aboutsummaryrefslogtreecommitdiff
path: root/drivers/infiniband/hw/mlx4/qp.c
diff options
context:
space:
mode:
authorJack Morgenstein <jackm@dev.mellanox.co.il>2009-06-05 10:36:24 -0700
committerRoland Dreier <rolandd@cisco.com>2009-06-05 10:36:24 -0700
commit2ac6bf4ddc87c3b6b609f8fa82f6ebbffeac12f4 (patch)
tree7de468eac9f948f85faebb2f86efbfd66737d8d9 /drivers/infiniband/hw/mlx4/qp.c
parentab6bf42e2339580b5d87746d0ff4da4b1578b03e (diff)
downloadlinux-linaro-stable-2ac6bf4ddc87c3b6b609f8fa82f6ebbffeac12f4.tar.gz
IB/mlx4: Add strong ordering to local inval and fast reg work requests
The ConnectX Programmer's Reference Manual states that the "SO" bit must be set when posting Fast Register and Local Invalidate send work requests. When this bit is set, the work request will be executed only after all previous work requests on the send queue have been executed. (If the bit is not set, Fast Register and Local Invalidate WQEs may begin execution too early, which violates the defined semantics for these operations) This fixes the issue with NFS/RDMA reported in <http://lists.openfabrics.org/pipermail/general/2009-April/059253.html> Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il> Cc: <stable@kernel.org> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Diffstat (limited to 'drivers/infiniband/hw/mlx4/qp.c')
-rw-r--r--drivers/infiniband/hw/mlx4/qp.c4
1 files changed, 4 insertions, 0 deletions
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 20724aee76f4..c4a02648c8af 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -1585,12 +1585,16 @@ int mlx4_ib_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
break;
case IB_WR_LOCAL_INV:
+ ctrl->srcrb_flags |=
+ cpu_to_be32(MLX4_WQE_CTRL_STRONG_ORDER);
set_local_inv_seg(wqe, wr->ex.invalidate_rkey);
wqe += sizeof (struct mlx4_wqe_local_inval_seg);
size += sizeof (struct mlx4_wqe_local_inval_seg) / 16;
break;
case IB_WR_FAST_REG_MR:
+ ctrl->srcrb_flags |=
+ cpu_to_be32(MLX4_WQE_CTRL_STRONG_ORDER);
set_fmr_seg(wqe, wr);
wqe += sizeof (struct mlx4_wqe_fmr_seg);
size += sizeof (struct mlx4_wqe_fmr_seg) / 16;