MLX5_IB_UMR_OCTOWORD;
}
-static __be64 frwr_mkey_mask(void)
+static __be64 frwr_mkey_mask(bool atomic)
{
u64 result;
MLX5_MKEY_MASK_LW |
MLX5_MKEY_MASK_RR |
MLX5_MKEY_MASK_RW |
- MLX5_MKEY_MASK_A |
MLX5_MKEY_MASK_SMALL_FENCE |
MLX5_MKEY_MASK_FREE;
+ if (atomic)
+ result |= MLX5_MKEY_MASK_A;
+
return cpu_to_be64(result);
}
}
static void set_reg_umr_seg(struct mlx5_wqe_umr_ctrl_seg *umr,
- struct mlx5_ib_mr *mr, u8 flags)
+ struct mlx5_ib_mr *mr, u8 flags, bool atomic)
{
int size = (mr->ndescs + mr->meta_ndescs) * mr->desc_size;
umr->flags = flags;
umr->xlt_octowords = cpu_to_be16(get_xlt_octo(size));
- umr->mkey_mask = frwr_mkey_mask();
+ umr->mkey_mask = frwr_mkey_mask(atomic);
}
static void set_linv_umr_seg(struct mlx5_wqe_umr_ctrl_seg *umr)
{
struct mlx5_ib_mr *mr = to_mmr(wr->mr);
struct mlx5_ib_pd *pd = to_mpd(qp->ibqp.pd);
+ struct mlx5_ib_dev *dev = to_mdev(pd->ibpd.device);
int mr_list_size = (mr->ndescs + mr->meta_ndescs) * mr->desc_size;
bool umr_inline = mr_list_size <= MLX5_IB_SQ_UMR_INLINE_THRESHOLD;
+ bool atomic = wr->access & IB_ACCESS_REMOTE_ATOMIC;
u8 flags = 0;
+ if (!mlx5_ib_can_use_umr(dev, atomic)) {
+ mlx5_ib_warn(to_mdev(qp->ibqp.device),
+ "Fast update of %s for MR is disabled\n",
+ (MLX5_CAP_GEN(dev->mdev,
+ umr_modify_entity_size_disabled)) ?
+ "entity size" :
+ "atomic access");
+ return -EINVAL;
+ }
+
if (unlikely(wr->wr.send_flags & IB_SEND_INLINE)) {
mlx5_ib_warn(to_mdev(qp->ibqp.device),
"Invalid IB_SEND_INLINE send flag\n");
if (umr_inline)
flags |= MLX5_UMR_INLINE;
- set_reg_umr_seg(*seg, mr, flags);
+ set_reg_umr_seg(*seg, mr, flags, atomic);
*seg += sizeof(struct mlx5_wqe_umr_ctrl_seg);
*size += sizeof(struct mlx5_wqe_umr_ctrl_seg) / 16;
handle_post_send_edge(&qp->sq, seg, *size, cur_edge);