in hw/mlx4/qp.c [952:1247]
static int create_qp_common(struct ib_pd *pd, struct ib_qp_init_attr *init_attr,
struct ib_udata *udata, int sqpn,
struct mlx4_ib_qp *qp)
{
struct mlx4_ib_dev *dev = to_mdev(pd->device);
int qpn;
int err;
struct mlx4_ib_ucontext *context = rdma_udata_to_drv_context(
udata, struct mlx4_ib_ucontext, ibucontext);
enum mlx4_ib_qp_type qp_type = (enum mlx4_ib_qp_type) init_attr->qp_type;
struct mlx4_ib_cq *mcq;
unsigned long flags;
/* When tunneling special qps, we use a plain UD qp */
if (sqpn) {
if (mlx4_is_mfunc(dev->dev) &&
(!mlx4_is_master(dev->dev) ||
!(init_attr->create_flags & MLX4_IB_SRIOV_SQP))) {
if (init_attr->qp_type == IB_QPT_GSI)
qp_type = MLX4_IB_QPT_PROXY_GSI;
else {
if (mlx4_is_master(dev->dev) ||
qp0_enabled_vf(dev->dev, sqpn))
qp_type = MLX4_IB_QPT_PROXY_SMI_OWNER;
else
qp_type = MLX4_IB_QPT_PROXY_SMI;
}
}
qpn = sqpn;
/* add extra sg entry for tunneling */
init_attr->cap.max_recv_sge++;
} else if (init_attr->create_flags & MLX4_IB_SRIOV_TUNNEL_QP) {
struct mlx4_ib_qp_tunnel_init_attr *tnl_init =
container_of(init_attr,
struct mlx4_ib_qp_tunnel_init_attr, init_attr);
if ((tnl_init->proxy_qp_type != IB_QPT_SMI &&
tnl_init->proxy_qp_type != IB_QPT_GSI) ||
!mlx4_is_master(dev->dev))
return -EINVAL;
if (tnl_init->proxy_qp_type == IB_QPT_GSI)
qp_type = MLX4_IB_QPT_TUN_GSI;
else if (tnl_init->slave == mlx4_master_func_num(dev->dev) ||
mlx4_vf_smi_enabled(dev->dev, tnl_init->slave,
tnl_init->port))
qp_type = MLX4_IB_QPT_TUN_SMI_OWNER;
else
qp_type = MLX4_IB_QPT_TUN_SMI;
/* we are definitely in the PPF here, since we are creating
* tunnel QPs. base_tunnel_sqpn is therefore valid. */
qpn = dev->dev->phys_caps.base_tunnel_sqpn + 8 * tnl_init->slave
+ tnl_init->proxy_qp_type * 2 + tnl_init->port - 1;
sqpn = qpn;
}
if (init_attr->qp_type == IB_QPT_SMI ||
init_attr->qp_type == IB_QPT_GSI || qp_type == MLX4_IB_QPT_SMI ||
qp_type == MLX4_IB_QPT_GSI ||
(qp_type & (MLX4_IB_QPT_PROXY_SMI | MLX4_IB_QPT_PROXY_SMI_OWNER |
MLX4_IB_QPT_PROXY_GSI | MLX4_IB_QPT_TUN_SMI_OWNER))) {
qp->sqp = kzalloc(sizeof(struct mlx4_ib_sqp), GFP_KERNEL);
if (!qp->sqp)
return -ENOMEM;
}
qp->mlx4_ib_qp_type = qp_type;
spin_lock_init(&qp->sq.lock);
spin_lock_init(&qp->rq.lock);
INIT_LIST_HEAD(&qp->gid_list);
INIT_LIST_HEAD(&qp->steering_rules);
qp->state = IB_QPS_RESET;
if (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR)
qp->sq_signal_bits = cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE);
if (udata) {
struct mlx4_ib_create_qp ucmd;
size_t copy_len;
int shift;
int n;
copy_len = sizeof(struct mlx4_ib_create_qp);
if (ib_copy_from_udata(&ucmd, udata, copy_len)) {
err = -EFAULT;
goto err;
}
qp->inl_recv_sz = ucmd.inl_recv_sz;
if (init_attr->create_flags & IB_QP_CREATE_SCATTER_FCS) {
if (!(dev->dev->caps.flags &
MLX4_DEV_CAP_FLAG_FCS_KEEP)) {
pr_debug("scatter FCS is unsupported\n");
err = -EOPNOTSUPP;
goto err;
}
qp->flags |= MLX4_IB_QP_SCATTER_FCS;
}
err = set_rq_size(dev, &init_attr->cap, udata,
qp_has_rq(init_attr), qp, qp->inl_recv_sz);
if (err)
goto err;
qp->sq_no_prefetch = ucmd.sq_no_prefetch;
err = set_user_sq_size(dev, qp, &ucmd);
if (err)
goto err;
qp->umem =
ib_umem_get(pd->device, ucmd.buf_addr, qp->buf_size, 0);
if (IS_ERR(qp->umem)) {
err = PTR_ERR(qp->umem);
goto err;
}
shift = mlx4_ib_umem_calc_optimal_mtt_size(qp->umem, 0, &n);
err = mlx4_mtt_init(dev->dev, n, shift, &qp->mtt);
if (err)
goto err_buf;
err = mlx4_ib_umem_write_mtt(dev, &qp->mtt, qp->umem);
if (err)
goto err_mtt;
if (qp_has_rq(init_attr)) {
err = mlx4_ib_db_map_user(udata, ucmd.db_addr, &qp->db);
if (err)
goto err_mtt;
}
qp->mqp.usage = MLX4_RES_USAGE_USER_VERBS;
} else {
err = set_rq_size(dev, &init_attr->cap, udata,
qp_has_rq(init_attr), qp, 0);
if (err)
goto err;
qp->sq_no_prefetch = 0;
if (init_attr->create_flags & IB_QP_CREATE_IPOIB_UD_LSO)
qp->flags |= MLX4_IB_QP_LSO;
if (init_attr->create_flags & IB_QP_CREATE_NETIF_QP) {
if (dev->steering_support ==
MLX4_STEERING_MODE_DEVICE_MANAGED)
qp->flags |= MLX4_IB_QP_NETIF;
else {
err = -EINVAL;
goto err;
}
}
err = set_kernel_sq_size(dev, &init_attr->cap, qp_type, qp);
if (err)
goto err;
if (qp_has_rq(init_attr)) {
err = mlx4_db_alloc(dev->dev, &qp->db, 0);
if (err)
goto err;
*qp->db.db = 0;
}
if (mlx4_buf_alloc(dev->dev, qp->buf_size, PAGE_SIZE * 2,
&qp->buf)) {
err = -ENOMEM;
goto err_db;
}
err = mlx4_mtt_init(dev->dev, qp->buf.npages, qp->buf.page_shift,
&qp->mtt);
if (err)
goto err_buf;
err = mlx4_buf_write_mtt(dev->dev, &qp->mtt, &qp->buf);
if (err)
goto err_mtt;
qp->sq.wrid = kvmalloc_array(qp->sq.wqe_cnt,
sizeof(u64), GFP_KERNEL);
qp->rq.wrid = kvmalloc_array(qp->rq.wqe_cnt,
sizeof(u64), GFP_KERNEL);
if (!qp->sq.wrid || !qp->rq.wrid) {
err = -ENOMEM;
goto err_wrid;
}
qp->mqp.usage = MLX4_RES_USAGE_DRIVER;
}
if (sqpn) {
if (qp->mlx4_ib_qp_type & (MLX4_IB_QPT_PROXY_SMI_OWNER |
MLX4_IB_QPT_PROXY_SMI | MLX4_IB_QPT_PROXY_GSI)) {
if (alloc_proxy_bufs(pd->device, qp)) {
err = -ENOMEM;
goto err_wrid;
}
}
} else {
/* Raw packet QPNs may not have bits 6,7 set in their qp_num;
* otherwise, the WQE BlueFlame setup flow wrongly causes
* VLAN insertion. */
if (init_attr->qp_type == IB_QPT_RAW_PACKET)
err = mlx4_qp_reserve_range(dev->dev, 1, 1, &qpn,
(init_attr->cap.max_send_wr ?
MLX4_RESERVE_ETH_BF_QP : 0) |
(init_attr->cap.max_recv_wr ?
MLX4_RESERVE_A0_QP : 0),
qp->mqp.usage);
else
if (qp->flags & MLX4_IB_QP_NETIF)
err = mlx4_ib_steer_qp_alloc(dev, 1, &qpn);
else
err = mlx4_qp_reserve_range(dev->dev, 1, 1,
&qpn, 0, qp->mqp.usage);
if (err)
goto err_proxy;
}
if (init_attr->create_flags & IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK)
qp->flags |= MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK;
err = mlx4_qp_alloc(dev->dev, qpn, &qp->mqp);
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
* a little bit when posting sends.
*/
qp->doorbell_qpn = swab32(qp->mqp.qpn << 8);
qp->mqp.event = mlx4_ib_qp_event;
spin_lock_irqsave(&dev->reset_flow_resource_lock, flags);
mlx4_ib_lock_cqs(to_mcq(init_attr->send_cq),
to_mcq(init_attr->recv_cq));
/* Maintain device to QPs access, needed for further handling
* via reset flow
*/
list_add_tail(&qp->qps_list, &dev->qp_list);
/* Maintain CQ to QPs access, needed for further handling
* via reset flow
*/
mcq = to_mcq(init_attr->send_cq);
list_add_tail(&qp->cq_send_list, &mcq->send_qp_list);
mcq = to_mcq(init_attr->recv_cq);
list_add_tail(&qp->cq_recv_list, &mcq->recv_qp_list);
mlx4_ib_unlock_cqs(to_mcq(init_attr->send_cq),
to_mcq(init_attr->recv_cq));
spin_unlock_irqrestore(&dev->reset_flow_resource_lock, flags);
return 0;
err_qpn:
if (!sqpn) {
if (qp->flags & MLX4_IB_QP_NETIF)
mlx4_ib_steer_qp_free(dev, qpn, 1);
else
mlx4_qp_release_range(dev->dev, qpn, 1);
}
err_proxy:
if (qp->mlx4_ib_qp_type == MLX4_IB_QPT_PROXY_GSI)
free_proxy_bufs(pd->device, qp);
err_wrid:
if (udata) {
if (qp_has_rq(init_attr))
mlx4_ib_db_unmap_user(context, &qp->db);
} else {
kvfree(qp->sq.wrid);
kvfree(qp->rq.wrid);
}
err_mtt:
mlx4_mtt_cleanup(dev->dev, &qp->mtt);
err_buf:
if (!qp->umem)
mlx4_buf_free(dev->dev, qp->buf_size, &qp->buf);
ib_umem_release(qp->umem);
err_db:
if (!udata && qp_has_rq(init_attr))
mlx4_db_free(dev->dev, &qp->db);
err:
kfree(qp->sqp);
return err;
}