_stop) {
 		/*
 		 * And try to be sure that tail reg update has happened too.
@@ -738,7 +737,7 @@ static int ipath_manage_rcvq(struct ipath_portdata *pd, unsigned subport,
 		 * in memory copy, since we could overwrite an update by the
 		 * chip if we did.
 		 */
-		tval = ipath_read_ureg32(dd, ur_rcvhdrtail, pd->port_port);
+		ipath_read_ureg32(dd, ur_rcvhdrtail, pd->port_port);
 	}
 	/* always; new head should be equal to new tail; see above */
 bail:

drivers/infiniband/hw/ipath/ipath_iba6110.c

@@ -1447,7 +1447,7 @@ static void ipath_ht_tidtemplate(struct ipath_devdata *dd)
 static int ipath_ht_early_init(struct ipath_devdata *dd)
 {
 	u32 __iomem *piobuf;
-	u32 pioincr, val32, egrsize;
+	u32 pioincr, val32;
 	int i;
 
 	/*
@@ -1467,7 +1467,6 @@ static int ipath_ht_early_init(struct ipath_devdata *dd)
 	 * errors interrupts if we ever see one).
 	 */
 	dd->ipath_rcvegrbufsize = dd->ipath_piosize2k;
-	egrsize = dd->ipath_rcvegrbufsize;
 
 	/*
 	 * the min() check here is currently a nop, but it may not

drivers/infiniband/hw/ipath/ipath_iba6120.c

@@ -602,7 +602,7 @@ static void ipath_pe_init_hwerrors(struct ipath_devdata *dd)
  */
 static int ipath_pe_bringup_serdes(struct ipath_devdata *dd)
 {
-	u64 val, tmp, config1, prev_val;
+	u64 val, config1, prev_val;
 	int ret = 0;
 
 	ipath_dbg("Trying to bringup serdes\n");
@@ -633,7 +633,7 @@ static int ipath_pe_bringup_serdes(struct ipath_devdata *dd)
 		| INFINIPATH_SERDC0_L1PWR_DN;
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_serdesconfig0, val);
 	/* be sure chip saw it */
-	tmp = ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
+	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
 	udelay(5);		/* need pll reset set at least for a bit */
 	/*
 	 * after PLL is reset, set the per-lane Resets and TxIdle and
@@ -647,7 +647,7 @@ static int ipath_pe_bringup_serdes(struct ipath_devdata *dd)
 		   "and txidle (%llx)\n", (unsigned long long) val);
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_serdesconfig0, val);
 	/* be sure chip saw it */
-	tmp = ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
+	ipath_read_kreg64(dd, dd->ipath_kregs->kr_scratch);
 	/* need PLL reset clear for at least 11 usec before lane
 	 * resets cleared; give it a few more to be sure */
 	udelay(15);
@@ -851,12 +851,12 @@ static int ipath_setup_pe_config(struct ipath_devdata *dd,
 	int pos, ret;
 
 	dd->ipath_msi_lo = 0;	/* used as a flag during reset processing */
-	dd->ipath_irq = pdev->irq;
 	ret = pci_enable_msi(dd->pcidev);
 	if (ret)
 		ipath_dev_err(dd, "pci_enable_msi failed: %d, "
 			      "interrupts may not work\n", ret);
 	/* continue even if it fails, we may still be OK... */
+	dd->ipath_irq = pdev->irq;
 
 	if ((pos = pci_find_capability(dd->pcidev, PCI_CAP_ID_MSI))) {
 		u16 control;

drivers/infiniband/hw/ipath/ipath_init_chip.c

@@ -347,10 +347,9 @@ done:
 static int init_chip_reset(struct ipath_devdata *dd,
 			   struct ipath_portdata **pdp)
 {
-	struct ipath_portdata *pd;
 	u32 rtmp;
 
-	*pdp = pd = dd->ipath_pd[0];
+	*pdp = dd->ipath_pd[0];
 	/* ensure chip does no sends or receives while we re-initialize */
 	dd->ipath_control = dd->ipath_sendctrl = dd->ipath_rcvctrl = 0U;
 	ipath_write_kreg(dd, dd->ipath_kregs->kr_rcvctrl, 0);

drivers/infiniband/hw/ipath/ipath_intr.c

@@ -598,10 +598,9 @@ static int handle_errors(struct ipath_devdata *dd, ipath_err_t errs)
 	 * on close
 	 */
 	if (errs & INFINIPATH_E_RRCVHDRFULL) {
-		int any;
 		u32 hd, tl;
 		ipath_stats.sps_hdrqfull++;
-		for (any = i = 0; i < dd->ipath_cfgports; i++) {
+		for (i = 0; i < dd->ipath_cfgports; i++) {
 			struct ipath_portdata *pd = dd->ipath_pd[i];
 			if (i == 0) {
 				hd = dd->ipath_port0head;

drivers/infiniband/hw/ipath/ipath_keys.c

@@ -134,7 +134,7 @@ int ipath_lkey_ok(struct ipath_qp *qp, struct ipath_sge *isge,
 	 */
 	if (sge->lkey == 0) {
 		isge->mr = NULL;
-		isge->vaddr = bus_to_virt(sge->addr);
+		isge->vaddr = (void *) sge->addr;
 		isge->length = sge->length;
 		isge->sge_length = sge->length;
 		ret = 1;
@@ -202,12 +202,12 @@ int ipath_rkey_ok(struct ipath_qp *qp, struct ipath_sge_state *ss,
 	int ret;
 
 	/*
-	 * We use RKEY == zero for physical addresses
-	 * (see ipath_get_dma_mr).
+	 * We use RKEY == zero for kernel virtual addresses
+	 * (see ipath_get_dma_mr and ipath_dma.c).
 	 */
 	if (rkey == 0) {
 		sge->mr = NULL;
-		sge->vaddr = phys_to_virt(vaddr);
+		sge->vaddr = (void *) vaddr;
 		sge->length = len;
 		sge->sge_length = len;
 		ss->sg_list = NULL;

drivers/infiniband/hw/ipath/ipath_mr.c

@@ -54,6 +54,8 @@ static inline struct ipath_fmr *to_ifmr(struct ib_fmr *ibfmr)
  * @acc: access flags
  *
  * Returns the memory region on success, otherwise returns an errno.
+ * Note that all DMA addresses should be created via the
+ * struct ib_dma_mapping_ops functions (see ipath_dma.c).
  */
 struct ib_mr *ipath_get_dma_mr(struct ib_pd *pd, int acc)
 {
@@ -149,8 +151,7 @@ struct ib_mr *ipath_reg_phys_mr(struct ib_pd *pd,
 	m = 0;
 	n = 0;
 	for (i = 0; i < num_phys_buf; i++) {
-		mr->mr.map[m]->segs[n].vaddr =
-			phys_to_virt(buffer_list[i].addr);
+		mr->mr.map[m]->segs[n].vaddr = (void *) buffer_list[i].addr;
 		mr->mr.map[m]->segs[n].length = buffer_list[i].size;
 		mr->mr.length += buffer_list[i].size;
 		n++;
@@ -347,7 +348,7 @@ int ipath_map_phys_fmr(struct ib_fmr *ibfmr, u64 * page_list,
 	n = 0;
 	ps = 1 << fmr->page_shift;
 	for (i = 0; i < list_len; i++) {
-		fmr->mr.map[m]->segs[n].vaddr = phys_to_virt(page_list[i]);
+		fmr->mr.map[m]->segs[n].vaddr = (void *) page_list[i];
 		fmr->mr.map[m]->segs[n].length = ps;
 		if (++n == IPATH_SEGSZ) {
 			m++;

drivers/infiniband/hw/ipath/ipath_sysfs.c

@@ -215,7 +215,6 @@ static ssize_t store_mlid(struct device *dev,
 			  size_t count)
 {
 	struct ipath_devdata *dd = dev_get_drvdata(dev);
-	int unit;
 	u16 mlid;
 	int ret;
 
@@ -223,8 +222,6 @@ static ssize_t store_mlid(struct device *dev,
 	if (ret < 0 || mlid < IPATH_MULTICAST_LID_BASE)
 		goto invalid;
 
-	unit = dd->ipath_unit;
-
 	dd->ipath_mlid = mlid;
 
 	goto bail;

drivers/infiniband/hw/ipath/ipath_verbs.c

@@ -1599,6 +1599,7 @@ int ipath_register_ib_device(struct ipath_devdata *dd)
 	dev->detach_mcast = ipath_multicast_detach;
 	dev->process_mad = ipath_process_mad;
 	dev->mmap = ipath_mmap;
+	dev->dma_ops = &ipath_dma_mapping_ops;
 
 	snprintf(dev->node_desc, sizeof(dev->node_desc),
 		 IPATH_IDSTR " %s", init_utsname()->nodename);

drivers/infiniband/hw/ipath/ipath_verbs.h

@@ -812,4 +812,6 @@ extern unsigned int ib_ipath_max_srq_wrs;
 
 extern const u32 ib_ipath_rnr_table[];
 
+extern struct ib_dma_mapping_ops ipath_dma_mapping_ops;
+
 #endif				/* IPATH_VERBS_H */

drivers/infiniband/ulp/ipoib/ipoib.h

@@ -105,12 +105,12 @@ struct ipoib_mcast;
 
 struct ipoib_rx_buf {
 	struct sk_buff *skb;
-	dma_addr_t	mapping;
+	u64		mapping;
 };
 
 struct ipoib_tx_buf {
 	struct sk_buff *skb;
-	DECLARE_PCI_UNMAP_ADDR(mapping)
+	u64		mapping;
 };
 
 /*

drivers/infiniband/ulp/ipoib/ipoib_ib.c

@@ -109,9 +109,8 @@ static int ipoib_ib_post_receive(struct net_device *dev, int id)
 	ret = ib_post_recv(priv->qp, &param, &bad_wr);
 	if (unlikely(ret)) {
 		ipoib_warn(priv, "receive failed for buf %d (%d)\n", id, ret);
-		dma_unmap_single(priv->ca->dma_device,
-				 priv->rx_ring[id].mapping,
-				 IPOIB_BUF_SIZE, DMA_FROM_DEVICE);
+		ib_dma_unmap_single(priv->ca, priv->rx_ring[id].mapping,
+				    IPOIB_BUF_SIZE, DMA_FROM_DEVICE);
 		dev_kfree_skb_any(priv->rx_ring[id].skb);
 		priv->rx_ring[id].skb = NULL;
 	}
@@ -123,7 +122,7 @@ static int ipoib_alloc_rx_skb(struct net_device *dev, int id)
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 	struct sk_buff *skb;
-	dma_addr_t addr;
+	u64 addr;
 
 	skb = dev_alloc_skb(IPOIB_BUF_SIZE + 4);
 	if (!skb)
@@ -136,10 +135,9 @@ static int ipoib_alloc_rx_skb(struct net_device *dev, int id)
 	 */
 	skb_reserve(skb, 4);
 
-	addr = dma_map_single(priv->ca->dma_device,
-			      skb->data, IPOIB_BUF_SIZE,
-			      DMA_FROM_DEVICE);
-	if (unlikely(dma_mapping_error(addr))) {
+	addr = ib_dma_map_single(priv->ca, skb->data, IPOIB_BUF_SIZE,
+				 DMA_FROM_DEVICE);
+	if (unlikely(ib_dma_mapping_error(priv->ca, addr))) {
 		dev_kfree_skb_any(skb);
 		return -EIO;
 	}
@@ -174,7 +172,7 @@ static void ipoib_ib_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 	unsigned int wr_id = wc->wr_id & ~IPOIB_OP_RECV;
 	struct sk_buff *skb;
-	dma_addr_t addr;
+	u64 addr;
 
 	ipoib_dbg_data(priv, "recv completion: id %d, op %d, status: %d\n",
 		       wr_id, wc->opcode, wc->status);
@@ -193,8 +191,8 @@ static void ipoib_ib_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
 			ipoib_warn(priv, "failed recv event "
 				   "(status=%d, wrid=%d vend_err %x)\n",
 				   wc->status, wr_id, wc->vendor_err);
-		dma_unmap_single(priv->ca->dma_device, addr,
-				 IPOIB_BUF_SIZE, DMA_FROM_DEVICE);
+		ib_dma_unmap_single(priv->ca, addr,
+				    IPOIB_BUF_SIZE, DMA_FROM_DEVICE);
 		dev_kfree_skb_any(skb);
 		priv->rx_ring[wr_id].skb = NULL;
 		return;
@@ -212,8 +210,7 @@ static void ipoib_ib_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
 	ipoib_dbg_data(priv, "received %d bytes, SLID 0x%04x\n",
 		       wc->byte_len, wc->slid);
 
-	dma_unmap_single(priv->ca->dma_device, addr,
-			 IPOIB_BUF_SIZE, DMA_FROM_DEVICE);
+	ib_dma_unmap_single(priv->ca, addr, IPOIB_BUF_SIZE, DMA_FROM_DEVICE);
 
 	skb_put(skb, wc->byte_len);
 	skb_pull(skb, IB_GRH_BYTES);
@@ -261,10 +258,8 @@ static void ipoib_ib_handle_tx_wc(struct net_device *dev, struct ib_wc *wc)
 
 	tx_req = &priv->tx_ring[wr_id];
 
-	dma_unmap_single(priv->ca->dma_device,
-			 pci_unmap_addr(tx_req, mapping),
-			 tx_req->skb->len,
-			 DMA_TO_DEVICE);
+	ib_dma_unmap_single(priv->ca, tx_req->mapping,
+			    tx_req->skb->len, DMA_TO_DEVICE);
 
 	++priv->stats.tx_packets;
 	priv->stats.tx_bytes += tx_req->skb->len;
@@ -311,7 +306,7 @@ void ipoib_ib_completion(struct ib_cq *cq, void *dev_ptr)
 static inline int post_send(struct ipoib_dev_priv *priv,
 			    unsigned int wr_id,
 			    struct ib_ah *address, u32 qpn,
-			    dma_addr_t addr, int len)
+			    u64 addr, int len)
 {
 	struct ib_send_wr *bad_wr;
 
@@ -330,7 +325,7 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb,
 {
 	struct ipoib_dev_priv *priv = netdev_priv(dev);
 	struct ipoib_tx_buf *tx_req;
-	dma_addr_t addr;
+	u64 addr;
 
 	if (unlikely(skb->len > dev->mtu + INFINIBAND_ALEN)) {
 		ipoib_warn(priv, "packet len %d (> %d) too long to send, dropping\n",
@@ -353,21 +348,20 @@ void ipoib_send(struct net_device *dev, struct sk_buff *skb,
 	 */
 	tx_req = &priv->tx_ring[priv->tx_head & (ipoib_sendq_size - 1)];
 	tx_req->skb = skb;
-	addr = dma_map_single(priv->ca->dma_device, skb->data, skb->len,
-			      DMA_TO_DEVICE);
-	if (unlikely(dma_mapping_error(addr))) {
+	addr = ib_dma_map_single(priv->ca, skb->data, skb->len,
+				 DMA_TO_DEVICE);
+	if (unlikely(ib_dma_mapping_error(priv->ca, addr))) {
 		++priv->stats.tx_errors;
 		dev_kfree_skb_any(skb);
 		return;
 	}
-	pci_unmap_addr_set(tx_req, mapping, addr);
+	tx_req->mapping = addr;
 
 	if (unlikely(post_send(priv, priv->tx_head & (ipoib_sendq_size - 1),
 			       address->ah, qpn, addr, skb->len))) {
 		ipoib_warn(priv, "post_send failed\n");
 		++priv->stats.tx_errors;
-		dma_unmap_single(priv->ca->dma_device, addr, skb->len,
-				 DMA_TO_DEVICE);
+		ib_dma_unmap_single(priv->ca, addr, skb->len, DMA_TO_DEVICE);
 		dev_kfree_skb_any(skb);
 	} else {
 		dev->trans_start = jiffies;
@@ -538,24 +532,27 @@ int ipoib_ib_dev_stop(struct net_device *dev)
 			while ((int) priv->tx_tail - (int) priv->tx_head < 0) {
 				tx_req = &priv->tx_ring[priv->tx_tail &
 							(ipoib_sendq_size - 1)];
-				dma_unmap_single(priv->ca->dma_device,
-						 pci_unmap_addr(tx_req, mapping),
-						 tx_req->skb->len,
-						 DMA_TO_DEVICE);
+				ib_dma_unmap_single(priv->ca,
+						    tx_req->mapping,
+						    tx_req->skb->len,
+						    DMA_TO_DEVICE);
 				dev_kfree_skb_any(tx_req->skb);
 				++priv->tx_tail;
 			}
 
-			for (i = 0; i < ipoib_recvq_size; ++i)
-				if (priv->rx_ring[i].skb) {
-					dma_unmap_single(priv->ca->dma_device,
-							 pci_unmap_addr(&priv->rx_ring[i],
-									mapping),
-							 IPOIB_BUF_SIZE,
-							 DMA_FROM_DEVICE);
-					dev_kfree_skb_any(priv->rx_ring[i].skb);
-					priv->rx_ring[i].skb = NULL;
-				}
+			for (i = 0; i < ipoib_recvq_size; ++i) {
+				struct ipoib_rx_buf *rx_req;
+
+				rx_req = &priv->rx_ring[i];
+				if (!rx_req->skb)
+					continue;
+				ib_dma_unmap_single(priv->ca,
+						    rx_req->mapping,
+						    IPOIB_BUF_SIZE,
+						    DMA_FROM_DEVICE);
+				dev_kfree_skb_any(rx_req->skb);
+				rx_req->skb = NULL;
+			}
 
 			goto timeout;
 		}

drivers/infiniband/ulp/ipoib/ipoib_main.c

@@ -497,8 +497,6 @@ static void neigh_add_path(struct sk_buff *skb, struct net_device *dev)
 		return;
 	}
 
-	skb_queue_head_init(&neigh->queue);
-
 	/*
 	 * We can only be called from ipoib_start_xmit, so we're
 	 * inside tx_lock -- no need to save/restore flags.
@@ -806,6 +804,7 @@ struct ipoib_neigh *ipoib_neigh_alloc(struct neighbour *neighbour)
 
 	neigh->neighbour = neighbour;
 	*to_ipoib_neigh(neighbour) = neigh;
+	skb_queue_head_init(&neigh->queue);
 
 	return neigh;
 }

drivers/infiniband/ulp/iser/iscsi_iser.h

@@ -182,7 +182,7 @@ struct iser_regd_buf {
 	struct iser_mem_reg     reg;        /* memory registration info        */
 	void                    *virt_addr;
 	struct iser_device      *device;    /* device->device for dma_unmap    */
-	dma_addr_t              dma_addr;   /* if non zero, addr for dma_unmap */
+	u64                     dma_addr;   /* if non zero, addr for dma_unmap */
 	enum dma_data_direction direction;  /* direction for dma_unmap	       */
 	unsigned int            data_size;
 	atomic_t                ref_count;  /* refcount, freed when dec to 0   */

drivers/infiniband/ulp/iser/iser_initiator.c

@@ -487,10 +487,8 @@ int iser_send_control(struct iscsi_conn *conn,
 	struct iscsi_iser_conn *iser_conn = conn->dd_data;
 	struct iser_desc *mdesc = mtask->dd_data;
 	struct iser_dto *send_dto = NULL;
-	unsigned int itt;
 	unsigned long data_seg_len;
 	int err = 0;
-	unsigned char opcode;
 	struct iser_regd_buf *regd_buf;
 	struct iser_device *device;
 
@@ -512,8 +510,6 @@ int iser_send_control(struct iscsi_conn *conn,
 
 	iser_reg_single(device, send_dto->regd[0], DMA_TO_DEVICE);
 
-	itt = ntohl(mtask->hdr->itt);
-	opcode = mtask->hdr->opcode & ISCSI_OPCODE_MASK;
 	data_seg_len = ntoh24(mtask->hdr->dlength);
 
 	if (data_seg_len > 0) {

drivers/infiniband/ulp/iser/iser_memory.c

@@ -52,7 +52,7 @@
  */
 int iser_regd_buff_release(struct iser_regd_buf *regd_buf)
 {
-	struct device *dma_device;
+	struct ib_device *dev;
 
 	if ((atomic_read(&regd_buf->ref_count) == 0) ||
 	    atomic_dec_and_test(&regd_buf->ref_count)) {
@@ -61,8 +61,8 @@ int iser_regd_buff_release(struct iser_regd_buf *regd_buf)
 			iser_unreg_mem(&regd_buf->reg);
 
 		if (regd_buf->dma_addr) {
-			dma_device = regd_buf->device->ib_device->dma_device;
-			dma_unmap_single(dma_device,
+			dev = regd_buf->device->ib_device;
+			ib_dma_unmap_single(dev,
 					 regd_buf->dma_addr,
 					 regd_buf->data_size,
 					 regd_buf->direction);
@@ -84,12 +84,12 @@ void iser_reg_single(struct iser_device *device,
 		     struct iser_regd_buf *regd_buf,
 		     enum dma_data_direction direction)
 {
-	dma_addr_t dma_addr;
+	u64 dma_addr;
 
-	dma_addr  = dma_map_single(device->ib_device->dma_device,
-				   regd_buf->virt_addr,
-				   regd_buf->data_size, direction);
-	BUG_ON(dma_mapping_error(dma_addr));
+	dma_addr = ib_dma_map_single(device->ib_device,
+				     regd_buf->virt_addr,
+				     regd_buf->data_size, direction);
+	BUG_ON(ib_dma_mapping_error(device->ib_device, dma_addr));
 
 	regd_buf->reg.lkey = device->mr->lkey;
 	regd_buf->reg.len  = regd_buf->data_size;
@@ -107,7 +107,7 @@ int iser_start_rdma_unaligned_sg(struct iscsi_iser_cmd_task  *iser_ctask,
 				 enum iser_data_dir cmd_dir)
 {
 	int dma_nents;
-	struct device *dma_device;
+	struct ib_device *dev;
 	char *mem = NULL;
 	struct iser_data_buf *data = &iser_ctask->data[cmd_dir];
 	unsigned long  cmd_data_len = data->data_len;
@@ -147,17 +147,12 @@ int iser_start_rdma_unaligned_sg(struct iscsi_iser_cmd_task  *iser_ctask,
 
 	iser_ctask->data_copy[cmd_dir].copy_buf  = mem;
 
-	dma_device = iser_ctask->iser_conn->ib_conn->device->ib_device->dma_device;
-
-	if (cmd_dir == ISER_DIR_OUT)
-		dma_nents = dma_map_sg(dma_device,
-				       &iser_ctask->data_copy[cmd_dir].sg_single,
-				       1, DMA_TO_DEVICE);
-	else
-		dma_nents = dma_map_sg(dma_device,
-				       &iser_ctask->data_copy[cmd_dir].sg_single,
-				       1, DMA_FROM_DEVICE);
-
+	dev = iser_ctask->iser_conn->ib_conn->device->ib_device;
+	dma_nents = ib_dma_map_sg(dev,
+				  &iser_ctask->data_copy[cmd_dir].sg_single,
+				  1,
+				  (cmd_dir == ISER_DIR_OUT) ?
+				  DMA_TO_DEVICE : DMA_FROM_DEVICE);
 	BUG_ON(dma_nents == 0);
 
 	iser_ctask->data_copy[cmd_dir].dma_nents = dma_nents;
@@ -170,19 +165,16 @@ int iser_start_rdma_unaligned_sg(struct iscsi_iser_cmd_task  *iser_ctask,
 void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_cmd_task *iser_ctask,
 				     enum iser_data_dir         cmd_dir)
 {
-	struct device *dma_device;
+	struct ib_device *dev;
 	struct iser_data_buf *mem_copy;
 	unsigned long  cmd_data_len;
 
-	dma_device = iser_ctask->iser_conn->ib_conn->device->ib_device->dma_device;
-	mem_copy   = &iser_ctask->data_copy[cmd_dir];
+	dev = iser_ctask->iser_conn->ib_conn->device->ib_device;
+	mem_copy = &iser_ctask->data_copy[cmd_dir];
 
-	if (cmd_dir == ISER_DIR_OUT)
-		dma_unmap_sg(dma_device, &mem_copy->sg_single, 1,
-			     DMA_TO_DEVICE);
-	else
-		dma_unmap_sg(dma_device, &mem_copy->sg_single, 1,
-			     DMA_FROM_DEVICE);
+	ib_dma_unmap_sg(dev, &mem_copy->sg_single, 1,
+			(cmd_dir == ISER_DIR_OUT) ?
+			DMA_TO_DEVICE : DMA_FROM_DEVICE);
 
 	if (cmd_dir == ISER_DIR_IN) {
 		char *mem;
@@ -231,11 +223,12 @@ void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_cmd_task *iser_ctask,
  * consecutive elements. Also, it handles one entry SG.
  */
 static int iser_sg_to_page_vec(struct iser_data_buf *data,
-			       struct iser_page_vec *page_vec)
+			       struct iser_page_vec *page_vec,
+			       struct ib_device *ibdev)
 {
 	struct scatterlist *sg = (struct scatterlist *)data->buf;
-	dma_addr_t first_addr, last_addr, page;
-	int start_aligned, end_aligned;
+	u64 first_addr, last_addr, page;
+	int end_aligned;
 	unsigned int cur_page = 0;
 	unsigned long total_sz = 0;
 	int i;
@@ -244,19 +237,21 @@ static int iser_sg_to_page_vec(struct iser_data_buf *data,
 	page_vec->offset = (u64) sg[0].offset & ~MASK_4K;
 
 	for (i = 0; i < data->dma_nents; i++) {
-		total_sz += sg_dma_len(&sg[i]);
+		unsigned int dma_len = ib_sg_dma_len(ibdev, &sg[i]);
+
+		total_sz += dma_len;
 
-		first_addr = sg_dma_address(&sg[i]);
-		last_addr  = first_addr + sg_dma_len(&sg[i]);
+		first_addr = ib_sg_dma_address(ibdev, &sg[i]);
+		last_addr  = first_addr + dma_len;
 
-		start_aligned = !(first_addr & ~MASK_4K);
 		end_aligned   = !(last_addr  & ~MASK_4K);
 
 		/* continue to collect page fragments till aligned or SG ends */
 		while (!end_aligned && (i + 1 < data->dma_nents)) {
 			i++;
-			total_sz += sg_dma_len(&sg[i]);
-			last_addr = sg_dma_address(&sg[i]) + sg_dma_len(&sg[i]);
+			dma_len = ib_sg_dma_len(ibdev, &sg[i]);
+			total_sz += dma_len;
+			last_addr = ib_sg_dma_address(ibdev, &sg[i]) + dma_len;
 			end_aligned = !(last_addr  & ~MASK_4K);
 		}
 
@@ -288,10 +283,11 @@ static int iser_sg_to_page_vec(struct iser_data_buf *data,
  * the number of entries which are aligned correctly. Supports the case where
  * consecutive SG elements are actually fragments of the same physcial page.
  */
-static unsigned int iser_data_buf_aligned_len(struct iser_data_buf *data)
+static unsigned int iser_data_buf_aligned_len(struct iser_data_buf *data,
+					      struct ib_device *ibdev)
 {
 	struct scatterlist *sg;
-	dma_addr_t end_addr, next_addr;
+	u64 end_addr, next_addr;
 	int i, cnt;
 	unsigned int ret_len = 0;
 
@@ -303,12 +299,12 @@ static unsigned int iser_data_buf_aligned_len(struct iser_data_buf *data)
 		   (unsigned long)page_to_phys(sg[i].page),
 		   (unsigned long)sg[i].offset,
 		   (unsigned long)sg[i].length); */
-		end_addr = sg_dma_address(&sg[i]) +
-			   sg_dma_len(&sg[i]);
+		end_addr = ib_sg_dma_address(ibdev, &sg[i]) +
+			   ib_sg_dma_len(ibdev, &sg[i]);
 		/* iser_dbg("Checking sg iobuf end address "
 		       "0x%08lX\n", end_addr); */
 		if (i + 1 < data->dma_nents) {
-			next_addr = sg_dma_address(&sg[i+1]);
+			next_addr = ib_sg_dma_address(ibdev, &sg[i+1]);
 			/* are i, i+1 fragments of the same page? */
 			if (end_addr == next_addr)
 				continue;
@@ -325,7 +321,8 @@ static unsigned int iser_data_buf_aligned_len(struct iser_data_buf *data)
 	return ret_len;
 }
 
-static void iser_data_buf_dump(struct iser_data_buf *data)
+static void iser_data_buf_dump(struct iser_data_buf *data,
+			       struct ib_device *ibdev)
 {
 	struct scatterlist *sg = (struct scatterlist *)data->buf;
 	int i;
@@ -333,9 +330,9 @@ static void iser_data_buf_dump(struct iser_data_buf *data)
 	for (i = 0; i < data->dma_nents; i++)
 		iser_err("sg[%d] dma_addr:0x%lX page:0x%p "
 			 "off:0x%x sz:0x%x dma_len:0x%x\n",
-			 i, (unsigned long)sg_dma_address(&sg[i]),
+			 i, (unsigned long)ib_sg_dma_address(ibdev, &sg[i]),
 			 sg[i].page, sg[i].offset,
-			 sg[i].length,sg_dma_len(&sg[i]));
+			 sg[i].length, ib_sg_dma_len(ibdev, &sg[i]));
 }
 
 static void iser_dump_page_vec(struct iser_page_vec *page_vec)
@@ -349,7 +346,8 @@ static void iser_dump_page_vec(struct iser_page_vec *page_vec)
 }
 
 static void iser_page_vec_build(struct iser_data_buf *data,
-				struct iser_page_vec *page_vec)
+				struct iser_page_vec *page_vec,
+				struct ib_device *ibdev)
 {
 	int page_vec_len = 0;
 
@@ -357,14 +355,14 @@ static void iser_page_vec_build(struct iser_data_buf *data,
 	page_vec->offset = 0;
 
 	iser_dbg("Translating sg sz: %d\n", data->dma_nents);
-	page_vec_len = iser_sg_to_page_vec(data,page_vec);
+	page_vec_len = iser_sg_to_page_vec(data, page_vec, ibdev);
 	iser_dbg("sg len %d page_vec_len %d\n", data->dma_nents,page_vec_len);
 
 	page_vec->length = page_vec_len;
 
 	if (page_vec_len * SIZE_4K < page_vec->data_size) {
 		iser_err("page_vec too short to hold this SG\n");
-		iser_data_buf_dump(data);
+		iser_data_buf_dump(data, ibdev);
 		iser_dump_page_vec(page_vec);
 		BUG();
 	}
@@ -375,13 +373,12 @@ int iser_dma_map_task_data(struct iscsi_iser_cmd_task *iser_ctask,
 			    enum   iser_data_dir       iser_dir,
 			    enum   dma_data_direction  dma_dir)
 {
-	struct device *dma_device;
+	struct ib_device *dev;
 
 	iser_ctask->dir[iser_dir] = 1;
-	dma_device =
-		iser_ctask->iser_conn->ib_conn->device->ib_device->dma_device;
+	dev = iser_ctask->iser_conn->ib_conn->device->ib_device;
 
-	data->dma_nents = dma_map_sg(dma_device, data->buf, data->size, dma_dir);
+	data->dma_nents = ib_dma_map_sg(dev, data->buf, data->size, dma_dir);
 	if (data->dma_nents == 0) {
 		iser_err("dma_map_sg failed!!!\n");
 		return -EINVAL;
@@ -391,20 +388,19 @@ int iser_dma_map_task_data(struct iscsi_iser_cmd_task *iser_ctask,
 
 void iser_dma_unmap_task_data(struct iscsi_iser_cmd_task *iser_ctask)
 {
-	struct device  *dma_device;
+	struct ib_device *dev;
 	struct iser_data_buf *data;
 
-	dma_device =
-		iser_ctask->iser_conn->ib_conn->device->ib_device->dma_device;
+	dev = iser_ctask->iser_conn->ib_conn->device->ib_device;
 
 	if (iser_ctask->dir[ISER_DIR_IN]) {
 		data = &iser_ctask->data[ISER_DIR_IN];
-		dma_unmap_sg(dma_device, data->buf, data->size, DMA_FROM_DEVICE);
+		ib_dma_unmap_sg(dev, data->buf, data->size, DMA_FROM_DEVICE);
 	}
 
 	if (iser_ctask->dir[ISER_DIR_OUT]) {
 		data = &iser_ctask->data[ISER_DIR_OUT];
-		dma_unmap_sg(dma_device, data->buf, data->size, DMA_TO_DEVICE);
+		ib_dma_unmap_sg(dev, data->buf, data->size, DMA_TO_DEVICE);
 	}
 }
 
@@ -419,6 +415,7 @@ int iser_reg_rdma_mem(struct iscsi_iser_cmd_task *iser_ctask,
 {
 	struct iser_conn     *ib_conn = iser_ctask->iser_conn->ib_conn;
 	struct iser_device   *device = ib_conn->device;
+	struct ib_device     *ibdev = device->ib_device;
 	struct iser_data_buf *mem = &iser_ctask->data[cmd_dir];
 	struct iser_regd_buf *regd_buf;
 	int aligned_len;
@@ -428,11 +425,11 @@ int iser_reg_rdma_mem(struct iscsi_iser_cmd_task *iser_ctask,
 
 	regd_buf = &iser_ctask->rdma_regd[cmd_dir];
 
-	aligned_len = iser_data_buf_aligned_len(mem);
+	aligned_len = iser_data_buf_aligned_len(mem, ibdev);
 	if (aligned_len != mem->dma_nents) {
 		iser_err("rdma alignment violation %d/%d aligned\n",
 			 aligned_len, mem->size);
-		iser_data_buf_dump(mem);
+		iser_data_buf_dump(mem, ibdev);
 
 		/* unmap the command data before accessing it */
 		iser_dma_unmap_task_data(iser_ctask);
@@ -450,8 +447,8 @@ int iser_reg_rdma_mem(struct iscsi_iser_cmd_task *iser_ctask,
 
 		regd_buf->reg.lkey = device->mr->lkey;
 		regd_buf->reg.rkey = device->mr->rkey;
-		regd_buf->reg.len  = sg_dma_len(&sg[0]);
-		regd_buf->reg.va   = sg_dma_address(&sg[0]);
+		regd_buf->reg.len  = ib_sg_dma_len(ibdev, &sg[0]);
+		regd_buf->reg.va   = ib_sg_dma_address(ibdev, &sg[0]);
 		regd_buf->reg.is_fmr = 0;
 
 		iser_dbg("PHYSICAL Mem.register: lkey: 0x%08X rkey: 0x%08X  "
@@ -461,10 +458,10 @@ int iser_reg_rdma_mem(struct iscsi_iser_cmd_task *iser_ctask,
 			 (unsigned long)regd_buf->reg.va,
 			 (unsigned long)regd_buf->reg.len);
 	} else { /* use FMR for multiple dma entries */
-		iser_page_vec_build(mem, ib_conn->page_vec);
+		iser_page_vec_build(mem, ib_conn->page_vec, ibdev);
 		err = iser_reg_page_vec(ib_conn, ib_conn->page_vec, &regd_buf->reg);
 		if (err) {
-			iser_data_buf_dump(mem);
+			iser_data_buf_dump(mem, ibdev);
 			iser_err("mem->dma_nents = %d (dlength = 0x%x)\n", mem->dma_nents,
 				 ntoh24(iser_ctask->desc.iscsi_header.dlength));
 			iser_err("page_vec: data_size = 0x%x, length = %d, offset = 0x%x\n",

drivers/infiniband/ulp/srp/ib_srp.c

@@ -122,9 +122,8 @@ static struct srp_iu *srp_alloc_iu(struct srp_host *host, size_t size,
 	if (!iu->buf)
 		goto out_free_iu;
 
-	iu->dma = dma_map_single(host->dev->dev->dma_device,
-				 iu->buf, size, direction);
-	if (dma_mapping_error(iu->dma))
+	iu->dma = ib_dma_map_single(host->dev->dev, iu->buf, size, direction);
+	if (ib_dma_mapping_error(host->dev->dev, iu->dma))
 		goto out_free_buf;
 
 	iu->size      = size;
@@ -145,8 +144,7 @@ static void srp_free_iu(struct srp_host *host, struct srp_iu *iu)
 	if (!iu)
 		return;
 
-	dma_unmap_single(host->dev->dev->dma_device,
-			 iu->dma, iu->size, iu->direction);
+	ib_dma_unmap_single(host->dev->dev, iu->dma, iu->size, iu->direction);
 	kfree(iu->buf);
 	kfree(iu);
 }
@@ -482,8 +480,8 @@ static void srp_unmap_data(struct scsi_cmnd *scmnd,
 		scat  = &req->fake_sg;
 	}
 
-	dma_unmap_sg(target->srp_host->dev->dev->dma_device, scat, nents,
-		     scmnd->sc_data_direction);
+	ib_dma_unmap_sg(target->srp_host->dev->dev, scat, nents,
+			scmnd->sc_data_direction);
 }
 
 static void srp_remove_req(struct srp_target_port *target, struct srp_request *req)
@@ -595,23 +593,26 @@ static int srp_map_fmr(struct srp_target_port *target, struct scatterlist *scat,
 	int i, j;
 	int ret;
 	struct srp_device *dev = target->srp_host->dev;
+	struct ib_device *ibdev = dev->dev;
 
 	if (!dev->fmr_pool)
 		return -ENODEV;
 
-	if ((sg_dma_address(&scat[0]) & ~dev->fmr_page_mask) &&
+	if ((ib_sg_dma_address(ibdev, &scat[0]) & ~dev->fmr_page_mask) &&
 	    mellanox_workarounds && !memcmp(&target->ioc_guid, mellanox_oui, 3))
 		return -EINVAL;
 
 	len = page_cnt = 0;
 	for (i = 0; i < sg_cnt; ++i) {
-		if (sg_dma_address(&scat[i]) & ~dev->fmr_page_mask) {
+		unsigned int dma_len = ib_sg_dma_len(ibdev, &scat[i]);
+
+		if (ib_sg_dma_address(ibdev, &scat[i]) & ~dev->fmr_page_mask) {
 			if (i > 0)
 				return -EINVAL;
 			else
 				++page_cnt;
 		}
-		if ((sg_dma_address(&scat[i]) + sg_dma_len(&scat[i])) &
+		if ((ib_sg_dma_address(ibdev, &scat[i]) + dma_len) &
 		    ~dev->fmr_page_mask) {
 			if (i < sg_cnt - 1)
 				return -EINVAL;
@@ -619,7 +620,7 @@ static int srp_map_fmr(struct srp_target_port *target, struct scatterlist *scat,
 				++page_cnt;
 		}
 
-		len += sg_dma_len(&scat[i]);
+		len += dma_len;
 	}
 
 	page_cnt += len >> dev->fmr_page_shift;
@@ -631,10 +632,14 @@ static int srp_map_fmr(struct srp_target_port *target, struct scatterlist *scat,
 		return -ENOMEM;
 
 	page_cnt = 0;
-	for (i = 0; i < sg_cnt; ++i)
-		for (j = 0; j < sg_dma_len(&scat[i]); j += dev->fmr_page_size)
+	for (i = 0; i < sg_cnt; ++i) {
+		unsigned int dma_len = ib_sg_dma_len(ibdev, &scat[i]);
+
+		for (j = 0; j < dma_len; j += dev->fmr_page_size)
 			dma_pages[page_cnt++] =
-				(sg_dma_address(&scat[i]) & dev->fmr_page_mask) + j;
+				(ib_sg_dma_address(ibdev, &scat[i]) &
+				 dev->fmr_page_mask) + j;
+	}
 
 	req->fmr = ib_fmr_pool_map_phys(dev->fmr_pool,
 					dma_pages, page_cnt, io_addr);
@@ -644,7 +649,8 @@ static int srp_map_fmr(struct srp_target_port *target, struct scatterlist *scat,
 		goto out;
 	}
 
-	buf->va  = cpu_to_be64(sg_dma_address(&scat[0]) & ~dev->fmr_page_mask);
+	buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, &scat[0]) &
+			       ~dev->fmr_page_mask);
 	buf->key = cpu_to_be32(req->fmr->fmr->rkey);
 	buf->len = cpu_to_be32(len);
 
@@ -663,6 +669,8 @@ static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target,
 	struct srp_cmd *cmd = req->cmd->buf;
 	int len, nents, count;
 	u8 fmt = SRP_DATA_DESC_DIRECT;
+	struct srp_device *dev;
+	struct ib_device *ibdev;
 
 	if (!scmnd->request_buffer || scmnd->sc_data_direction == DMA_NONE)
 		return sizeof (struct srp_cmd);
@@ -687,8 +695,10 @@ static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target,
 		sg_init_one(scat, scmnd->request_buffer, scmnd->request_bufflen);
 	}
 
-	count = dma_map_sg(target->srp_host->dev->dev->dma_device,
-			   scat, nents, scmnd->sc_data_direction);
+	dev = target->srp_host->dev;
+	ibdev = dev->dev;
+
+	count = ib_dma_map_sg(ibdev, scat, nents, scmnd->sc_data_direction);
 
 	fmt = SRP_DATA_DESC_DIRECT;
 	len = sizeof (struct srp_cmd) +	sizeof (struct srp_direct_buf);
@@ -702,9 +712,9 @@ static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target,
 		 */
 		struct srp_direct_buf *buf = (void *) cmd->add_data;
 
-		buf->va  = cpu_to_be64(sg_dma_address(scat));
-		buf->key = cpu_to_be32(target->srp_host->dev->mr->rkey);
-		buf->len = cpu_to_be32(sg_dma_len(scat));
+		buf->va  = cpu_to_be64(ib_sg_dma_address(ibdev, scat));
+		buf->key = cpu_to_be32(dev->mr->rkey);
+		buf->len = cpu_to_be32(ib_sg_dma_len(ibdev, scat));
 	} else if (srp_map_fmr(target, scat, count, req,
 			       (void *) cmd->add_data)) {
 		/*
@@ -722,13 +732,14 @@ static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target,
 			count * sizeof (struct srp_direct_buf);
 
 		for (i = 0; i < count; ++i) {
+			unsigned int dma_len = ib_sg_dma_len(ibdev, &scat[i]);
+
 			buf->desc_list[i].va  =
-				cpu_to_be64(sg_dma_address(&scat[i]));
+				cpu_to_be64(ib_sg_dma_address(ibdev, &scat[i]));
 			buf->desc_list[i].key =
-				cpu_to_be32(target->srp_host->dev->mr->rkey);
-			buf->desc_list[i].len =
-				cpu_to_be32(sg_dma_len(&scat[i]));
-			datalen += sg_dma_len(&scat[i]);
+				cpu_to_be32(dev->mr->rkey);
+			buf->desc_list[i].len = cpu_to_be32(dma_len);
+			datalen += dma_len;
 		}
 
 		if (scmnd->sc_data_direction == DMA_TO_DEVICE)
@@ -808,13 +819,15 @@ static void srp_process_rsp(struct srp_target_port *target, struct srp_rsp *rsp)
 
 static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc)
 {
+	struct ib_device *dev;
 	struct srp_iu *iu;
 	u8 opcode;
 
 	iu = target->rx_ring[wc->wr_id & ~SRP_OP_RECV];
 
-	dma_sync_single_for_cpu(target->srp_host->dev->dev->dma_device, iu->dma,
-				target->max_ti_iu_len, DMA_FROM_DEVICE);
+	dev = target->srp_host->dev->dev;
+	ib_dma_sync_single_for_cpu(dev, iu->dma, target->max_ti_iu_len,
+				   DMA_FROM_DEVICE);
 
 	opcode = *(u8 *) iu->buf;
 
@@ -850,8 +863,8 @@ static void srp_handle_recv(struct srp_target_port *target, struct ib_wc *wc)
 		break;
 	}
 
-	dma_sync_single_for_device(target->srp_host->dev->dev->dma_device, iu->dma,
-				   target->max_ti_iu_len, DMA_FROM_DEVICE);
+	ib_dma_sync_single_for_device(dev, iu->dma, target->max_ti_iu_len,
+				      DMA_FROM_DEVICE);
 }
 
 static void srp_completion(struct ib_cq *cq, void *target_ptr)
@@ -969,6 +982,7 @@ static int srp_queuecommand(struct scsi_cmnd *scmnd,
 	struct srp_request *req;
 	struct srp_iu *iu;
 	struct srp_cmd *cmd;
+	struct ib_device *dev;
 	int len;
 
 	if (target->state == SRP_TARGET_CONNECTING)
@@ -985,8 +999,9 @@ static int srp_queuecommand(struct scsi_cmnd *scmnd,
 	if (!iu)
 		goto err;
 
-	dma_sync_single_for_cpu(target->srp_host->dev->dev->dma_device, iu->dma,
-				srp_max_iu_len, DMA_TO_DEVICE);
+	dev = target->srp_host->dev->dev;
+	ib_dma_sync_single_for_cpu(dev, iu->dma, srp_max_iu_len,
+				   DMA_TO_DEVICE);
 
 	req = list_entry(target->free_reqs.next, struct srp_request, list);
 
@@ -1018,8 +1033,8 @@ static int srp_queuecommand(struct scsi_cmnd *scmnd,
 		goto err_unmap;
 	}
 
-	dma_sync_single_for_device(target->srp_host->dev->dev->dma_device, iu->dma,
-				   srp_max_iu_len, DMA_TO_DEVICE);
+	ib_dma_sync_single_for_device(dev, iu->dma, srp_max_iu_len,
+				      DMA_TO_DEVICE);
 
 	if (__srp_post_send(target, iu, len)) {
 		printk(KERN_ERR PFX "Send failed\n");

drivers/infiniband/ulp/srp/ib_srp.h

@@ -161,7 +161,7 @@ struct srp_target_port {
 };
 
 struct srp_iu {
-	dma_addr_t		dma;
+	u64			dma;
 	void		       *buf;
 	size_t			size;
 	enum dma_data_direction	direction;

include/rdma/ib_marshall.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2005 Intel Corporation.  All rights reserved.
+ * Copyright (c) 2005-2006 Intel Corporation.  All rights reserved.
  *
  * This software is available to you under a choice of one of two
  * licenses.  You may choose to be licensed under the terms of the GNU
@@ -41,6 +41,9 @@
 void ib_copy_qp_attr_to_user(struct ib_uverbs_qp_attr *dst,
 			     struct ib_qp_attr *src);
 
+void ib_copy_ah_attr_to_user(struct ib_uverbs_ah_attr *dst,
+			     struct ib_ah_attr *src);
+
 void ib_copy_path_rec_to_user(struct ib_user_path_rec *dst,
 			      struct ib_sa_path_rec *src);
 

include/rdma/ib_verbs.h

@@ -43,6 +43,8 @@
 
 #include <linux/types.h>
 #include <linux/device.h>
+#include <linux/mm.h>
+#include <linux/dma-mapping.h>
 
 #include <asm/atomic.h>
 #include <asm/scatterlist.h>
@@ -848,6 +850,49 @@ struct ib_cache {
 	u8                     *lmc_cache;
 };
 
+struct ib_dma_mapping_ops {
+	int		(*mapping_error)(struct ib_device *dev,
+					 u64 dma_addr);
+	u64		(*map_single)(struct ib_device *dev,
+				      void *ptr, size_t size,
+				      enum dma_data_direction direction);
+	void		(*unmap_single)(struct ib_device *dev,
+					u64 addr, size_t size,
+					enum dma_data_direction direction);
+	u64		(*map_page)(struct ib_device *dev,
+				    struct page *page, unsigned long offset,
+				    size_t size,
+				    enum dma_data_direction direction);
+	void		(*unmap_page)(struct ib_device *dev,
+				      u64 addr, size_t size,
+				      enum dma_data_direction direction);
+	int		(*map_sg)(struct ib_device *dev,
+				  struct scatterlist *sg, int nents,
+				  enum dma_data_direction direction);
+	void		(*unmap_sg)(struct ib_device *dev,
+				    struct scatterlist *sg, int nents,
+				    enum dma_data_direction direction);
+	u64		(*dma_address)(struct ib_device *dev,
+				       struct scatterlist *sg);
+	unsigned int	(*dma_len)(struct ib_device *dev,
+				   struct scatterlist *sg);
+	void		(*sync_single_for_cpu)(struct ib_device *dev,
+					       u64 dma_handle,
+					       size_t size,
+				               enum dma_data_direction dir);
+	void		(*sync_single_for_device)(struct ib_device *dev,
+						  u64 dma_handle,
+						  size_t size,
+						  enum dma_data_direction dir);
+	void		*(*alloc_coherent)(struct ib_device *dev,
+					   size_t size,
+					   u64 *dma_handle,
+					   gfp_t flag);
+	void		(*free_coherent)(struct ib_device *dev,
+					 size_t size, void *cpu_addr,
+					 u64 dma_handle);
+};
+
 struct iw_cm_verbs;
 
 struct ib_device {
@@ -992,6 +1037,8 @@ struct ib_device {
 						  struct ib_mad *in_mad,
 						  struct ib_mad *out_mad);
 
+	struct ib_dma_mapping_ops   *dma_ops;
+
 	struct module               *owner;
 	struct class_device          class_dev;
 	struct kobject               ports_parent;
@@ -1395,9 +1442,215 @@ static inline int ib_req_ncomp_notif(struct ib_cq *cq, int wc_cnt)
  *   usable for DMA.
  * @pd: The protection domain associated with the memory region.
  * @mr_access_flags: Specifies the memory access rights.
+ *
+ * Note that the ib_dma_*() functions defined below must be used
+ * to create/destroy addresses used with the Lkey or Rkey returned
+ * by ib_get_dma_mr().
  */
 struct ib_mr *ib_get_dma_mr(struct ib_pd *pd, int mr_access_flags);
 
+/**
+ * ib_dma_mapping_error - check a DMA addr for error
+ * @dev: The device for which the dma_addr was created
+ * @dma_addr: The DMA address to check
+ */
+static inline int ib_dma_mapping_error(struct ib_device *dev, u64 dma_addr)
+{
+	return dev->dma_ops ?
+		dev->dma_ops->mapping_error(dev, dma_addr) :
+		dma_mapping_error(dma_addr);
+}
+
+/**
+ * ib_dma_map_single - Map a kernel virtual address to DMA address
+ * @dev: The device for which the dma_addr is to be created
+ * @cpu_addr: The kernel virtual address
+ * @size: The size of the region in bytes
+ * @direction: The direction of the DMA
+ */
+static inline u64 ib_dma_map_single(struct ib_device *dev,
+				    void *cpu_addr, size_t size,
+				    enum dma_data_direction direction)
+{
+	return dev->dma_ops ?
+		dev->dma_ops->map_single(dev, cpu_addr, size, direction) :
+		dma_map_single(dev->dma_device, cpu_addr, size, direction);
+}
+
+/**
+ * ib_dma_unmap_single - Destroy a mapping created by ib_dma_map_single()
+ * @dev: The device for which the DMA address was created
+ * @addr: The DMA address
+ * @size: The size of the region in bytes
+ * @direction: The direction of the DMA
+ */
+static inline void ib_dma_unmap_single(struct ib_device *dev,
+				       u64 addr, size_t size,
+				       enum dma_data_direction direction)
+{
+	dev->dma_ops ?
+		dev->dma_ops->unmap_single(dev, addr, size, direction) :
+		dma_unmap_single(dev->dma_device, addr, size, direction);
+}
+
+/**
+ * ib_dma_map_page - Map a physical page to DMA address
+ * @dev: The device for which the dma_addr is to be created
+ * @page: The page to be mapped
+ * @offset: The offset within the page
+ * @size: The size of the region in bytes
+ * @direction: The direction of the DMA
+ */
+static inline u64 ib_dma_map_page(struct ib_device *dev,
+				  struct page *page,
+				  unsigned long offset,
+				  size_t size,
+					 enum dma_data_direction direction)
+{
+	return dev->dma_ops ?
+		dev->dma_ops->map_page(dev, page, offset, size, direction) :
+		dma_map_page(dev->dma_device, page, offset, size, direction);
+}
+
+/**
+ * ib_dma_unmap_page - Destroy a mapping created by ib_dma_map_page()
+ * @dev: The device for which the DMA address was created
+ * @addr: The DMA address
+ * @size: The size of the region in bytes
+ * @direction: The direction of the DMA
+ */
+static inline void ib_dma_unmap_page(struct ib_device *dev,
+				     u64 addr, size_t size,
+				     enum dma_data_direction direction)
+{
+	dev->dma_ops ?
+		dev->dma_ops->unmap_page(dev, addr, size, direction) :
+		dma_unmap_page(dev->dma_device, addr, size, direction);
+}
+
+/**
+ * ib_dma_map_sg - Map a scatter/gather list to DMA addresses
+ * @dev: The device for which the DMA addresses are to be created
+ * @sg: The array of scatter/gather entries
+ * @nents: The number of scatter/gather entries
+ * @direction: The direction of the DMA
+ */
+static inline int ib_dma_map_sg(struct ib_device *dev,
+				struct scatterlist *sg, int nents,
+				enum dma_data_direction direction)
+{
+	return dev->dma_ops ?
+		dev->dma_ops->map_sg(dev, sg, nents, direction) :
+		dma_map_sg(dev->dma_device, sg, nents, direction);
+}
+
+/**
+ * ib_dma_unmap_sg - Unmap a scatter/gather list of DMA addresses
+ * @dev: The device for which the DMA addresses were created
+ * @sg: The array of scatter/gather entries
+ * @nents: The number of scatter/gather entries
+ * @direction: The direction of the DMA
+ */
+static inline void ib_dma_unmap_sg(struct ib_device *dev,
+				   struct scatterlist *sg, int nents,
+				   enum dma_data_direction direction)
+{
+	dev->dma_ops ?
+		dev->dma_ops->unmap_sg(dev, sg, nents, direction) :
+		dma_unmap_sg(dev->dma_device, sg, nents, direction);
+}
+
+/**
+ * ib_sg_dma_address - Return the DMA address from a scatter/gather entry
+ * @dev: The device for which the DMA addresses were created
+ * @sg: The scatter/gather entry
+ */
+static inline u64 ib_sg_dma_address(struct ib_device *dev,
+				    struct scatterlist *sg)
+{
+	return dev->dma_ops ?
+		dev->dma_ops->dma_address(dev, sg) : sg_dma_address(sg);
+}
+
+/**
+ * ib_sg_dma_len - Return the DMA length from a scatter/gather entry
+ * @dev: The device for which the DMA addresses were created
+ * @sg: The scatter/gather entry
+ */
+static inline unsigned int ib_sg_dma_len(struct ib_device *dev,
+					 struct scatterlist *sg)
+{
+	return dev->dma_ops ?
+		dev->dma_ops->dma_len(dev, sg) : sg_dma_len(sg);
+}
+
+/**
+ * ib_dma_sync_single_for_cpu - Prepare DMA region to be accessed by CPU
+ * @dev: The device for which the DMA address was created
+ * @addr: The DMA address
+ * @size: The size of the region in bytes
+ * @dir: The direction of the DMA
+ */
+static inline void ib_dma_sync_single_for_cpu(struct ib_device *dev,
+					      u64 addr,
+					      size_t size,
+					      enum dma_data_direction dir)
+{
+	dev->dma_ops ?
+		dev->dma_ops->sync_single_for_cpu(dev, addr, size, dir) :
+		dma_sync_single_for_cpu(dev->dma_device, addr, size, dir);
+}
+
+/**
+ * ib_dma_sync_single_for_device - Prepare DMA region to be accessed by device
+ * @dev: The device for which the DMA address was created
+ * @addr: The DMA address
+ * @size: The size of the region in bytes
+ * @dir: The direction of the DMA
+ */
+static inline void ib_dma_sync_single_for_device(struct ib_device *dev,
+						 u64 addr,
+						 size_t size,
+						 enum dma_data_direction dir)
+{
+	dev->dma_ops ?
+		dev->dma_ops->sync_single_for_device(dev, addr, size, dir) :
+		dma_sync_single_for_device(dev->dma_device, addr, size, dir);
+}
+
+/**
+ * ib_dma_alloc_coherent - Allocate memory and map it for DMA
+ * @dev: The device for which the DMA address is requested
+ * @size: The size of the region to allocate in bytes
+ * @dma_handle: A pointer for returning the DMA address of the region
+ * @flag: memory allocator flags
+ */
+static inline void *ib_dma_alloc_coherent(struct ib_device *dev,
+					   size_t size,
+					   u64 *dma_handle,
+					   gfp_t flag)
+{
+	return dev->dma_ops ?
+		dev->dma_ops->alloc_coherent(dev, size, dma_handle, flag) :
+		dma_alloc_coherent(dev->dma_device, size, dma_handle, flag);
+}
+
+/**
+ * ib_dma_free_coherent - Free memory allocated by ib_dma_alloc_coherent()
+ * @dev: The device for which the DMA addresses were allocated
+ * @size: The size of the region
+ * @cpu_addr: the address returned by ib_dma_alloc_coherent()
+ * @dma_handle: the DMA address returned by ib_dma_alloc_coherent()
+ */
+static inline void ib_dma_free_coherent(struct ib_device *dev,
+					size_t size, void *cpu_addr,
+					u64 dma_handle)
+{
+	dev->dma_ops ?
+		dev->dma_ops->free_coherent(dev, size, cpu_addr, dma_handle) :
+		dma_free_coherent(dev->dma_device, size, cpu_addr, dma_handle);
+}
+
 /**
  * ib_reg_phys_mr - Prepares a virtually addressed memory region for use
  *   by an HCA.

include/rdma/rdma_cm.h

@@ -77,11 +77,34 @@ struct rdma_route {
 	int num_paths;
 };
 
+struct rdma_conn_param {
+	const void *private_data;
+	u8 private_data_len;
+	u8 responder_resources;
+	u8 initiator_depth;
+	u8 flow_control;
+	u8 retry_count;		/* ignored when accepting */
+	u8 rnr_retry_count;
+	/* Fields below ignored if a QP is created on the rdma_cm_id. */
+	u8 srq;
+	u32 qp_num;
+};
+
+struct rdma_ud_param {
+	const void *private_data;
+	u8 private_data_len;
+	struct ib_ah_attr ah_attr;
+	u32 qp_num;
+	u32 qkey;
+};
+
 struct rdma_cm_event {
 	enum rdma_cm_event_type	 event;
 	int			 status;
-	void			*private_data;
-	u8			 private_data_len;
+	union {
+		struct rdma_conn_param	conn;
+		struct rdma_ud_param	ud;
+	} param;
 };
 
 struct rdma_cm_id;
@@ -204,25 +227,17 @@ void rdma_destroy_qp(struct rdma_cm_id *id);
 int rdma_init_qp_attr(struct rdma_cm_id *id, struct ib_qp_attr *qp_attr,
 		       int *qp_attr_mask);
 
-struct rdma_conn_param {
-	const void *private_data;
-	u8 private_data_len;
-	u8 responder_resources;
-	u8 initiator_depth;
-	u8 flow_control;
-	u8 retry_count;		/* ignored when accepting */
-	u8 rnr_retry_count;
-	/* Fields below ignored if a QP is created on the rdma_cm_id. */
-	u8 srq;
-	u32 qp_num;
-	enum ib_qp_type qp_type;
-};
-
 /**
  * rdma_connect - Initiate an active connection request.
+ * @id: Connection identifier to connect.
+ * @conn_param: Connection information used for connected QPs.
  *
  * Users must have resolved a route for the rdma_cm_id to connect with
  * by having called rdma_resolve_route before calling this routine.
+ *
+ * This call will either connect to a remote QP or obtain remote QP
+ * information for unconnected rdma_cm_id's.  The actual operation is
+ * based on the rdma_cm_id's port space.
  */
 int rdma_connect(struct rdma_cm_id *id, struct rdma_conn_param *conn_param);
 
@@ -252,6 +267,21 @@ int rdma_listen(struct rdma_cm_id *id, int backlog);
  */
 int rdma_accept(struct rdma_cm_id *id, struct rdma_conn_param *conn_param);
 
+/**
+ * rdma_notify - Notifies the RDMA CM of an asynchronous event that has
+ * occurred on the connection.
+ * @id: Connection identifier to transition to established.
+ * @event: Asynchronous event.
+ *
+ * This routine should be invoked by users to notify the CM of relevant
+ * communication events.  Events that should be reported to the CM and
+ * when to report them are:
+ *
+ * IB_EVENT_COMM_EST - Used when a message is received on a connected
+ *    QP before an RTU has been received.
+ */
+int rdma_notify(struct rdma_cm_id *id, enum ib_event_type event);
+
 /**
  * rdma_reject - Called to reject a connection request or response.
  */

include/rdma/rdma_cm_ib.h

@@ -44,4 +44,7 @@
 int rdma_set_ib_paths(struct rdma_cm_id *id,
 		      struct ib_sa_path_rec *path_rec, int num_paths);
 
+/* Global qkey for UD QPs and multicast groups. */
+#define RDMA_UD_QKEY 0x01234567
+
 #endif /* RDMA_CM_IB_H */

include/rdma/rdma_user_cm.h

@@ -0,0 +1,206 @@
+/*
+ * Copyright (c) 2005-2006 Intel Corporation.  All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * 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:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer 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.
+ */
+
+#ifndef RDMA_USER_CM_H
+#define RDMA_USER_CM_H
+
+#include <linux/types.h>
+#include <linux/in6.h>
+#include <rdma/ib_user_verbs.h>
+#include <rdma/ib_user_sa.h>
+
+#define RDMA_USER_CM_ABI_VERSION	3
+
+#define RDMA_MAX_PRIVATE_DATA		256
+
+enum {
+	RDMA_USER_CM_CMD_CREATE_ID,
+	RDMA_USER_CM_CMD_DESTROY_ID,
+	RDMA_USER_CM_CMD_BIND_ADDR,
+	RDMA_USER_CM_CMD_RESOLVE_ADDR,
+	RDMA_USER_CM_CMD_RESOLVE_ROUTE,
+	RDMA_USER_CM_CMD_QUERY_ROUTE,
+	RDMA_USER_CM_CMD_CONNECT,
+	RDMA_USER_CM_CMD_LISTEN,
+	RDMA_USER_CM_CMD_ACCEPT,
+	RDMA_USER_CM_CMD_REJECT,
+	RDMA_USER_CM_CMD_DISCONNECT,
+	RDMA_USER_CM_CMD_INIT_QP_ATTR,
+	RDMA_USER_CM_CMD_GET_EVENT,
+	RDMA_USER_CM_CMD_GET_OPTION,
+	RDMA_USER_CM_CMD_SET_OPTION,
+	RDMA_USER_CM_CMD_NOTIFY
+};
+
+/*
+ * command ABI structures.
+ */
+struct rdma_ucm_cmd_hdr {
+	__u32 cmd;
+	__u16 in;
+	__u16 out;
+};
+
+struct rdma_ucm_create_id {
+	__u64 uid;
+	__u64 response;
+	__u16 ps;
+	__u8  reserved[6];
+};
+
+struct rdma_ucm_create_id_resp {
+	__u32 id;
+};
+
+struct rdma_ucm_destroy_id {
+	__u64 response;
+	__u32 id;
+	__u32 reserved;
+};
+
+struct rdma_ucm_destroy_id_resp {
+	__u32 events_reported;
+};
+
+struct rdma_ucm_bind_addr {
+	__u64 response;
+	struct sockaddr_in6 addr;
+	__u32 id;
+};
+
+struct rdma_ucm_resolve_addr {
+	struct sockaddr_in6 src_addr;
+	struct sockaddr_in6 dst_addr;
+	__u32 id;
+	__u32 timeout_ms;
+};
+
+struct rdma_ucm_resolve_route {
+	__u32 id;
+	__u32 timeout_ms;
+};
+
+struct rdma_ucm_query_route {
+	__u64 response;
+	__u32 id;
+	__u32 reserved;
+};
+
+struct rdma_ucm_query_route_resp {
+	__u64 node_guid;
+	struct ib_user_path_rec ib_route[2];
+	struct sockaddr_in6 src_addr;
+	struct sockaddr_in6 dst_addr;
+	__u32 num_paths;
+	__u8 port_num;
+	__u8 reserved[3];
+};
+
+struct rdma_ucm_conn_param {
+	__u32 qp_num;
+	__u32 reserved;
+	__u8  private_data[RDMA_MAX_PRIVATE_DATA];
+	__u8  private_data_len;
+	__u8  srq;
+	__u8  responder_resources;
+	__u8  initiator_depth;
+	__u8  flow_control;
+	__u8  retry_count;
+	__u8  rnr_retry_count;
+	__u8  valid;
+};
+
+struct rdma_ucm_ud_param {
+	__u32 qp_num;
+	__u32 qkey;
+	struct ib_uverbs_ah_attr ah_attr;
+	__u8  private_data[RDMA_MAX_PRIVATE_DATA];
+	__u8  private_data_len;
+	__u8  reserved[7];
+};
+
+struct rdma_ucm_connect {
+	struct rdma_ucm_conn_param conn_param;
+	__u32 id;
+	__u32 reserved;
+};
+
+struct rdma_ucm_listen {
+	__u32 id;
+	__u32 backlog;
+};
+
+struct rdma_ucm_accept {
+	__u64 uid;
+	struct rdma_ucm_conn_param conn_param;
+	__u32 id;
+	__u32 reserved;
+};
+
+struct rdma_ucm_reject {
+	__u32 id;
+	__u8  private_data_len;
+	__u8  reserved[3];
+	__u8  private_data[RDMA_MAX_PRIVATE_DATA];
+};
+
+struct rdma_ucm_disconnect {
+	__u32 id;
+};
+
+struct rdma_ucm_init_qp_attr {
+	__u64 response;
+	__u32 id;
+	__u32 qp_state;
+};
+
+struct rdma_ucm_notify {
+	__u32 id;
+	__u32 event;
+};
+
+struct rdma_ucm_get_event {
+	__u64 response;
+};
+
+struct rdma_ucm_event_resp {
+	__u64 uid;
+	__u32 id;
+	__u32 event;
+	__u32 status;
+	union {
+		struct rdma_ucm_conn_param conn;
+		struct rdma_ucm_ud_param   ud;
+	} param;
+};
+
+#endif /* RDMA_USER_CM_H */