IB/mlx4: Use flow counters on IBoE ports

Allocate flow counter per Ethernet/IBoE port, and attach this counter
to all the QPs created on that port.  Based on patch by Eli Cohen
<eli@mellanox.co.il>.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@purestorage.com>

4 files changed, 29 insertions, 4 deletions

diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c
index 621c409433e2..fa643f4f4e28 100644
--- a/drivers/infiniband/hw/mlx4/main.c
+++ b/drivers/infiniband/hw/mlx4/main.c
@@ -1098,11 +1098,21 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
         if (init_node_data(ibdev))
                 goto err_map;
 
+        for (i = 0; i < ibdev->num_ports; ++i) {
+                if (mlx4_ib_port_link_layer(&ibdev->ib_dev, i + 1) ==
+                                                IB_LINK_LAYER_ETHERNET) {
+                        err = mlx4_counter_alloc(ibdev->dev, &ibdev->counters[i]);
+                        if (err)
+                                ibdev->counters[i] = -1;
+                } else
+                                ibdev->counters[i] = -1;
+        }
+
         spin_lock_init(&ibdev->sm_lock);
         mutex_init(&ibdev->cap_mask_mutex);
 
         if (ib_register_device(&ibdev->ib_dev, NULL))
-                goto err_map;
+                goto err_counter;
 
         if (mlx4_ib_mad_init(ibdev))
                 goto err_reg;
@@ -1132,6 +1142,10 @@ err_notif:
 err_reg:
         ib_unregister_device(&ibdev->ib_dev);
 
+err_counter:
+        for (; i; --i)
+                mlx4_counter_free(ibdev->dev, ibdev->counters[i - 1]);
+
 err_map:
         iounmap(ibdev->uar_map);
 
@@ -1160,7 +1174,8 @@ static void mlx4_ib_remove(struct mlx4_dev *dev, void *ibdev_ptr)
                 ibdev->iboe.nb.notifier_call = NULL;
         }
         iounmap(ibdev->uar_map);
-
+        for (p = 0; p < ibdev->num_ports; ++p)
+                mlx4_counter_free(ibdev->dev, ibdev->counters[p]);
         mlx4_foreach_port(p, dev, MLX4_PORT_TYPE_IB)
                 mlx4_CLOSE_PORT(dev, p);
 
diff --git a/drivers/infiniband/hw/mlx4/mlx4_ib.h b/drivers/infiniband/hw/mlx4/mlx4_ib.h
index 2a322f21049f..e4bf2cff8662 100644
--- a/drivers/infiniband/hw/mlx4/mlx4_ib.h
+++ b/drivers/infiniband/hw/mlx4/mlx4_ib.h
@@ -193,6 +193,7 @@ struct mlx4_ib_dev {
         struct mutex            cap_mask_mutex;
         bool                    ib_active;
         struct mlx4_ib_iboe     iboe;
+        int                     counters[MLX4_MAX_PORTS];
 };
 
 static inline struct mlx4_ib_dev *to_mdev(struct ib_device *ibdev)
diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c
index 2001f20a4361..3a91d9d8dc51 100644
--- a/drivers/infiniband/hw/mlx4/qp.c
+++ b/drivers/infiniband/hw/mlx4/qp.c
@@ -893,7 +893,6 @@ static int mlx4_set_path(struct mlx4_ib_dev *dev, const struct ib_ah_attr *ah,
                         --path->static_rate;
         } else
                 path->static_rate = 0;
-        path->counter_index = 0xff;
 
         if (ah->ah_flags & IB_AH_GRH) {
                 if (ah->grh.sgid_index >= dev->dev->caps.gid_table_len[port]) {
@@ -1034,6 +1033,15 @@ static int __mlx4_ib_modify_qp(struct ib_qp *ibqp,
                 }
         }
 
+        if (cur_state == IB_QPS_INIT && new_state == IB_QPS_RTR) {
+                if (dev->counters[qp->port - 1] != -1) {
+                        context->pri_path.counter_index =
+                                                dev->counters[qp->port - 1];
+                        optpar |= MLX4_QP_OPTPAR_COUNTER_INDEX;
+                } else
+                        context->pri_path.counter_index = 0xff;
+        }
+
         if (attr_mask & IB_QP_PKEY_INDEX) {
                 context->pri_path.pkey_index = attr->pkey_index;
                 optpar |= MLX4_QP_OPTPAR_PKEY_INDEX;
diff --git a/include/linux/mlx4/qp.h b/include/linux/mlx4/qp.h
index 6486d880bdbf..4001c8249dbb 100644
--- a/include/linux/mlx4/qp.h
+++ b/include/linux/mlx4/qp.h
@@ -54,7 +54,8 @@ enum mlx4_qp_optpar {
         MLX4_QP_OPTPAR_RETRY_COUNT              = 1 << 12,
         MLX4_QP_OPTPAR_RNR_RETRY                = 1 << 13,
         MLX4_QP_OPTPAR_ACK_TIMEOUT              = 1 << 14,
-        MLX4_QP_OPTPAR_SCHED_QUEUE              = 1 << 16
+        MLX4_QP_OPTPAR_SCHED_QUEUE              = 1 << 16,
+        MLX4_QP_OPTPAR_COUNTER_INDEX            = 1 << 20
 };
 
 enum mlx4_qp_state {