Esempio n. 1
0
/*
 * This needs to be very careful to not leave IS_ERR pointers around for
 * cleanup to trip over.
 */
static int rds_ib_setup_qp(struct rds_connection *conn)
{
	struct rds_ib_connection *ic = conn->c_transport_data;
	struct ib_device *dev = ic->i_cm_id->device;
	struct ib_qp_init_attr attr;
	struct rds_ib_device *rds_ibdev;
	int ret;

	/* rds_ib_add_one creates a rds_ib_device object per IB device,
	 * and allocates a protection domain, memory range and FMR pool
	 * for each.  If that fails for any reason, it will not register
	 * the rds_ibdev at all.
	 */
	rds_ibdev = ib_get_client_data(dev, &rds_ib_client);
	if (rds_ibdev == NULL) {
		if (printk_ratelimit())
			printk(KERN_NOTICE "RDS/IB: No client_data for device %s\n",
					dev->name);
		return -EOPNOTSUPP;
	}

	if (rds_ibdev->max_wrs < ic->i_send_ring.w_nr + 1)
		rds_ib_ring_resize(&ic->i_send_ring, rds_ibdev->max_wrs - 1);
	if (rds_ibdev->max_wrs < ic->i_recv_ring.w_nr + 1)
		rds_ib_ring_resize(&ic->i_recv_ring, rds_ibdev->max_wrs - 1);

	/* Protection domain and memory range */
	ic->i_pd = rds_ibdev->pd;
	ic->i_mr = rds_ibdev->mr;

	ic->i_send_cq = ib_create_cq(dev, rds_ib_send_cq_comp_handler,
				     rds_ib_cq_event_handler, conn,
				     ic->i_send_ring.w_nr + 1, 0);
	if (IS_ERR(ic->i_send_cq)) {
		ret = PTR_ERR(ic->i_send_cq);
		ic->i_send_cq = NULL;
		rdsdebug("ib_create_cq send failed: %d\n", ret);
		goto out;
	}

	ic->i_recv_cq = ib_create_cq(dev, rds_ib_recv_cq_comp_handler,
				     rds_ib_cq_event_handler, conn,
				     ic->i_recv_ring.w_nr, 0);
	if (IS_ERR(ic->i_recv_cq)) {
		ret = PTR_ERR(ic->i_recv_cq);
		ic->i_recv_cq = NULL;
		rdsdebug("ib_create_cq recv failed: %d\n", ret);
		goto out;
	}

	ret = ib_req_notify_cq(ic->i_send_cq, IB_CQ_NEXT_COMP);
	if (ret) {
		rdsdebug("ib_req_notify_cq send failed: %d\n", ret);
		goto out;
	}

	ret = ib_req_notify_cq(ic->i_recv_cq, IB_CQ_SOLICITED);
	if (ret) {
		rdsdebug("ib_req_notify_cq recv failed: %d\n", ret);
		goto out;
	}

	/* XXX negotiate max send/recv with remote? */
	memset(&attr, 0, sizeof(attr));
	attr.event_handler = rds_ib_qp_event_handler;
	attr.qp_context = conn;
	/* + 1 to allow for the single ack message */
	attr.cap.max_send_wr = ic->i_send_ring.w_nr + 1;
	attr.cap.max_recv_wr = ic->i_recv_ring.w_nr + 1;
	attr.cap.max_send_sge = rds_ibdev->max_sge;
	attr.cap.max_recv_sge = RDS_IB_RECV_SGE;
	attr.sq_sig_type = IB_SIGNAL_REQ_WR;
	attr.qp_type = IB_QPT_RC;
	attr.send_cq = ic->i_send_cq;
	attr.recv_cq = ic->i_recv_cq;

	/*
	 * XXX this can fail if max_*_wr is too large?  Are we supposed
	 * to back off until we get a value that the hardware can support?
	 */
	ret = rdma_create_qp(ic->i_cm_id, ic->i_pd, &attr);
	if (ret) {
		rdsdebug("rdma_create_qp failed: %d\n", ret);
		goto out;
	}

	ic->i_send_hdrs = ib_dma_alloc_coherent(dev,
					   ic->i_send_ring.w_nr *
						sizeof(struct rds_header),
					   &ic->i_send_hdrs_dma, GFP_KERNEL);
	if (ic->i_send_hdrs == NULL) {
		ret = -ENOMEM;
		rdsdebug("ib_dma_alloc_coherent send failed\n");
		goto out;
	}

	ic->i_recv_hdrs = ib_dma_alloc_coherent(dev,
					   ic->i_recv_ring.w_nr *
						sizeof(struct rds_header),
					   &ic->i_recv_hdrs_dma, GFP_KERNEL);
	if (ic->i_recv_hdrs == NULL) {
		ret = -ENOMEM;
		rdsdebug("ib_dma_alloc_coherent recv failed\n");
		goto out;
	}

	ic->i_ack = ib_dma_alloc_coherent(dev, sizeof(struct rds_header),
				       &ic->i_ack_dma, GFP_KERNEL);
	if (ic->i_ack == NULL) {
		ret = -ENOMEM;
		rdsdebug("ib_dma_alloc_coherent ack failed\n");
		goto out;
	}

	ic->i_sends = vmalloc(ic->i_send_ring.w_nr * sizeof(struct rds_ib_send_work));
	if (ic->i_sends == NULL) {
		ret = -ENOMEM;
		rdsdebug("send allocation failed\n");
		goto out;
	}
	memset(ic->i_sends, 0, ic->i_send_ring.w_nr * sizeof(struct rds_ib_send_work));

	ic->i_recvs = vmalloc(ic->i_recv_ring.w_nr * sizeof(struct rds_ib_recv_work));
	if (ic->i_recvs == NULL) {
		ret = -ENOMEM;
		rdsdebug("recv allocation failed\n");
		goto out;
	}
	memset(ic->i_recvs, 0, ic->i_recv_ring.w_nr * sizeof(struct rds_ib_recv_work));

	rds_ib_recv_init_ack(ic);

	rdsdebug("conn %p pd %p mr %p cq %p %p\n", conn, ic->i_pd, ic->i_mr,
		 ic->i_send_cq, ic->i_recv_cq);

out:
	return ret;
}
Esempio n. 2
0
/*
 * This needs to be very careful to not leave IS_ERR pointers around for
 * cleanup to trip over.
 */
static int rds_ib_setup_qp(struct rds_connection *conn)
{
	struct rds_ib_connection *ic = conn->c_transport_data;
	struct ib_device *dev = ic->i_cm_id->device;
	struct ib_qp_init_attr attr;
	struct ib_cq_init_attr cq_attr = {};
	struct rds_ib_device *rds_ibdev;
	int ret, fr_queue_space;

	/*
	 * It's normal to see a null device if an incoming connection races
	 * with device removal, so we don't print a warning.
	 */
	rds_ibdev = rds_ib_get_client_data(dev);
	if (!rds_ibdev)
		return -EOPNOTSUPP;

	/* The fr_queue_space is currently set to 512, to add extra space on
	 * completion queue and send queue. This extra space is used for FRMR
	 * registration and invalidation work requests
	 */
	fr_queue_space = (rds_ibdev->use_fastreg ? RDS_IB_DEFAULT_FR_WR : 0);

	/* add the conn now so that connection establishment has the dev */
	rds_ib_add_conn(rds_ibdev, conn);

	if (rds_ibdev->max_wrs < ic->i_send_ring.w_nr + 1)
		rds_ib_ring_resize(&ic->i_send_ring, rds_ibdev->max_wrs - 1);
	if (rds_ibdev->max_wrs < ic->i_recv_ring.w_nr + 1)
		rds_ib_ring_resize(&ic->i_recv_ring, rds_ibdev->max_wrs - 1);

	/* Protection domain and memory range */
	ic->i_pd = rds_ibdev->pd;

	cq_attr.cqe = ic->i_send_ring.w_nr + fr_queue_space + 1;

	ic->i_send_cq = ib_create_cq(dev, rds_ib_cq_comp_handler_send,
				     rds_ib_cq_event_handler, conn,
				     &cq_attr);
	if (IS_ERR(ic->i_send_cq)) {
		ret = PTR_ERR(ic->i_send_cq);
		ic->i_send_cq = NULL;
		rdsdebug("ib_create_cq send failed: %d\n", ret);
		goto out;
	}

	cq_attr.cqe = ic->i_recv_ring.w_nr;
	ic->i_recv_cq = ib_create_cq(dev, rds_ib_cq_comp_handler_recv,
				     rds_ib_cq_event_handler, conn,
				     &cq_attr);
	if (IS_ERR(ic->i_recv_cq)) {
		ret = PTR_ERR(ic->i_recv_cq);
		ic->i_recv_cq = NULL;
		rdsdebug("ib_create_cq recv failed: %d\n", ret);
		goto out;
	}

	ret = ib_req_notify_cq(ic->i_send_cq, IB_CQ_NEXT_COMP);
	if (ret) {
		rdsdebug("ib_req_notify_cq send failed: %d\n", ret);
		goto out;
	}

	ret = ib_req_notify_cq(ic->i_recv_cq, IB_CQ_SOLICITED);
	if (ret) {
		rdsdebug("ib_req_notify_cq recv failed: %d\n", ret);
		goto out;
	}

	/* XXX negotiate max send/recv with remote? */
	memset(&attr, 0, sizeof(attr));
	attr.event_handler = rds_ib_qp_event_handler;
	attr.qp_context = conn;
	/* + 1 to allow for the single ack message */
	attr.cap.max_send_wr = ic->i_send_ring.w_nr + fr_queue_space + 1;
	attr.cap.max_recv_wr = ic->i_recv_ring.w_nr + 1;
	attr.cap.max_send_sge = rds_ibdev->max_sge;
	attr.cap.max_recv_sge = RDS_IB_RECV_SGE;
	attr.sq_sig_type = IB_SIGNAL_REQ_WR;
	attr.qp_type = IB_QPT_RC;
	attr.send_cq = ic->i_send_cq;
	attr.recv_cq = ic->i_recv_cq;
	atomic_set(&ic->i_fastreg_wrs, RDS_IB_DEFAULT_FR_WR);

	/*
	 * XXX this can fail if max_*_wr is too large?  Are we supposed
	 * to back off until we get a value that the hardware can support?
	 */
	ret = rdma_create_qp(ic->i_cm_id, ic->i_pd, &attr);
	if (ret) {
		rdsdebug("rdma_create_qp failed: %d\n", ret);
		goto out;
	}

	ic->i_send_hdrs = ib_dma_alloc_coherent(dev,
					   ic->i_send_ring.w_nr *
						sizeof(struct rds_header),
					   &ic->i_send_hdrs_dma, GFP_KERNEL);
	if (!ic->i_send_hdrs) {
		ret = -ENOMEM;
		rdsdebug("ib_dma_alloc_coherent send failed\n");
		goto out;
	}

	ic->i_recv_hdrs = ib_dma_alloc_coherent(dev,
					   ic->i_recv_ring.w_nr *
						sizeof(struct rds_header),
					   &ic->i_recv_hdrs_dma, GFP_KERNEL);
	if (!ic->i_recv_hdrs) {
		ret = -ENOMEM;
		rdsdebug("ib_dma_alloc_coherent recv failed\n");
		goto out;
	}

	ic->i_ack = ib_dma_alloc_coherent(dev, sizeof(struct rds_header),
				       &ic->i_ack_dma, GFP_KERNEL);
	if (!ic->i_ack) {
		ret = -ENOMEM;
		rdsdebug("ib_dma_alloc_coherent ack failed\n");
		goto out;
	}

	ic->i_sends = vzalloc_node(ic->i_send_ring.w_nr * sizeof(struct rds_ib_send_work),
				   ibdev_to_node(dev));
	if (!ic->i_sends) {
		ret = -ENOMEM;
		rdsdebug("send allocation failed\n");
		goto out;
	}

	ic->i_recvs = vzalloc_node(ic->i_recv_ring.w_nr * sizeof(struct rds_ib_recv_work),
				   ibdev_to_node(dev));
	if (!ic->i_recvs) {
		ret = -ENOMEM;
		rdsdebug("recv allocation failed\n");
		goto out;
	}

	rds_ib_recv_init_ack(ic);

	rdsdebug("conn %p pd %p cq %p %p\n", conn, ic->i_pd,
		 ic->i_send_cq, ic->i_recv_cq);

out:
	rds_ib_dev_put(rds_ibdev);
	return ret;
}
static int rds_ib_setup_qp(struct rds_connection *conn)
{
	struct rds_ib_connection *ic = conn->c_transport_data;
	struct ib_device *dev = ic->i_cm_id->device;
	struct ib_qp_init_attr attr;
	struct rds_ib_device *rds_ibdev;
	int ret;

	rds_ibdev = rds_ib_get_client_data(dev);
	if (!rds_ibdev)
		return -EOPNOTSUPP;

	
	rds_ib_add_conn(rds_ibdev, conn);

	if (rds_ibdev->max_wrs < ic->i_send_ring.w_nr + 1)
		rds_ib_ring_resize(&ic->i_send_ring, rds_ibdev->max_wrs - 1);
	if (rds_ibdev->max_wrs < ic->i_recv_ring.w_nr + 1)
		rds_ib_ring_resize(&ic->i_recv_ring, rds_ibdev->max_wrs - 1);

	
	ic->i_pd = rds_ibdev->pd;
	ic->i_mr = rds_ibdev->mr;

	ic->i_send_cq = ib_create_cq(dev, rds_ib_send_cq_comp_handler,
				     rds_ib_cq_event_handler, conn,
				     ic->i_send_ring.w_nr + 1, 0);
	if (IS_ERR(ic->i_send_cq)) {
		ret = PTR_ERR(ic->i_send_cq);
		ic->i_send_cq = NULL;
		rdsdebug("ib_create_cq send failed: %d\n", ret);
		goto out;
	}

	ic->i_recv_cq = ib_create_cq(dev, rds_ib_recv_cq_comp_handler,
				     rds_ib_cq_event_handler, conn,
				     ic->i_recv_ring.w_nr, 0);
	if (IS_ERR(ic->i_recv_cq)) {
		ret = PTR_ERR(ic->i_recv_cq);
		ic->i_recv_cq = NULL;
		rdsdebug("ib_create_cq recv failed: %d\n", ret);
		goto out;
	}

	ret = ib_req_notify_cq(ic->i_send_cq, IB_CQ_NEXT_COMP);
	if (ret) {
		rdsdebug("ib_req_notify_cq send failed: %d\n", ret);
		goto out;
	}

	ret = ib_req_notify_cq(ic->i_recv_cq, IB_CQ_SOLICITED);
	if (ret) {
		rdsdebug("ib_req_notify_cq recv failed: %d\n", ret);
		goto out;
	}

	
	memset(&attr, 0, sizeof(attr));
	attr.event_handler = rds_ib_qp_event_handler;
	attr.qp_context = conn;
	
	attr.cap.max_send_wr = ic->i_send_ring.w_nr + 1;
	attr.cap.max_recv_wr = ic->i_recv_ring.w_nr + 1;
	attr.cap.max_send_sge = rds_ibdev->max_sge;
	attr.cap.max_recv_sge = RDS_IB_RECV_SGE;
	attr.sq_sig_type = IB_SIGNAL_REQ_WR;
	attr.qp_type = IB_QPT_RC;
	attr.send_cq = ic->i_send_cq;
	attr.recv_cq = ic->i_recv_cq;

	ret = rdma_create_qp(ic->i_cm_id, ic->i_pd, &attr);
	if (ret) {
		rdsdebug("rdma_create_qp failed: %d\n", ret);
		goto out;
	}

	ic->i_send_hdrs = ib_dma_alloc_coherent(dev,
					   ic->i_send_ring.w_nr *
						sizeof(struct rds_header),
					   &ic->i_send_hdrs_dma, GFP_KERNEL);
	if (!ic->i_send_hdrs) {
		ret = -ENOMEM;
		rdsdebug("ib_dma_alloc_coherent send failed\n");
		goto out;
	}

	ic->i_recv_hdrs = ib_dma_alloc_coherent(dev,
					   ic->i_recv_ring.w_nr *
						sizeof(struct rds_header),
					   &ic->i_recv_hdrs_dma, GFP_KERNEL);
	if (!ic->i_recv_hdrs) {
		ret = -ENOMEM;
		rdsdebug("ib_dma_alloc_coherent recv failed\n");
		goto out;
	}

	ic->i_ack = ib_dma_alloc_coherent(dev, sizeof(struct rds_header),
				       &ic->i_ack_dma, GFP_KERNEL);
	if (!ic->i_ack) {
		ret = -ENOMEM;
		rdsdebug("ib_dma_alloc_coherent ack failed\n");
		goto out;
	}

	ic->i_sends = vzalloc_node(ic->i_send_ring.w_nr * sizeof(struct rds_ib_send_work),
				   ibdev_to_node(dev));
	if (!ic->i_sends) {
		ret = -ENOMEM;
		rdsdebug("send allocation failed\n");
		goto out;
	}

	ic->i_recvs = vzalloc_node(ic->i_recv_ring.w_nr * sizeof(struct rds_ib_recv_work),
				   ibdev_to_node(dev));
	if (!ic->i_recvs) {
		ret = -ENOMEM;
		rdsdebug("recv allocation failed\n");
		goto out;
	}

	rds_ib_recv_init_ack(ic);

	rdsdebug("conn %p pd %p mr %p cq %p %p\n", conn, ic->i_pd, ic->i_mr,
		 ic->i_send_cq, ic->i_recv_cq);

out:
	rds_ib_dev_put(rds_ibdev);
	return ret;
}