IB/mlx5: Fix send work queue size calculation
authorEli Cohen <eli@mellanox.com>
Wed, 11 Sep 2013 13:35:22 +0000 (16:35 +0300)
committerRoland Dreier <roland@purestorage.com>
Thu, 10 Oct 2013 16:23:55 +0000 (09:23 -0700)
1. Make sure wqe_cnt does not exceed the limit published by firmware.

2. There is no requirement that the number of outstanding work
   requests will be a power of two. Remove the ilog2 in the
   calculation of sq.max_post to fix that.

3. Add case for IB_QPT_XRC_TGT in sq_overhead and return 0 as XRC
   target QPs do not have a send queue.

Signed-off-by: Eli Cohen <eli@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
drivers/infiniband/hw/mlx5/qp.c

index 045f8cd..05d53f1 100644 (file)
@@ -203,7 +203,7 @@ static int sq_overhead(enum ib_qp_type qp_type)
 
        switch (qp_type) {
        case IB_QPT_XRC_INI:
-               size = sizeof(struct mlx5_wqe_xrc_seg);
+               size += sizeof(struct mlx5_wqe_xrc_seg);
                /* fall through */
        case IB_QPT_RC:
                size += sizeof(struct mlx5_wqe_ctrl_seg) +
@@ -211,20 +211,23 @@ static int sq_overhead(enum ib_qp_type qp_type)
                        sizeof(struct mlx5_wqe_raddr_seg);
                break;
 
+       case IB_QPT_XRC_TGT:
+               return 0;
+
        case IB_QPT_UC:
-               size = sizeof(struct mlx5_wqe_ctrl_seg) +
+               size += sizeof(struct mlx5_wqe_ctrl_seg) +
                        sizeof(struct mlx5_wqe_raddr_seg);
                break;
 
        case IB_QPT_UD:
        case IB_QPT_SMI:
        case IB_QPT_GSI:
-               size = sizeof(struct mlx5_wqe_ctrl_seg) +
+               size += sizeof(struct mlx5_wqe_ctrl_seg) +
                        sizeof(struct mlx5_wqe_datagram_seg);
                break;
 
        case MLX5_IB_QPT_REG_UMR:
-               size = sizeof(struct mlx5_wqe_ctrl_seg) +
+               size += sizeof(struct mlx5_wqe_ctrl_seg) +
                        sizeof(struct mlx5_wqe_umr_ctrl_seg) +
                        sizeof(struct mlx5_mkey_seg);
                break;
@@ -270,7 +273,8 @@ static int calc_sq_size(struct mlx5_ib_dev *dev, struct ib_qp_init_attr *attr,
                return wqe_size;
 
        if (wqe_size > dev->mdev.caps.max_sq_desc_sz) {
-               mlx5_ib_dbg(dev, "\n");
+               mlx5_ib_dbg(dev, "wqe_size(%d) > max_sq_desc_sz(%d)\n",
+                           wqe_size, dev->mdev.caps.max_sq_desc_sz);
                return -EINVAL;
        }
 
@@ -280,9 +284,15 @@ static int calc_sq_size(struct mlx5_ib_dev *dev, struct ib_qp_init_attr *attr,
 
        wq_size = roundup_pow_of_two(attr->cap.max_send_wr * wqe_size);
        qp->sq.wqe_cnt = wq_size / MLX5_SEND_WQE_BB;
+       if (qp->sq.wqe_cnt > dev->mdev.caps.max_wqes) {
+               mlx5_ib_dbg(dev, "wqe count(%d) exceeds limits(%d)\n",
+                           qp->sq.wqe_cnt, dev->mdev.caps.max_wqes);
+               return -ENOMEM;
+       }
        qp->sq.wqe_shift = ilog2(MLX5_SEND_WQE_BB);
        qp->sq.max_gs = attr->cap.max_send_sge;
-       qp->sq.max_post = 1 << ilog2(wq_size / wqe_size);
+       qp->sq.max_post = wq_size / wqe_size;
+       attr->cap.max_send_wr = qp->sq.max_post;
 
        return wq_size;
 }