提交 4931c6ef 编写于 作者: S Saeed Mahameed 提交者: David S. Miller

net/mlx4_en: Optimized single ring steering

Avoid touching RX QP RSS context when loading with only
one RX ring, to allow optimized A0 RX steering.

Enable by:
- loading mlx4_core with module param: log_num_mgm_entry_size = -6.
- then: ethtool -L <interface> rx 1

Performance tests:
Tested on ConnectX3Pro, Intel(R) Xeon(R) CPU E5-2680 v3 @ 2.50GHz

XDP_DROP packet rate:
-------------------------------------
     | Before    | After     | Gain |
IPv4 | 20.5 Mpps | 28.1 Mpps |  37% |
IPv6 | 18.4 Mpps | 28.1 Mpps |  53% |
-------------------------------------
Signed-off-by: NSaeed Mahameed <saeedm@mellanox.com>
Signed-off-by: NTariq Toukan <tariqt@mellanox.com>
Cc: kernel-team@fb.com
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>
上级 cf97050d
...@@ -125,9 +125,9 @@ void mlx4_en_update_loopback_state(struct net_device *dev, ...@@ -125,9 +125,9 @@ void mlx4_en_update_loopback_state(struct net_device *dev,
priv->flags |= MLX4_EN_FLAG_ENABLE_HW_LOOPBACK; priv->flags |= MLX4_EN_FLAG_ENABLE_HW_LOOPBACK;
mutex_lock(&priv->mdev->state_lock); mutex_lock(&priv->mdev->state_lock);
if (priv->mdev->dev->caps.flags2 & if ((priv->mdev->dev->caps.flags2 &
MLX4_DEV_CAP_FLAG2_UPDATE_QP_SRC_CHECK_LB && MLX4_DEV_CAP_FLAG2_UPDATE_QP_SRC_CHECK_LB) &&
priv->rss_map.indir_qp.qpn) { priv->rss_map.indir_qp && priv->rss_map.indir_qp->qpn) {
int i; int i;
int err = 0; int err = 0;
int loopback = !!(features & NETIF_F_LOOPBACK); int loopback = !!(features & NETIF_F_LOOPBACK);
......
...@@ -596,6 +596,8 @@ static int mlx4_en_get_qp(struct mlx4_en_priv *priv) ...@@ -596,6 +596,8 @@ static int mlx4_en_get_qp(struct mlx4_en_priv *priv)
return err; return err;
} }
en_info(priv, "Steering Mode %d\n", dev->caps.steering_mode);
if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) { if (dev->caps.steering_mode == MLX4_STEERING_MODE_A0) {
int base_qpn = mlx4_get_base_qpn(dev, priv->port); int base_qpn = mlx4_get_base_qpn(dev, priv->port);
*qpn = base_qpn + index; *qpn = base_qpn + index;
...@@ -1010,7 +1012,7 @@ static void mlx4_en_do_multicast(struct mlx4_en_priv *priv, ...@@ -1010,7 +1012,7 @@ static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
memcpy(&mc_list[10], mclist->addr, ETH_ALEN); memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
mc_list[5] = priv->port; mc_list[5] = priv->port;
err = mlx4_multicast_detach(mdev->dev, err = mlx4_multicast_detach(mdev->dev,
&priv->rss_map.indir_qp, priv->rss_map.indir_qp,
mc_list, mc_list,
MLX4_PROT_ETH, MLX4_PROT_ETH,
mclist->reg_id); mclist->reg_id);
...@@ -1032,7 +1034,7 @@ static void mlx4_en_do_multicast(struct mlx4_en_priv *priv, ...@@ -1032,7 +1034,7 @@ static void mlx4_en_do_multicast(struct mlx4_en_priv *priv,
/* needed for B0 steering support */ /* needed for B0 steering support */
mc_list[5] = priv->port; mc_list[5] = priv->port;
err = mlx4_multicast_attach(mdev->dev, err = mlx4_multicast_attach(mdev->dev,
&priv->rss_map.indir_qp, priv->rss_map.indir_qp,
mc_list, mc_list,
priv->port, 0, priv->port, 0,
MLX4_PROT_ETH, MLX4_PROT_ETH,
...@@ -1742,7 +1744,7 @@ int mlx4_en_start_port(struct net_device *dev) ...@@ -1742,7 +1744,7 @@ int mlx4_en_start_port(struct net_device *dev)
/* Attach rx QP to bradcast address */ /* Attach rx QP to bradcast address */
eth_broadcast_addr(&mc_list[10]); eth_broadcast_addr(&mc_list[10]);
mc_list[5] = priv->port; /* needed for B0 steering support */ mc_list[5] = priv->port; /* needed for B0 steering support */
if (mlx4_multicast_attach(mdev->dev, &priv->rss_map.indir_qp, mc_list, if (mlx4_multicast_attach(mdev->dev, priv->rss_map.indir_qp, mc_list,
priv->port, 0, MLX4_PROT_ETH, priv->port, 0, MLX4_PROT_ETH,
&priv->broadcast_id)) &priv->broadcast_id))
mlx4_warn(mdev, "Failed Attaching Broadcast\n"); mlx4_warn(mdev, "Failed Attaching Broadcast\n");
...@@ -1866,12 +1868,12 @@ void mlx4_en_stop_port(struct net_device *dev, int detach) ...@@ -1866,12 +1868,12 @@ void mlx4_en_stop_port(struct net_device *dev, int detach)
/* Detach All multicasts */ /* Detach All multicasts */
eth_broadcast_addr(&mc_list[10]); eth_broadcast_addr(&mc_list[10]);
mc_list[5] = priv->port; /* needed for B0 steering support */ mc_list[5] = priv->port; /* needed for B0 steering support */
mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mc_list, mlx4_multicast_detach(mdev->dev, priv->rss_map.indir_qp, mc_list,
MLX4_PROT_ETH, priv->broadcast_id); MLX4_PROT_ETH, priv->broadcast_id);
list_for_each_entry(mclist, &priv->curr_list, list) { list_for_each_entry(mclist, &priv->curr_list, list) {
memcpy(&mc_list[10], mclist->addr, ETH_ALEN); memcpy(&mc_list[10], mclist->addr, ETH_ALEN);
mc_list[5] = priv->port; mc_list[5] = priv->port;
mlx4_multicast_detach(mdev->dev, &priv->rss_map.indir_qp, mlx4_multicast_detach(mdev->dev, priv->rss_map.indir_qp,
mc_list, MLX4_PROT_ETH, mclist->reg_id); mc_list, MLX4_PROT_ETH, mclist->reg_id);
if (mclist->tunnel_reg_id) if (mclist->tunnel_reg_id)
mlx4_flow_detach(mdev->dev, mclist->tunnel_reg_id); mlx4_flow_detach(mdev->dev, mclist->tunnel_reg_id);
......
...@@ -1099,11 +1099,14 @@ int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv) ...@@ -1099,11 +1099,14 @@ int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv)
int i, qpn; int i, qpn;
int err = 0; int err = 0;
int good_qps = 0; int good_qps = 0;
u8 flags;
en_dbg(DRV, priv, "Configuring rss steering\n"); en_dbg(DRV, priv, "Configuring rss steering\n");
flags = priv->rx_ring_num == 1 ? MLX4_RESERVE_A0_QP : 0;
err = mlx4_qp_reserve_range(mdev->dev, priv->rx_ring_num, err = mlx4_qp_reserve_range(mdev->dev, priv->rx_ring_num,
priv->rx_ring_num, priv->rx_ring_num,
&rss_map->base_qpn, 0); &rss_map->base_qpn, flags);
if (err) { if (err) {
en_err(priv, "Failed reserving %d qps\n", priv->rx_ring_num); en_err(priv, "Failed reserving %d qps\n", priv->rx_ring_num);
return err; return err;
...@@ -1120,13 +1123,28 @@ int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv) ...@@ -1120,13 +1123,28 @@ int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv)
++good_qps; ++good_qps;
} }
if (priv->rx_ring_num == 1) {
rss_map->indir_qp = &rss_map->qps[0];
priv->base_qpn = rss_map->indir_qp->qpn;
en_info(priv, "Optimized Non-RSS steering\n");
return 0;
}
rss_map->indir_qp = kzalloc(sizeof(*rss_map->indir_qp), GFP_KERNEL);
if (!rss_map->indir_qp) {
err = -ENOMEM;
goto rss_err;
}
/* Configure RSS indirection qp */ /* Configure RSS indirection qp */
err = mlx4_qp_alloc(mdev->dev, priv->base_qpn, &rss_map->indir_qp, GFP_KERNEL); err = mlx4_qp_alloc(mdev->dev, priv->base_qpn, rss_map->indir_qp,
GFP_KERNEL);
if (err) { if (err) {
en_err(priv, "Failed to allocate RSS indirection QP\n"); en_err(priv, "Failed to allocate RSS indirection QP\n");
goto rss_err; goto rss_err;
} }
rss_map->indir_qp.event = mlx4_en_sqp_event;
rss_map->indir_qp->event = mlx4_en_sqp_event;
mlx4_en_fill_qp_context(priv, 0, 0, 0, 1, priv->base_qpn, mlx4_en_fill_qp_context(priv, 0, 0, 0, 1, priv->base_qpn,
priv->rx_ring[0]->cqn, -1, &context); priv->rx_ring[0]->cqn, -1, &context);
...@@ -1164,8 +1182,9 @@ int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv) ...@@ -1164,8 +1182,9 @@ int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv)
err = -EINVAL; err = -EINVAL;
goto indir_err; goto indir_err;
} }
err = mlx4_qp_to_ready(mdev->dev, &priv->res.mtt, &context, err = mlx4_qp_to_ready(mdev->dev, &priv->res.mtt, &context,
&rss_map->indir_qp, &rss_map->indir_state); rss_map->indir_qp, &rss_map->indir_state);
if (err) if (err)
goto indir_err; goto indir_err;
...@@ -1173,9 +1192,11 @@ int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv) ...@@ -1173,9 +1192,11 @@ int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv)
indir_err: indir_err:
mlx4_qp_modify(mdev->dev, NULL, rss_map->indir_state, mlx4_qp_modify(mdev->dev, NULL, rss_map->indir_state,
MLX4_QP_STATE_RST, NULL, 0, 0, &rss_map->indir_qp); MLX4_QP_STATE_RST, NULL, 0, 0, rss_map->indir_qp);
mlx4_qp_remove(mdev->dev, &rss_map->indir_qp); mlx4_qp_remove(mdev->dev, rss_map->indir_qp);
mlx4_qp_free(mdev->dev, &rss_map->indir_qp); mlx4_qp_free(mdev->dev, rss_map->indir_qp);
kfree(rss_map->indir_qp);
rss_map->indir_qp = NULL;
rss_err: rss_err:
for (i = 0; i < good_qps; i++) { for (i = 0; i < good_qps; i++) {
mlx4_qp_modify(mdev->dev, NULL, rss_map->state[i], mlx4_qp_modify(mdev->dev, NULL, rss_map->state[i],
...@@ -1193,10 +1214,15 @@ void mlx4_en_release_rss_steer(struct mlx4_en_priv *priv) ...@@ -1193,10 +1214,15 @@ void mlx4_en_release_rss_steer(struct mlx4_en_priv *priv)
struct mlx4_en_rss_map *rss_map = &priv->rss_map; struct mlx4_en_rss_map *rss_map = &priv->rss_map;
int i; int i;
mlx4_qp_modify(mdev->dev, NULL, rss_map->indir_state, if (priv->rx_ring_num > 1) {
MLX4_QP_STATE_RST, NULL, 0, 0, &rss_map->indir_qp); mlx4_qp_modify(mdev->dev, NULL, rss_map->indir_state,
mlx4_qp_remove(mdev->dev, &rss_map->indir_qp); MLX4_QP_STATE_RST, NULL, 0, 0,
mlx4_qp_free(mdev->dev, &rss_map->indir_qp); rss_map->indir_qp);
mlx4_qp_remove(mdev->dev, rss_map->indir_qp);
mlx4_qp_free(mdev->dev, rss_map->indir_qp);
kfree(rss_map->indir_qp);
rss_map->indir_qp = NULL;
}
for (i = 0; i < priv->rx_ring_num; i++) { for (i = 0; i < priv->rx_ring_num; i++) {
mlx4_qp_modify(mdev->dev, NULL, rss_map->state[i], mlx4_qp_modify(mdev->dev, NULL, rss_map->state[i],
......
...@@ -2356,8 +2356,8 @@ static int mlx4_init_hca(struct mlx4_dev *dev) ...@@ -2356,8 +2356,8 @@ static int mlx4_init_hca(struct mlx4_dev *dev)
MLX4_A0_STEERING_TABLE_SIZE; MLX4_A0_STEERING_TABLE_SIZE;
} }
mlx4_dbg(dev, "DMFS high rate steer mode is: %s\n", mlx4_info(dev, "DMFS high rate steer mode is: %s\n",
dmfs_high_rate_steering_mode_str( dmfs_high_rate_steering_mode_str(
dev->caps.dmfs_high_steer_mode)); dev->caps.dmfs_high_steer_mode));
} }
} else { } else {
......
...@@ -431,7 +431,7 @@ struct mlx4_en_rss_map { ...@@ -431,7 +431,7 @@ struct mlx4_en_rss_map {
int base_qpn; int base_qpn;
struct mlx4_qp qps[MAX_RX_RINGS]; struct mlx4_qp qps[MAX_RX_RINGS];
enum mlx4_qp_state state[MAX_RX_RINGS]; enum mlx4_qp_state state[MAX_RX_RINGS];
struct mlx4_qp indir_qp; struct mlx4_qp *indir_qp;
enum mlx4_qp_state indir_state; enum mlx4_qp_state indir_state;
}; };
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册