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

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (62 commits)
  mlx4_core: Deprecate log_num_vlan module param
  IB/mlx4: Don't set VLAN in IBoE WQEs' control segment
  IB/mlx4: Enable 4K mtu for IBoE
  RDMA/cxgb4: Mark QP in error before disabling the queue in firmware
  RDMA/cxgb4: Serialize calls to CQ's comp_handler
  RDMA/cxgb3: Serialize calls to CQ's comp_handler
  IB/qib: Fix issue with link states and QSFP cables
  IB/mlx4: Configure extended active speeds
  mlx4_core: Add extended port capabilities support
  IB/qib: Hold links until tuning data is available
  IB/qib: Clean up checkpatch issue
  IB/qib: Remove s_lock around header validation
  IB/qib: Precompute timeout jiffies to optimize latency
  IB/qib: Use RCU for qpn lookup
  IB/qib: Eliminate divide/mod in converting idx to egr buf pointer
  IB/qib: Decode path MTU optimization
  IB/qib: Optimize RC/UC code by IB operation
  IPoIB: Use the right function to do DMA unmap pages
  RDMA/cxgb4: Use correct QID in insert_recv_cqe()
  RDMA/cxgb4: Make sure flush CQ entries are collected on connection close
  ...

4 files changed, 214 insertions, 46 deletions

diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index fa643f4f4e28..77f3dbc0aaa1 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -128,6 +128,8 @@ static int mlx4_ib_query_device(struct ib_device *ibdev,
 	    (dev->dev->caps.bmme_flags & MLX4_BMME_FLAG_REMOTE_INV) &&
 	    (dev->dev->caps.bmme_flags & MLX4_BMME_FLAG_FAST_REG_WR))
 		props->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS;
+	if (dev->dev->caps.flags & MLX4_DEV_CAP_FLAG_XRC)
+		props->device_cap_flags |= IB_DEVICE_XRC;
 
 	props->vendor_id	   = be32_to_cpup((__be32 *) (out_mad->data + 36)) &
 		0xffffff;
@@ -181,8 +183,12 @@ mlx4_ib_port_link_layer(struct ib_device *device, u8 port_num)
 
 static int ib_link_query_port(struct ib_device *ibdev, u8 port,
 			      struct ib_port_attr *props,
+			      struct ib_smp *in_mad,
 			      struct ib_smp *out_mad)
 {
+	int ext_active_speed;
+	int err;
+
 	props->lid		= be16_to_cpup((__be16 *) (out_mad->data + 16));
 	props->lmc		= out_mad->data[34] & 0x7;
 	props->sm_lid		= be16_to_cpup((__be16 *) (out_mad->data + 18));
@@ -203,6 +209,39 @@ static int ib_link_query_port(struct ib_device *ibdev, u8 port,
 	props->max_vl_num	= out_mad->data[37] >> 4;
 	props->init_type_reply	= out_mad->data[41] >> 4;
 
+	/* Check if extended speeds (EDR/FDR/...) are supported */
+	if (props->port_cap_flags & IB_PORT_EXTENDED_SPEEDS_SUP) {
+		ext_active_speed = out_mad->data[62] >> 4;
+
+		switch (ext_active_speed) {
+		case 1:
+			props->active_speed = 16; /* FDR */
+			break;
+		case 2:
+			props->active_speed = 32; /* EDR */
+			break;
+		}
+	}
+
+	/* If reported active speed is QDR, check if is FDR-10 */
+	if (props->active_speed == 4) {
+		if (to_mdev(ibdev)->dev->caps.ext_port_cap[port] &
+		    MLX_EXT_PORT_CAP_FLAG_EXTENDED_PORT_INFO) {
+			init_query_mad(in_mad);
+			in_mad->attr_id = MLX4_ATTR_EXTENDED_PORT_INFO;
+			in_mad->attr_mod = cpu_to_be32(port);
+
+			err = mlx4_MAD_IFC(to_mdev(ibdev), 1, 1, port,
+					   NULL, NULL, in_mad, out_mad);
+			if (err)
+				return err;
+
+			/* Checking LinkSpeedActive for FDR-10 */
+			if (out_mad->data[15] & 0x1)
+				props->active_speed = 8;
+		}
+	}
+
 	return 0;
 }
 
@@ -227,7 +266,7 @@ static int eth_link_query_port(struct ib_device *ibdev, u8 port,
 	props->pkey_tbl_len	= 1;
 	props->bad_pkey_cntr	= be16_to_cpup((__be16 *) (out_mad->data + 46));
 	props->qkey_viol_cntr	= be16_to_cpup((__be16 *) (out_mad->data + 48));
-	props->max_mtu		= IB_MTU_2048;
+	props->max_mtu		= IB_MTU_4096;
 	props->subnet_timeout	= 0;
 	props->max_vl_num	= out_mad->data[37] >> 4;
 	props->init_type_reply	= 0;
@@ -274,7 +313,7 @@ static int mlx4_ib_query_port(struct ib_device *ibdev, u8 port,
 		goto out;
 
 	err = mlx4_ib_port_link_layer(ibdev, port) == IB_LINK_LAYER_INFINIBAND ?
-		ib_link_query_port(ibdev, port, props, out_mad) :
+		ib_link_query_port(ibdev, port, props, in_mad, out_mad) :
 		eth_link_query_port(ibdev, port, props, out_mad);
 
 out:
@@ -566,6 +605,57 @@ static int mlx4_ib_dealloc_pd(struct ib_pd *pd)
 	return 0;
 }
 
+static struct ib_xrcd *mlx4_ib_alloc_xrcd(struct ib_device *ibdev,
+					  struct ib_ucontext *context,
+					  struct ib_udata *udata)
+{
+	struct mlx4_ib_xrcd *xrcd;
+	int err;
+
+	if (!(to_mdev(ibdev)->dev->caps.flags & MLX4_DEV_CAP_FLAG_XRC))
+		return ERR_PTR(-ENOSYS);
+
+	xrcd = kmalloc(sizeof *xrcd, GFP_KERNEL);
+	if (!xrcd)
+		return ERR_PTR(-ENOMEM);
+
+	err = mlx4_xrcd_alloc(to_mdev(ibdev)->dev, &xrcd->xrcdn);
+	if (err)
+		goto err1;
+
+	xrcd->pd = ib_alloc_pd(ibdev);
+	if (IS_ERR(xrcd->pd)) {
+		err = PTR_ERR(xrcd->pd);
+		goto err2;
+	}
+
+	xrcd->cq = ib_create_cq(ibdev, NULL, NULL, xrcd, 1, 0);
+	if (IS_ERR(xrcd->cq)) {
+		err = PTR_ERR(xrcd->cq);
+		goto err3;
+	}
+
+	return &xrcd->ibxrcd;
+
+err3:
+	ib_dealloc_pd(xrcd->pd);
+err2:
+	mlx4_xrcd_free(to_mdev(ibdev)->dev, xrcd->xrcdn);
+err1:
+	kfree(xrcd);
+	return ERR_PTR(err);
+}
+
+static int mlx4_ib_dealloc_xrcd(struct ib_xrcd *xrcd)
+{
+	ib_destroy_cq(to_mxrcd(xrcd)->cq);
+	ib_dealloc_pd(to_mxrcd(xrcd)->pd);
+	mlx4_xrcd_free(to_mdev(xrcd->device)->dev, to_mxrcd(xrcd)->xrcdn);
+	kfree(xrcd);
+
+	return 0;
+}
+
 static int add_gid_entry(struct ib_qp *ibqp, union ib_gid *gid)
 {
 	struct mlx4_ib_qp *mqp = to_mqp(ibqp);
@@ -1044,7 +1134,9 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 		(1ull << IB_USER_VERBS_CMD_CREATE_SRQ)		|
 		(1ull << IB_USER_VERBS_CMD_MODIFY_SRQ)		|
 		(1ull << IB_USER_VERBS_CMD_QUERY_SRQ)		|
-		(1ull << IB_USER_VERBS_CMD_DESTROY_SRQ);
+		(1ull << IB_USER_VERBS_CMD_DESTROY_SRQ)		|
+		(1ull << IB_USER_VERBS_CMD_CREATE_XSRQ)		|
+		(1ull << IB_USER_VERBS_CMD_OPEN_QP);
 
 	ibdev->ib_dev.query_device	= mlx4_ib_query_device;
 	ibdev->ib_dev.query_port	= mlx4_ib_query_port;
@@ -1093,6 +1185,14 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 	ibdev->ib_dev.unmap_fmr		= mlx4_ib_unmap_fmr;
 	ibdev->ib_dev.dealloc_fmr	= mlx4_ib_fmr_dealloc;
 
+	if (dev->caps.flags & MLX4_DEV_CAP_FLAG_XRC) {
+		ibdev->ib_dev.alloc_xrcd = mlx4_ib_alloc_xrcd;
+		ibdev->ib_dev.dealloc_xrcd = mlx4_ib_dealloc_xrcd;
+		ibdev->ib_dev.uverbs_cmd_mask |=
+			(1ull << IB_USER_VERBS_CMD_OPEN_XRCD) |
+			(1ull << IB_USER_VERBS_CMD_CLOSE_XRCD);
+	}
+
 	spin_lock_init(&iboe->lock);
 
 	if (init_node_data(ibdev))
diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h
index e4bf2cff8662..ed80345c99ae 100644
--- a/drivers/infiniband/hw/mlx4/mlx4_ib.h
+++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h
@@ -56,6 +56,13 @@ struct mlx4_ib_pd {
 	u32			pdn;
 };
 
+struct mlx4_ib_xrcd {
+	struct ib_xrcd		ibxrcd;
+	u32			xrcdn;
+	struct ib_pd	       *pd;
+	struct ib_cq	       *cq;
+};
+
 struct mlx4_ib_cq_buf {
 	struct mlx4_buf		buf;
 	struct mlx4_mtt		mtt;
@@ -138,6 +145,7 @@ struct mlx4_ib_qp {
 	struct mlx4_mtt		mtt;
 	int			buf_size;
 	struct mutex		mutex;
+	u16			xrcdn;
 	u32			flags;
 	u8			port;
 	u8			alt_port;
@@ -211,6 +219,11 @@ static inline struct mlx4_ib_pd *to_mpd(struct ib_pd *ibpd)
 	return container_of(ibpd, struct mlx4_ib_pd, ibpd);
 }
 
+static inline struct mlx4_ib_xrcd *to_mxrcd(struct ib_xrcd *ibxrcd)
+{
+	return container_of(ibxrcd, struct mlx4_ib_xrcd, ibxrcd);
+}
+
 static inline struct mlx4_ib_cq *to_mcq(struct ib_cq *ibcq)
 {
 	return container_of(ibcq, struct mlx4_ib_cq, ibcq);
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 3a91d9d8dc51..a16f0c8e6f3f 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -302,15 +302,14 @@ static int send_wqe_overhead(enum ib_qp_type type, u32 flags)
 }
 
 static int set_rq_size(struct mlx4_ib_dev *dev, struct ib_qp_cap *cap,
-		       int is_user, int has_srq, struct mlx4_ib_qp *qp)
+		       int is_user, int has_rq, struct mlx4_ib_qp *qp)
 {
 	/* Sanity check RQ size before proceeding */
 	if (cap->max_recv_wr  > dev->dev->caps.max_wqes  ||
 	    cap->max_recv_sge > dev->dev->caps.max_rq_sg)
 		return -EINVAL;
 
-	if (has_srq) {
-		/* QPs attached to an SRQ should have no RQ */
+	if (!has_rq) {
 		if (cap->max_recv_wr)
 			return -EINVAL;
 
@@ -463,6 +462,14 @@ static int set_user_sq_size(struct mlx4_ib_dev *dev,
 	return 0;
 }
 
+static int qp_has_rq(struct ib_qp_init_attr *attr)
+{
+	if (attr->qp_type == IB_QPT_XRC_INI || attr->qp_type == IB_QPT_XRC_TGT)
+		return 0;
+
+	return !attr->srq;
+}
+
 static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 			    struct ib_qp_init_attr *init_attr,
 			    struct ib_udata *udata, int sqpn, struct mlx4_ib_qp *qp)
@@ -479,7 +486,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 	if (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR)
 		qp->sq_signal_bits = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE);
 
-	err = set_rq_size(dev, &init_attr->cap, !!pd->uobject, !!init_attr->srq, qp);
+	err = set_rq_size(dev, &init_attr->cap, !!pd->uobject, qp_has_rq(init_attr), qp);
 	if (err)
 		goto err;
 
@@ -513,7 +520,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 		if (err)
 			goto err_mtt;
 
-		if (!init_attr->srq) {
+		if (qp_has_rq(init_attr)) {
 			err = mlx4_ib_db_map_user(to_mucontext(pd->uobject->context),
 						  ucmd.db_addr, &qp->db);
 			if (err)
@@ -532,7 +539,7 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 		if (err)
 			goto err;
 
-		if (!init_attr->srq) {
+		if (qp_has_rq(init_attr)) {
 			err = mlx4_db_alloc(dev->dev, &qp->db, 0);
 			if (err)
 				goto err;
@@ -575,6 +582,9 @@ static int create_qp_common(struct mlx4_ib_dev *dev, struct ib_pd *pd,
 	if (err)
 		goto err_qpn;
 
+	if (init_attr->qp_type == IB_QPT_XRC_TGT)
+		qp->mqp.qpn |= (1 << 23);
+
 	/*
 	 * Hardware wants QPN written in big-endian order (after
 	 * shifting) for send doorbell.  Precompute this value to save
@@ -592,9 +602,8 @@ err_qpn:
 
 err_wrid:
 	if (pd->uobject) {
-		if (!init_attr->srq)
-			mlx4_ib_db_unmap_user(to_mucontext(pd->uobject->context),
-					      &qp->db);
+		if (qp_has_rq(init_attr))
+			mlx4_ib_db_unmap_user(to_mucontext(pd->uobject->context), &qp->db);
 	} else {
 		kfree(qp->sq.wrid);
 		kfree(qp->rq.wrid);
@@ -610,7 +619,7 @@ err_buf:
 		mlx4_buf_free(dev->dev, qp->buf_size, &qp->buf);
 
 err_db:
-	if (!pd->uobject && !init_attr->srq)
+	if (!pd->uobject && qp_has_rq(init_attr))
 		mlx4_db_free(dev->dev, &qp->db);
 
 err:
@@ -671,6 +680,33 @@ static void del_gid_entries(struct mlx4_ib_qp *qp)
 	}
 }
 
+static struct mlx4_ib_pd *get_pd(struct mlx4_ib_qp *qp)
+{
+	if (qp->ibqp.qp_type == IB_QPT_XRC_TGT)
+		return to_mpd(to_mxrcd(qp->ibqp.xrcd)->pd);
+	else
+		return to_mpd(qp->ibqp.pd);
+}
+
+static void get_cqs(struct mlx4_ib_qp *qp,
+		    struct mlx4_ib_cq **send_cq, struct mlx4_ib_cq **recv_cq)
+{
+	switch (qp->ibqp.qp_type) {
+	case IB_QPT_XRC_TGT:
+		*send_cq = to_mcq(to_mxrcd(qp->ibqp.xrcd)->cq);
+		*recv_cq = *send_cq;
+		break;
+	case IB_QPT_XRC_INI:
+		*send_cq = to_mcq(qp->ibqp.send_cq);
+		*recv_cq = *send_cq;
+		break;
+	default:
+		*send_cq = to_mcq(qp->ibqp.send_cq);
+		*recv_cq = to_mcq(qp->ibqp.recv_cq);
+		break;
+	}
+}
+
 static void destroy_qp_common(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp,
 			      int is_user)
 {
@@ -682,8 +718,7 @@ static void destroy_qp_common(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp,
 			printk(KERN_WARNING "mlx4_ib: modify QP %06x to RESET failed.\n",
 			       qp->mqp.qpn);
 
-	send_cq = to_mcq(qp->ibqp.send_cq);
-	recv_cq = to_mcq(qp->ibqp.recv_cq);
+	get_cqs(qp, &send_cq, &recv_cq);
 
 	mlx4_ib_lock_cqs(send_cq, recv_cq);
 
@@ -706,7 +741,7 @@ static void destroy_qp_common(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp,
 	mlx4_mtt_cleanup(dev->dev, &qp->mtt);
 
 	if (is_user) {
-		if (!qp->ibqp.srq)
+		if (qp->rq.wqe_cnt)
 			mlx4_ib_db_unmap_user(to_mucontext(qp->ibqp.uobject->context),
 					      &qp->db);
 		ib_umem_release(qp->umem);
@@ -714,7 +749,7 @@ static void destroy_qp_common(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp,
 		kfree(qp->sq.wrid);
 		kfree(qp->rq.wrid);
 		mlx4_buf_free(dev->dev, qp->buf_size, &qp->buf);
-		if (!qp->ibqp.srq)
+		if (qp->rq.wqe_cnt)
 			mlx4_db_free(dev->dev, &qp->db);
 	}
 
@@ -725,10 +760,10 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 				struct ib_qp_init_attr *init_attr,
 				struct ib_udata *udata)
 {
-	struct mlx4_ib_dev *dev = to_mdev(pd->device);
 	struct mlx4_ib_sqp *sqp;
 	struct mlx4_ib_qp *qp;
 	int err;
+	u16 xrcdn = 0;
 
 	/*
 	 * We only support LSO and multicast loopback blocking, and
@@ -739,10 +774,20 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 		return ERR_PTR(-EINVAL);
 
 	if (init_attr->create_flags &&
-	    (pd->uobject || init_attr->qp_type != IB_QPT_UD))
+	    (udata || init_attr->qp_type != IB_QPT_UD))
 		return ERR_PTR(-EINVAL);
 
 	switch (init_attr->qp_type) {
+	case IB_QPT_XRC_TGT:
+		pd = to_mxrcd(init_attr->xrcd)->pd;
+		xrcdn = to_mxrcd(init_attr->xrcd)->xrcdn;
+		init_attr->send_cq = to_mxrcd(init_attr->xrcd)->cq;
+		/* fall through */
+	case IB_QPT_XRC_INI:
+		if (!(to_mdev(pd->device)->dev->caps.flags & MLX4_DEV_CAP_FLAG_XRC))
+			return ERR_PTR(-ENOSYS);
+		init_attr->recv_cq = init_attr->send_cq;
+		/* fall through */
 	case IB_QPT_RC:
 	case IB_QPT_UC:
 	case IB_QPT_UD:
@@ -751,13 +796,14 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 		if (!qp)
 			return ERR_PTR(-ENOMEM);
 
-		err = create_qp_common(dev, pd, init_attr, udata, 0, qp);
+		err = create_qp_common(to_mdev(pd->device), pd, init_attr, udata, 0, qp);
 		if (err) {
 			kfree(qp);
 			return ERR_PTR(err);
 		}
 
 		qp->ibqp.qp_num = qp->mqp.qpn;
+		qp->xrcdn = xrcdn;
 
 		break;
 	}
@@ -765,7 +811,7 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 	case IB_QPT_GSI:
 	{
 		/* Userspace is not allowed to create special QPs: */
-		if (pd->uobject)
+		if (udata)
 			return ERR_PTR(-EINVAL);
 
 		sqp = kzalloc(sizeof *sqp, GFP_KERNEL);
@@ -774,8 +820,8 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 
 		qp = &sqp->qp;
 
-		err = create_qp_common(dev, pd, init_attr, udata,
-				       dev->dev->caps.sqp_start +
+		err = create_qp_common(to_mdev(pd->device), pd, init_attr, udata,
+				       to_mdev(pd->device)->dev->caps.sqp_start +
 				       (init_attr->qp_type == IB_QPT_SMI ? 0 : 2) +
 				       init_attr->port_num - 1,
 				       qp);
@@ -801,11 +847,13 @@ int mlx4_ib_destroy_qp(struct ib_qp *qp)
 {
 	struct mlx4_ib_dev *dev = to_mdev(qp->device);
 	struct mlx4_ib_qp *mqp = to_mqp(qp);
+	struct mlx4_ib_pd *pd;
 
 	if (is_qp0(dev, mqp))
 		mlx4_CLOSE_PORT(dev->dev, mqp->port);
 
-	destroy_qp_common(dev, mqp, !!qp->pd->uobject);
+	pd = get_pd(mqp);
+	destroy_qp_common(dev, mqp, !!pd->ibpd.uobject);
 
 	if (is_sqp(dev, mqp))
 		kfree(to_msqp(mqp));
@@ -821,6 +869,8 @@ static int to_mlx4_st(enum ib_qp_type type)
 	case IB_QPT_RC:		return MLX4_QP_ST_RC;
 	case IB_QPT_UC:		return MLX4_QP_ST_UC;
 	case IB_QPT_UD:		return MLX4_QP_ST_UD;
+	case IB_QPT_XRC_INI:
+	case IB_QPT_XRC_TGT:	return MLX4_QP_ST_XRC;
 	case IB_QPT_SMI:
 	case IB_QPT_GSI:	return MLX4_QP_ST_MLX;
 	default:		return -1;
@@ -959,6 +1009,8 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 {
 	struct mlx4_ib_dev *dev = to_mdev(ibqp->device);
 	struct mlx4_ib_qp *qp = to_mqp(ibqp);
+	struct mlx4_ib_pd *pd;
+	struct mlx4_ib_cq *send_cq, *recv_cq;
 	struct mlx4_qp_context *context;
 	enum mlx4_qp_optpar optpar = 0;
 	int sqd_event;
@@ -1014,8 +1066,10 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 		context->sq_size_stride = ilog2(qp->sq.wqe_cnt) << 3;
 	context->sq_size_stride |= qp->sq.wqe_shift - 4;
 
-	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
+	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) {
 		context->sq_size_stride |= !!qp->sq_no_prefetch << 7;
+		context->xrcd = cpu_to_be32((u32) qp->xrcdn);
+	}
 
 	if (qp->ibqp.uobject)
 		context->usr_page = cpu_to_be32(to_mucontext(ibqp->uobject->context)->uar.index);
@@ -1079,8 +1133,12 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 		optpar |= MLX4_QP_OPTPAR_ALT_ADDR_PATH;
 	}
 
-	context->pd	    = cpu_to_be32(to_mpd(ibqp->pd)->pdn);
-	context->params1    = cpu_to_be32(MLX4_IB_ACK_REQ_FREQ << 28);
+	pd = get_pd(qp);
+	get_cqs(qp, &send_cq, &recv_cq);
+	context->pd       = cpu_to_be32(pd->pdn);
+	context->cqn_send = cpu_to_be32(send_cq->mcq.cqn);
+	context->cqn_recv = cpu_to_be32(recv_cq->mcq.cqn);
+	context->params1  = cpu_to_be32(MLX4_IB_ACK_REQ_FREQ << 28);
 
 	/* Set "fast registration enabled" for all kernel QPs */
 	if (!qp->ibqp.uobject)
@@ -1106,8 +1164,6 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 	if (attr_mask & IB_QP_SQ_PSN)
 		context->next_send_psn = cpu_to_be32(attr->sq_psn);
 
-	context->cqn_send = cpu_to_be32(to_mcq(ibqp->send_cq)->mcq.cqn);
-
 	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC) {
 		if (attr->max_dest_rd_atomic)
 			context->params2 |=
@@ -1130,8 +1186,6 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 	if (attr_mask & IB_QP_RQ_PSN)
 		context->rnr_nextrecvpsn |= cpu_to_be32(attr->rq_psn);
 
-	context->cqn_recv = cpu_to_be32(to_mcq(ibqp->recv_cq)->mcq.cqn);
-
 	if (attr_mask & IB_QP_QKEY) {
 		context->qkey = cpu_to_be32(attr->qkey);
 		optpar |= MLX4_QP_OPTPAR_Q_KEY;
@@ -1140,7 +1194,7 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 	if (ibqp->srq)
 		context->srqn = cpu_to_be32(1 << 24 | to_msrq(ibqp->srq)->msrq.srqn);
 
-	if (!ibqp->srq && cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
+	if (qp->rq.wqe_cnt && cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT)
 		context->db_rec_addr = cpu_to_be64(qp->db.dma);
 
 	if (cur_state == IB_QPS_INIT &&
@@ -1225,17 +1279,17 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
 	 * entries and reinitialize the QP.
 	 */
 	if (new_state == IB_QPS_RESET && !ibqp->uobject) {
-		mlx4_ib_cq_clean(to_mcq(ibqp->recv_cq), qp->mqp.qpn,
+		mlx4_ib_cq_clean(recv_cq, qp->mqp.qpn,
 				 ibqp->srq ? to_msrq(ibqp->srq): NULL);
-		if (ibqp->send_cq != ibqp->recv_cq)
-			mlx4_ib_cq_clean(to_mcq(ibqp->send_cq), qp->mqp.qpn, NULL);
+		if (send_cq != recv_cq)
+			mlx4_ib_cq_clean(send_cq, qp->mqp.qpn, NULL);
 
 		qp->rq.head = 0;
 		qp->rq.tail = 0;
 		qp->sq.head = 0;
 		qp->sq.tail = 0;
 		qp->sq_next_wqe = 0;
-		if (!ibqp->srq)
+		if (qp->rq.wqe_cnt)
 			*qp->db.db  = 0;
 	}
 
@@ -1547,14 +1601,13 @@ static void set_masked_atomic_seg(struct mlx4_wqe_masked_atomic_seg *aseg,
 }
 
 static void set_datagram_seg(struct mlx4_wqe_datagram_seg *dseg,
-			     struct ib_send_wr *wr, __be16 *vlan)
+			     struct ib_send_wr *wr)
 {
 	memcpy(dseg->av, &to_mah(wr->wr.ud.ah)->av, sizeof (struct mlx4_av));
 	dseg->dqpn = cpu_to_be32(wr->wr.ud.remote_qpn);
 	dseg->qkey = cpu_to_be32(wr->wr.ud.remote_qkey);
 	dseg->vlan = to_mah(wr->wr.ud.ah)->av.eth.vlan;
 	memcpy(dseg->mac, to_mah(wr->wr.ud.ah)->av.eth.mac, 6);
-	*vlan = dseg->vlan;
 }
 
 static void set_mlx_icrc_seg(void *dseg)
@@ -1657,7 +1710,6 @@ int mlx4_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 	__be32 uninitialized_var(lso_hdr_sz);
 	__be32 blh;
 	int i;
-	__be16 vlan = cpu_to_be16(0xffff);
 
 	spin_lock_irqsave(&qp->sq.lock, flags);
 
@@ -1761,7 +1813,7 @@ int mlx4_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 			break;
 
 		case IB_QPT_UD:
-			set_datagram_seg(wqe, wr, &vlan);
+			set_datagram_seg(wqe, wr);
 			wqe  += sizeof (struct mlx4_wqe_datagram_seg);
 			size += sizeof (struct mlx4_wqe_datagram_seg) / 16;
 
@@ -1824,11 +1876,6 @@ int mlx4_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 		ctrl->fence_size = (wr->send_flags & IB_SEND_FENCE ?
 				    MLX4_WQE_CTRL_FENCE : 0) | size;
 
-		if (be16_to_cpu(vlan) < 0x1000) {
-			ctrl->ins_vlan = 1 << 6;
-			ctrl->vlan_tag = vlan;
-		}
-
 		/*
 		 * Make sure descriptor is fully written before
 		 * setting ownership bit (because HW can start
diff --git a/drivers/infiniband/hw/mlx4/srq.c b/drivers/infiniband/hw/mlx4/srq.c
index 818b7ecace5e..39542f3703b8 100644
--- a/drivers/infiniband/hw/mlx4/srq.c
+++ b/drivers/infiniband/hw/mlx4/srq.c
@@ -76,6 +76,8 @@ struct ib_srq *mlx4_ib_create_srq(struct ib_pd *pd,
 	struct mlx4_ib_srq *srq;
 	struct mlx4_wqe_srq_next_seg *next;
 	struct mlx4_wqe_data_seg *scatter;
+	u32 cqn;
+	u16 xrcdn;
 	int desc_size;
 	int buf_size;
 	int err;
@@ -174,12 +176,18 @@ struct ib_srq *mlx4_ib_create_srq(struct ib_pd *pd,
 		}
 	}
 
-	err = mlx4_srq_alloc(dev->dev, to_mpd(pd)->pdn, &srq->mtt,
+	cqn = (init_attr->srq_type == IB_SRQT_XRC) ?
+		to_mcq(init_attr->ext.xrc.cq)->mcq.cqn : 0;
+	xrcdn = (init_attr->srq_type == IB_SRQT_XRC) ?
+		to_mxrcd(init_attr->ext.xrc.xrcd)->xrcdn :
+		(u16) dev->dev->caps.reserved_xrcds;
+	err = mlx4_srq_alloc(dev->dev, to_mpd(pd)->pdn, cqn, xrcdn, &srq->mtt,
 			     srq->db.dma, &srq->msrq);
 	if (err)
 		goto err_wrid;
 
 	srq->msrq.event = mlx4_ib_srq_event;
+	srq->ibsrq.ext.xrc.srq_num = srq->msrq.srqn;
 
 	if (pd->uobject)
 		if (ib_copy_to_udata(udata, &srq->msrq.srqn, sizeof (__u32))) {