From patchwork Tue Apr 24 10:36:01 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Tal Gilboa X-Patchwork-Id: 903371 X-Patchwork-Delegate: davem@davemloft.net Return-Path: X-Original-To: patchwork-incoming-netdev@ozlabs.org Delivered-To: patchwork-incoming-netdev@ozlabs.org Authentication-Results: ozlabs.org; spf=none (mailfrom) smtp.mailfrom=vger.kernel.org (client-ip=209.132.180.67; helo=vger.kernel.org; envelope-from=netdev-owner@vger.kernel.org; receiver=) Authentication-Results: ozlabs.org; dmarc=fail (p=none dis=none) header.from=mellanox.com Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by ozlabs.org (Postfix) with ESMTP id 40Vh8Q4GKbz9ryr for ; Tue, 24 Apr 2018 21:35:18 +1000 (AEST) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757225AbeDXLfQ (ORCPT ); Tue, 24 Apr 2018 07:35:16 -0400 Received: from mail-il-dmz.mellanox.com ([193.47.165.129]:57697 "EHLO mellanox.co.il" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1757215AbeDXLfM (ORCPT ); Tue, 24 Apr 2018 07:35:12 -0400 Received: from Internal Mail-Server by MTLPINE1 (envelope-from talgi@mellanox.com) with ESMTPS (AES256-SHA encrypted); 24 Apr 2018 14:36:32 +0300 Received: from gen-l-vrt-691.mtl.labs.mlnx (gen-l-vrt-691.mtl.labs.mlnx [10.141.69.1]) by labmailer.mlnx (8.13.8/8.13.8) with ESMTP id w3OBZ6D1030425; Tue, 24 Apr 2018 14:35:06 +0300 Received: from gen-l-vrt-691.mtl.labs.mlnx (localhost [127.0.0.1]) by gen-l-vrt-691.mtl.labs.mlnx (8.14.7/8.14.7) with ESMTP id w3OAalR9041649; Tue, 24 Apr 2018 13:36:47 +0300 Received: (from talgi@localhost) by gen-l-vrt-691.mtl.labs.mlnx (8.14.7/8.14.7/Submit) id w3OAalQI041648; Tue, 24 Apr 2018 13:36:47 +0300 From: Tal Gilboa To: David Miller Cc: netdev@vger.kernel.org, Tariq Toukan , Tal Gilboa , Saeed Mahameed , Florian Fainelli , Andy Gospodarek Subject: [PATCH net-next V3 1/3] net/dim: Rename *_get_profile() functions to *_get_rx_moderation() Date: Tue, 24 Apr 2018 13:36:01 +0300 Message-Id: <1524566163-41563-2-git-send-email-talgi@mellanox.com> X-Mailer: git-send-email 1.8.3.1 In-Reply-To: <1524566163-41563-1-git-send-email-talgi@mellanox.com> References: <1524566163-41563-1-git-send-email-talgi@mellanox.com> Sender: netdev-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org Preparation for introducing adaptive TX to net DIM. Signed-off-by: Tal Gilboa Reviewed-by: Tariq Toukan --- drivers/net/ethernet/broadcom/bcmsysport.c | 6 +++--- drivers/net/ethernet/broadcom/bnxt/bnxt_dim.c | 8 ++++---- drivers/net/ethernet/broadcom/genet/bcmgenet.c | 6 +++--- drivers/net/ethernet/mellanox/mlx5/core/en_dim.c | 6 +++--- drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 6 ++++-- include/linux/net_dim.h | 12 ++++++------ 6 files changed, 23 insertions(+), 21 deletions(-) diff --git a/drivers/net/ethernet/broadcom/bcmsysport.c b/drivers/net/ethernet/broadcom/bcmsysport.c index f9a3c1a..effc651 100644 --- a/drivers/net/ethernet/broadcom/bcmsysport.c +++ b/drivers/net/ethernet/broadcom/bcmsysport.c @@ -654,7 +654,7 @@ static int bcm_sysport_set_coalesce(struct net_device *dev, pkts = priv->rx_max_coalesced_frames; if (ec->use_adaptive_rx_coalesce && !priv->dim.use_dim) { - moder = net_dim_get_def_profile(priv->dim.dim.mode); + moder = net_dim_get_def_rx_moderation(priv->dim.dim.mode); usecs = moder.usec; pkts = moder.pkts; } @@ -1064,7 +1064,7 @@ static void bcm_sysport_dim_work(struct work_struct *work) struct bcm_sysport_priv *priv = container_of(ndim, struct bcm_sysport_priv, dim); struct net_dim_cq_moder cur_profile = - net_dim_get_profile(dim->mode, dim->profile_ix); + net_dim_get_rx_moderation(dim->mode, dim->profile_ix); bcm_sysport_set_rx_coalesce(priv, cur_profile.usec, cur_profile.pkts); dim->state = NET_DIM_START_MEASURE; @@ -1437,7 +1437,7 @@ static void bcm_sysport_init_rx_coalesce(struct bcm_sysport_priv *priv) /* If DIM was enabled, re-apply default parameters */ if (dim->use_dim) { - moder = net_dim_get_def_profile(dim->dim.mode); + moder = net_dim_get_def_rx_moderation(dim->dim.mode); usecs = moder.usec; pkts = moder.pkts; } diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_dim.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_dim.c index 408dd19..afa97c8 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt_dim.c +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_dim.c @@ -21,11 +21,11 @@ void bnxt_dim_work(struct work_struct *work) struct bnxt_napi *bnapi = container_of(cpr, struct bnxt_napi, cp_ring); - struct net_dim_cq_moder cur_profile = net_dim_get_profile(dim->mode, - dim->profile_ix); + struct net_dim_cq_moder cur_moder = + net_dim_get_rx_moderation(dim->mode, dim->profile_ix); - cpr->rx_ring_coal.coal_ticks = cur_profile.usec; - cpr->rx_ring_coal.coal_bufs = cur_profile.pkts; + cpr->rx_ring_coal.coal_ticks = cur_moder.usec; + cpr->rx_ring_coal.coal_bufs = cur_moder.pkts; bnxt_hwrm_set_ring_coal(bnapi->bp, bnapi); dim->state = NET_DIM_START_MEASURE; diff --git a/drivers/net/ethernet/broadcom/genet/bcmgenet.c b/drivers/net/ethernet/broadcom/genet/bcmgenet.c index 0445f2c..20c1681 100644 --- a/drivers/net/ethernet/broadcom/genet/bcmgenet.c +++ b/drivers/net/ethernet/broadcom/genet/bcmgenet.c @@ -652,7 +652,7 @@ static void bcmgenet_set_ring_rx_coalesce(struct bcmgenet_rx_ring *ring, pkts = ring->rx_max_coalesced_frames; if (ec->use_adaptive_rx_coalesce && !ring->dim.use_dim) { - moder = net_dim_get_def_profile(ring->dim.dim.mode); + moder = net_dim_get_def_rx_moderation(ring->dim.dim.mode); usecs = moder.usec; pkts = moder.pkts; } @@ -1925,7 +1925,7 @@ static void bcmgenet_dim_work(struct work_struct *work) struct bcmgenet_rx_ring *ring = container_of(ndim, struct bcmgenet_rx_ring, dim); struct net_dim_cq_moder cur_profile = - net_dim_get_profile(dim->mode, dim->profile_ix); + net_dim_get_rx_moderation(dim->mode, dim->profile_ix); bcmgenet_set_rx_coalesce(ring, cur_profile.usec, cur_profile.pkts); dim->state = NET_DIM_START_MEASURE; @@ -2102,7 +2102,7 @@ static void bcmgenet_init_rx_coalesce(struct bcmgenet_rx_ring *ring) /* If DIM was enabled, re-apply default parameters */ if (dim->use_dim) { - moder = net_dim_get_def_profile(dim->dim.mode); + moder = net_dim_get_def_rx_moderation(dim->dim.mode); usecs = moder.usec; pkts = moder.pkts; } diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c index 602851a..1b286e1 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c @@ -38,11 +38,11 @@ void mlx5e_rx_dim_work(struct work_struct *work) struct net_dim *dim = container_of(work, struct net_dim, work); struct mlx5e_rq *rq = container_of(dim, struct mlx5e_rq, dim); - struct net_dim_cq_moder cur_profile = net_dim_get_profile(dim->mode, - dim->profile_ix); + struct net_dim_cq_moder cur_moder = + net_dim_get_rx_moderation(dim->mode, dim->profile_ix); mlx5_core_modify_cq_moderation(rq->mdev, &rq->cq.mcq, - cur_profile.usec, cur_profile.pkts); + cur_moder.usec, cur_moder.pkts); dim->state = NET_DIM_START_MEASURE; } diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c index f100374..a69cc1c 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c @@ -4119,12 +4119,14 @@ void mlx5e_set_rx_cq_mode_params(struct mlx5e_params *params, u8 cq_period_mode) switch (cq_period_mode) { case MLX5_CQ_PERIOD_MODE_START_FROM_CQE: params->rx_cq_moderation = - net_dim_get_def_profile(NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE); + net_dim_get_def_rx_moderation( + NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE); break; case MLX5_CQ_PERIOD_MODE_START_FROM_EQE: default: params->rx_cq_moderation = - net_dim_get_def_profile(NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE); + net_dim_get_def_rx_moderation( + NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE); } } diff --git a/include/linux/net_dim.h b/include/linux/net_dim.h index 29ed8fd..7ca3c4d 100644 --- a/include/linux/net_dim.h +++ b/include/linux/net_dim.h @@ -129,17 +129,17 @@ enum { NET_DIM_CQE_PROFILES, }; -static inline struct net_dim_cq_moder net_dim_get_profile(u8 cq_period_mode, - int ix) +static inline struct net_dim_cq_moder +net_dim_get_rx_moderation(u8 cq_period_mode, int ix) { - struct net_dim_cq_moder cq_moder; + struct net_dim_cq_moder cq_moder = profile[cq_period_mode][ix]; - cq_moder = profile[cq_period_mode][ix]; cq_moder.cq_period_mode = cq_period_mode; return cq_moder; } -static inline struct net_dim_cq_moder net_dim_get_def_profile(u8 rx_cq_period_mode) +static inline struct net_dim_cq_moder +net_dim_get_def_rx_moderation(u8 rx_cq_period_mode) { int default_profile_ix; @@ -148,7 +148,7 @@ static inline struct net_dim_cq_moder net_dim_get_def_profile(u8 rx_cq_period_mo else /* NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE */ default_profile_ix = NET_DIM_DEF_PROFILE_EQE; - return net_dim_get_profile(rx_cq_period_mode, default_profile_ix); + return net_dim_get_rx_moderation(rx_cq_period_mode, default_profile_ix); } static inline bool net_dim_on_top(struct net_dim *dim) From patchwork Tue Apr 24 10:36:02 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Tal Gilboa X-Patchwork-Id: 903372 X-Patchwork-Delegate: davem@davemloft.net Return-Path: X-Original-To: patchwork-incoming-netdev@ozlabs.org Delivered-To: patchwork-incoming-netdev@ozlabs.org Authentication-Results: ozlabs.org; spf=none (mailfrom) smtp.mailfrom=vger.kernel.org (client-ip=209.132.180.67; helo=vger.kernel.org; envelope-from=netdev-owner@vger.kernel.org; receiver=) Authentication-Results: ozlabs.org; dmarc=fail (p=none dis=none) header.from=mellanox.com Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by ozlabs.org (Postfix) with ESMTP id 40Vh8d3W8Mz9ry1 for ; Tue, 24 Apr 2018 21:35:29 +1000 (AEST) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757228AbeDXLf2 (ORCPT ); Tue, 24 Apr 2018 07:35:28 -0400 Received: from mail-il-dmz.mellanox.com ([193.47.165.129]:57690 "EHLO mellanox.co.il" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1757217AbeDXLfM (ORCPT ); Tue, 24 Apr 2018 07:35:12 -0400 Received: from Internal Mail-Server by MTLPINE1 (envelope-from talgi@mellanox.com) with ESMTPS (AES256-SHA encrypted); 24 Apr 2018 14:36:32 +0300 Received: from gen-l-vrt-691.mtl.labs.mlnx (gen-l-vrt-691.mtl.labs.mlnx [10.141.69.1]) by labmailer.mlnx (8.13.8/8.13.8) with ESMTP id w3OBZ6Cx030425; Tue, 24 Apr 2018 14:35:06 +0300 Received: from gen-l-vrt-691.mtl.labs.mlnx (localhost [127.0.0.1]) by gen-l-vrt-691.mtl.labs.mlnx (8.14.7/8.14.7) with ESMTP id w3OAanUg041653; Tue, 24 Apr 2018 13:36:49 +0300 Received: (from talgi@localhost) by gen-l-vrt-691.mtl.labs.mlnx (8.14.7/8.14.7/Submit) id w3OAanc8041652; Tue, 24 Apr 2018 13:36:49 +0300 From: Tal Gilboa To: David Miller Cc: netdev@vger.kernel.org, Tariq Toukan , Tal Gilboa , Saeed Mahameed , Florian Fainelli , Andy Gospodarek Subject: [PATCH net-next V3 2/3] net/dim: Support adaptive TX moderation Date: Tue, 24 Apr 2018 13:36:02 +0300 Message-Id: <1524566163-41563-3-git-send-email-talgi@mellanox.com> X-Mailer: git-send-email 1.8.3.1 In-Reply-To: <1524566163-41563-1-git-send-email-talgi@mellanox.com> References: <1524566163-41563-1-git-send-email-talgi@mellanox.com> Sender: netdev-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org Interrupt moderation for TX traffic requires different profiles than RX interrupt moderation. The main goal here is to reduce interrupt rate and allow better payload aggregation by keeping SKBs in the TX queue a bit longer. Ping-pong behavior would get a profile with a short timer, so latency wouldn't increase for these scenarios. There might be a slight degradation in bandwidth for single stream with large message sizes, since net.ipv4.tcp_limit_output_bytes is limiting the allowed TX traffic, but with many streams performance is always improved. Signed-off-by: Tal Gilboa Reviewed-by: Tariq Toukan --- include/linux/net_dim.h | 63 +++++++++++++++++++++++++++++++++++++++---------- 1 file changed, 50 insertions(+), 13 deletions(-) diff --git a/include/linux/net_dim.h b/include/linux/net_dim.h index 7ca3c4d..db99240 100644 --- a/include/linux/net_dim.h +++ b/include/linux/net_dim.h @@ -103,11 +103,12 @@ enum { #define NET_DIM_PARAMS_NUM_PROFILES 5 /* Adaptive moderation profiles */ #define NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE 256 +#define NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE 128 #define NET_DIM_DEF_PROFILE_CQE 1 #define NET_DIM_DEF_PROFILE_EQE 1 /* All profiles sizes must be NET_PARAMS_DIM_NUM_PROFILES */ -#define NET_DIM_EQE_PROFILES { \ +#define NET_DIM_RX_EQE_PROFILES { \ {1, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \ {8, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \ {64, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \ @@ -115,7 +116,7 @@ enum { {256, NET_DIM_DEFAULT_RX_CQ_MODERATION_PKTS_FROM_EQE}, \ } -#define NET_DIM_CQE_PROFILES { \ +#define NET_DIM_RX_CQE_PROFILES { \ {2, 256}, \ {8, 128}, \ {16, 64}, \ @@ -123,32 +124,68 @@ enum { {64, 64} \ } +#define NET_DIM_TX_EQE_PROFILES { \ + {1, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE}, \ + {8, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE}, \ + {32, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE}, \ + {64, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE}, \ + {128, NET_DIM_DEFAULT_TX_CQ_MODERATION_PKTS_FROM_EQE} \ +} + +#define NET_DIM_TX_CQE_PROFILES { \ + {5, 128}, \ + {8, 64}, \ + {16, 32}, \ + {32, 32}, \ + {64, 32} \ +} + static const struct net_dim_cq_moder -profile[NET_DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = { - NET_DIM_EQE_PROFILES, - NET_DIM_CQE_PROFILES, +rx_profile[NET_DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = { + NET_DIM_RX_EQE_PROFILES, + NET_DIM_RX_CQE_PROFILES, +}; + +static const struct net_dim_cq_moder +tx_profile[NET_DIM_CQ_PERIOD_NUM_MODES][NET_DIM_PARAMS_NUM_PROFILES] = { + NET_DIM_TX_EQE_PROFILES, + NET_DIM_TX_CQE_PROFILES, }; static inline struct net_dim_cq_moder net_dim_get_rx_moderation(u8 cq_period_mode, int ix) { - struct net_dim_cq_moder cq_moder = profile[cq_period_mode][ix]; + struct net_dim_cq_moder cq_moder = rx_profile[cq_period_mode][ix]; cq_moder.cq_period_mode = cq_period_mode; return cq_moder; } static inline struct net_dim_cq_moder -net_dim_get_def_rx_moderation(u8 rx_cq_period_mode) +net_dim_get_def_rx_moderation(u8 cq_period_mode) +{ + u8 profile_ix = cq_period_mode == NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE ? + NET_DIM_DEF_PROFILE_CQE : NET_DIM_DEF_PROFILE_EQE; + + return net_dim_get_rx_moderation(cq_period_mode, profile_ix); +} + +static inline struct net_dim_cq_moder +net_dim_get_tx_moderation(u8 cq_period_mode, int ix) { - int default_profile_ix; + struct net_dim_cq_moder cq_moder = tx_profile[cq_period_mode][ix]; - if (rx_cq_period_mode == NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE) - default_profile_ix = NET_DIM_DEF_PROFILE_CQE; - else /* NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE */ - default_profile_ix = NET_DIM_DEF_PROFILE_EQE; + cq_moder.cq_period_mode = cq_period_mode; + return cq_moder; +} + +static inline struct net_dim_cq_moder +net_dim_get_def_tx_moderation(u8 cq_period_mode) +{ + u8 profile_ix = cq_period_mode == NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE ? + NET_DIM_DEF_PROFILE_CQE : NET_DIM_DEF_PROFILE_EQE; - return net_dim_get_rx_moderation(rx_cq_period_mode, default_profile_ix); + return net_dim_get_tx_moderation(cq_period_mode, profile_ix); } static inline bool net_dim_on_top(struct net_dim *dim) From patchwork Tue Apr 24 10:36:03 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Tal Gilboa X-Patchwork-Id: 903373 X-Patchwork-Delegate: davem@davemloft.net Return-Path: X-Original-To: patchwork-incoming-netdev@ozlabs.org Delivered-To: patchwork-incoming-netdev@ozlabs.org Authentication-Results: ozlabs.org; spf=none (mailfrom) smtp.mailfrom=vger.kernel.org (client-ip=209.132.180.67; helo=vger.kernel.org; envelope-from=netdev-owner@vger.kernel.org; receiver=) Authentication-Results: ozlabs.org; dmarc=fail (p=none dis=none) header.from=mellanox.com Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by ozlabs.org (Postfix) with ESMTP id 40Vh8f5SrXz9ryr for ; Tue, 24 Apr 2018 21:35:30 +1000 (AEST) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757232AbeDXLf2 (ORCPT ); Tue, 24 Apr 2018 07:35:28 -0400 Received: from mail-il-dmz.mellanox.com ([193.47.165.129]:57698 "EHLO mellanox.co.il" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1757221AbeDXLfM (ORCPT ); Tue, 24 Apr 2018 07:35:12 -0400 Received: from Internal Mail-Server by MTLPINE1 (envelope-from talgi@mellanox.com) with ESMTPS (AES256-SHA encrypted); 24 Apr 2018 14:36:32 +0300 Received: from gen-l-vrt-691.mtl.labs.mlnx (gen-l-vrt-691.mtl.labs.mlnx [10.141.69.1]) by labmailer.mlnx (8.13.8/8.13.8) with ESMTP id w3OBZ6D3030425; Tue, 24 Apr 2018 14:35:06 +0300 Received: from gen-l-vrt-691.mtl.labs.mlnx (localhost [127.0.0.1]) by gen-l-vrt-691.mtl.labs.mlnx (8.14.7/8.14.7) with ESMTP id w3OAapPj041656; Tue, 24 Apr 2018 13:36:51 +0300 Received: (from talgi@localhost) by gen-l-vrt-691.mtl.labs.mlnx (8.14.7/8.14.7/Submit) id w3OAaonl041655; Tue, 24 Apr 2018 13:36:50 +0300 From: Tal Gilboa To: David Miller Cc: netdev@vger.kernel.org, Tariq Toukan , Tal Gilboa , Saeed Mahameed , Florian Fainelli , Andy Gospodarek Subject: [PATCH net-next V3 3/3] net/mlx5e: Enable adaptive-TX moderation Date: Tue, 24 Apr 2018 13:36:03 +0300 Message-Id: <1524566163-41563-4-git-send-email-talgi@mellanox.com> X-Mailer: git-send-email 1.8.3.1 In-Reply-To: <1524566163-41563-1-git-send-email-talgi@mellanox.com> References: <1524566163-41563-1-git-send-email-talgi@mellanox.com> Sender: netdev-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org Add support for adaptive TX moderation. This greatly reduces TX interrupt rate and increases bandwidth, mostly for TCP bandwidth over ARM architecture (below). There is a slight single stream TCP with very large message sizes degradation (x86). In this case if there's any moderation on transmitted packets the bandwidth would reduce due to hitting TCP output limit. Since this is a synthetic case, this is still worth doing. Performance improvement (ConnectX-4Lx 40GbE, ARM) TCP 64B bandwidth with 1-50 streams increased 6-35%. TCP 64B bandwidth with 100-500 streams increased 20-70%. Performance improvement (ConnectX-5 100GbE, x86) Bandwidth: increased up to 40% (1024B with 10s of streams). Interrupt rate: reduced up to 50% (1024B with 1000s of streams). Performance degradation (ConnectX-5 100GbE, x86) Bandwidth: up to 10% decrease single stream TCP (1MB message size from 51Gb/s to 47Gb/s). Signed-off-by: Tal Gilboa Reviewed-by: Tariq Toukan Acked-by: Saeed Mahameed --- drivers/net/ethernet/mellanox/mlx5/core/en.h | 4 ++ drivers/net/ethernet/mellanox/mlx5/core/en_dim.c | 24 +++++-- .../net/ethernet/mellanox/mlx5/core/en_ethtool.c | 35 +++++++--- drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 81 +++++++++++++--------- drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c | 37 +++++++--- 5 files changed, 125 insertions(+), 56 deletions(-) diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en.h b/drivers/net/ethernet/mellanox/mlx5/core/en.h index 3317a4d..56c748c 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en.h +++ b/drivers/net/ethernet/mellanox/mlx5/core/en.h @@ -241,6 +241,7 @@ struct mlx5e_params { bool vlan_strip_disable; bool scatter_fcs_en; bool rx_dim_enabled; + bool tx_dim_enabled; u32 lro_timeout; u32 pflags; struct bpf_prog *xdp_prog; @@ -330,6 +331,7 @@ enum { MLX5E_SQ_STATE_ENABLED, MLX5E_SQ_STATE_RECOVERING, MLX5E_SQ_STATE_IPSEC, + MLX5E_SQ_STATE_AM, }; struct mlx5e_sq_wqe_info { @@ -342,6 +344,7 @@ struct mlx5e_txqsq { /* dirtied @completion */ u16 cc; u32 dma_fifo_cc; + struct net_dim dim; /* Adaptive Moderation */ /* dirtied @xmit */ u16 pc ____cacheline_aligned_in_smp; @@ -1111,4 +1114,5 @@ void mlx5e_build_nic_params(struct mlx5_core_dev *mdev, u16 max_channels, u16 mtu); u8 mlx5e_params_calculate_tx_min_inline(struct mlx5_core_dev *mdev); void mlx5e_rx_dim_work(struct work_struct *work); +void mlx5e_tx_dim_work(struct work_struct *work); #endif /* __MLX5_EN_H__ */ diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c index 1b286e1..d67adf7 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_dim.c @@ -33,16 +33,30 @@ #include #include "en.h" +static void +mlx5e_complete_dim_work(struct net_dim *dim, struct net_dim_cq_moder moder, + struct mlx5_core_dev *mdev, struct mlx5_core_cq *mcq) +{ + mlx5_core_modify_cq_moderation(mdev, mcq, moder.usec, moder.pkts); + dim->state = NET_DIM_START_MEASURE; +} + void mlx5e_rx_dim_work(struct work_struct *work) { - struct net_dim *dim = container_of(work, struct net_dim, - work); + struct net_dim *dim = container_of(work, struct net_dim, work); struct mlx5e_rq *rq = container_of(dim, struct mlx5e_rq, dim); struct net_dim_cq_moder cur_moder = net_dim_get_rx_moderation(dim->mode, dim->profile_ix); - mlx5_core_modify_cq_moderation(rq->mdev, &rq->cq.mcq, - cur_moder.usec, cur_moder.pkts); + mlx5e_complete_dim_work(dim, cur_moder, rq->mdev, &rq->cq.mcq); +} - dim->state = NET_DIM_START_MEASURE; +void mlx5e_tx_dim_work(struct work_struct *work) +{ + struct net_dim *dim = container_of(work, struct net_dim, work); + struct mlx5e_txqsq *sq = container_of(dim, struct mlx5e_txqsq, dim); + struct net_dim_cq_moder cur_moder = + net_dim_get_tx_moderation(dim->mode, dim->profile_ix); + + mlx5e_complete_dim_work(dim, cur_moder, sq->cq.mdev, &sq->cq.mcq); } diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c b/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c index 37fd024..2b786c4 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_ethtool.c @@ -389,14 +389,20 @@ static int mlx5e_set_channels(struct net_device *dev, int mlx5e_ethtool_get_coalesce(struct mlx5e_priv *priv, struct ethtool_coalesce *coal) { + struct net_dim_cq_moder *rx_moder, *tx_moder; + if (!MLX5_CAP_GEN(priv->mdev, cq_moderation)) return -EOPNOTSUPP; - coal->rx_coalesce_usecs = priv->channels.params.rx_cq_moderation.usec; - coal->rx_max_coalesced_frames = priv->channels.params.rx_cq_moderation.pkts; - coal->tx_coalesce_usecs = priv->channels.params.tx_cq_moderation.usec; - coal->tx_max_coalesced_frames = priv->channels.params.tx_cq_moderation.pkts; - coal->use_adaptive_rx_coalesce = priv->channels.params.rx_dim_enabled; + rx_moder = &priv->channels.params.rx_cq_moderation; + coal->rx_coalesce_usecs = rx_moder->usec; + coal->rx_max_coalesced_frames = rx_moder->pkts; + coal->use_adaptive_rx_coalesce = priv->channels.params.rx_dim_enabled; + + tx_moder = &priv->channels.params.tx_cq_moderation; + coal->tx_coalesce_usecs = tx_moder->usec; + coal->tx_max_coalesced_frames = tx_moder->pkts; + coal->use_adaptive_tx_coalesce = priv->channels.params.tx_dim_enabled; return 0; } @@ -438,6 +444,7 @@ static int mlx5e_get_coalesce(struct net_device *netdev, int mlx5e_ethtool_set_coalesce(struct mlx5e_priv *priv, struct ethtool_coalesce *coal) { + struct net_dim_cq_moder *rx_moder, *tx_moder; struct mlx5_core_dev *mdev = priv->mdev; struct mlx5e_channels new_channels = {}; int err = 0; @@ -463,11 +470,15 @@ int mlx5e_ethtool_set_coalesce(struct mlx5e_priv *priv, mutex_lock(&priv->state_lock); new_channels.params = priv->channels.params; - new_channels.params.tx_cq_moderation.usec = coal->tx_coalesce_usecs; - new_channels.params.tx_cq_moderation.pkts = coal->tx_max_coalesced_frames; - new_channels.params.rx_cq_moderation.usec = coal->rx_coalesce_usecs; - new_channels.params.rx_cq_moderation.pkts = coal->rx_max_coalesced_frames; - new_channels.params.rx_dim_enabled = !!coal->use_adaptive_rx_coalesce; + rx_moder = &new_channels.params.rx_cq_moderation; + rx_moder->usec = coal->rx_coalesce_usecs; + rx_moder->pkts = coal->rx_max_coalesced_frames; + new_channels.params.rx_dim_enabled = !!coal->use_adaptive_rx_coalesce; + + tx_moder = &new_channels.params.tx_cq_moderation; + tx_moder->usec = coal->tx_coalesce_usecs; + tx_moder->pkts = coal->tx_max_coalesced_frames; + new_channels.params.tx_dim_enabled = !!coal->use_adaptive_tx_coalesce; if (!test_bit(MLX5E_STATE_OPENED, &priv->state)) { priv->channels.params = new_channels.params; @@ -475,7 +486,9 @@ int mlx5e_ethtool_set_coalesce(struct mlx5e_priv *priv, } /* we are opened */ - reset = !!coal->use_adaptive_rx_coalesce != priv->channels.params.rx_dim_enabled; + reset = (!!coal->use_adaptive_rx_coalesce != priv->channels.params.rx_dim_enabled) || + (!!coal->use_adaptive_tx_coalesce != priv->channels.params.tx_dim_enabled); + if (!reset) { mlx5e_set_priv_channels_coalesce(priv, coal); priv->channels.params = new_channels.params; diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c index a69cc1c..f1fe490 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c @@ -1025,6 +1025,9 @@ static int mlx5e_alloc_txqsq(struct mlx5e_channel *c, if (err) goto err_sq_wq_destroy; + INIT_WORK(&sq->dim.work, mlx5e_tx_dim_work); + sq->dim.mode = params->tx_cq_moderation.cq_period_mode; + sq->edge = (sq->wq.sz_m1 + 1) - MLX5_SEND_WQE_MAX_WQEBBS; return 0; @@ -1188,6 +1191,9 @@ static int mlx5e_open_txqsq(struct mlx5e_channel *c, if (tx_rate) mlx5e_set_sq_maxrate(c->netdev, sq, tx_rate); + if (params->tx_dim_enabled) + sq->state |= BIT(MLX5E_SQ_STATE_AM); + return 0; err_free_txqsq: @@ -4084,18 +4090,48 @@ static bool slow_pci_heuristic(struct mlx5_core_dev *mdev) link_speed > MLX5E_SLOW_PCI_RATIO * pci_bw; } -void mlx5e_set_tx_cq_mode_params(struct mlx5e_params *params, u8 cq_period_mode) +static struct net_dim_cq_moder mlx5e_get_def_tx_moderation(u8 cq_period_mode) { - params->tx_cq_moderation.cq_period_mode = cq_period_mode; + struct net_dim_cq_moder moder; + + moder.cq_period_mode = cq_period_mode; + moder.pkts = MLX5E_PARAMS_DEFAULT_TX_CQ_MODERATION_PKTS; + moder.usec = MLX5E_PARAMS_DEFAULT_TX_CQ_MODERATION_USEC; + if (cq_period_mode == MLX5_CQ_PERIOD_MODE_START_FROM_CQE) + moder.usec = MLX5E_PARAMS_DEFAULT_TX_CQ_MODERATION_USEC_FROM_CQE; + + return moder; +} - params->tx_cq_moderation.pkts = - MLX5E_PARAMS_DEFAULT_TX_CQ_MODERATION_PKTS; - params->tx_cq_moderation.usec = - MLX5E_PARAMS_DEFAULT_TX_CQ_MODERATION_USEC; +static struct net_dim_cq_moder mlx5e_get_def_rx_moderation(u8 cq_period_mode) +{ + struct net_dim_cq_moder moder; + moder.cq_period_mode = cq_period_mode; + moder.pkts = MLX5E_PARAMS_DEFAULT_RX_CQ_MODERATION_PKTS; + moder.usec = MLX5E_PARAMS_DEFAULT_RX_CQ_MODERATION_USEC; if (cq_period_mode == MLX5_CQ_PERIOD_MODE_START_FROM_CQE) - params->tx_cq_moderation.usec = - MLX5E_PARAMS_DEFAULT_TX_CQ_MODERATION_USEC_FROM_CQE; + moder.usec = MLX5E_PARAMS_DEFAULT_RX_CQ_MODERATION_USEC_FROM_CQE; + + return moder; +} + +static u8 mlx5_to_net_dim_cq_period_mode(u8 cq_period_mode) +{ + return cq_period_mode == MLX5_CQ_PERIOD_MODE_START_FROM_CQE ? + NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE : + NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE; +} + +void mlx5e_set_tx_cq_mode_params(struct mlx5e_params *params, u8 cq_period_mode) +{ + if (params->tx_dim_enabled) { + u8 dim_period_mode = mlx5_to_net_dim_cq_period_mode(cq_period_mode); + + params->tx_cq_moderation = net_dim_get_def_tx_moderation(dim_period_mode); + } else { + params->tx_cq_moderation = mlx5e_get_def_tx_moderation(cq_period_mode); + } MLX5E_SET_PFLAG(params, MLX5E_PFLAG_TX_CQE_BASED_MODER, params->tx_cq_moderation.cq_period_mode == @@ -4104,30 +4140,12 @@ void mlx5e_set_tx_cq_mode_params(struct mlx5e_params *params, u8 cq_period_mode) void mlx5e_set_rx_cq_mode_params(struct mlx5e_params *params, u8 cq_period_mode) { - params->rx_cq_moderation.cq_period_mode = cq_period_mode; - - params->rx_cq_moderation.pkts = - MLX5E_PARAMS_DEFAULT_RX_CQ_MODERATION_PKTS; - params->rx_cq_moderation.usec = - MLX5E_PARAMS_DEFAULT_RX_CQ_MODERATION_USEC; - - if (cq_period_mode == MLX5_CQ_PERIOD_MODE_START_FROM_CQE) - params->rx_cq_moderation.usec = - MLX5E_PARAMS_DEFAULT_RX_CQ_MODERATION_USEC_FROM_CQE; - if (params->rx_dim_enabled) { - switch (cq_period_mode) { - case MLX5_CQ_PERIOD_MODE_START_FROM_CQE: - params->rx_cq_moderation = - net_dim_get_def_rx_moderation( - NET_DIM_CQ_PERIOD_MODE_START_FROM_CQE); - break; - case MLX5_CQ_PERIOD_MODE_START_FROM_EQE: - default: - params->rx_cq_moderation = - net_dim_get_def_rx_moderation( - NET_DIM_CQ_PERIOD_MODE_START_FROM_EQE); - } + u8 dim_period_mode = mlx5_to_net_dim_cq_period_mode(cq_period_mode); + + params->rx_cq_moderation = net_dim_get_def_rx_moderation(dim_period_mode); + } else { + params->rx_cq_moderation = mlx5e_get_def_rx_moderation(cq_period_mode); } MLX5E_SET_PFLAG(params, MLX5E_PFLAG_RX_CQE_BASED_MODER, @@ -4191,6 +4209,7 @@ void mlx5e_build_nic_params(struct mlx5_core_dev *mdev, MLX5_CQ_PERIOD_MODE_START_FROM_CQE : MLX5_CQ_PERIOD_MODE_START_FROM_EQE; params->rx_dim_enabled = MLX5_CAP_GEN(mdev, cq_moderation); + params->tx_dim_enabled = MLX5_CAP_GEN(mdev, cq_moderation); mlx5e_set_rx_cq_mode_params(params, rx_cq_period_mode); mlx5e_set_tx_cq_mode_params(params, MLX5_CQ_PERIOD_MODE_START_FROM_EQE); diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c index f292bb3..900661d 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_txrx.c @@ -44,6 +44,30 @@ static inline bool mlx5e_channel_no_affinity_change(struct mlx5e_channel *c) return cpumask_test_cpu(current_cpu, aff); } +static void mlx5e_handle_tx_dim(struct mlx5e_txqsq *sq) +{ + struct net_dim_sample dim_sample; + + if (unlikely(!MLX5E_TEST_BIT(sq->state, MLX5E_SQ_STATE_AM))) + return; + + net_dim_sample(sq->cq.event_ctr, sq->stats.packets, sq->stats.bytes, + &dim_sample); + net_dim(&sq->dim, dim_sample); +} + +static void mlx5e_handle_rx_dim(struct mlx5e_rq *rq) +{ + struct net_dim_sample dim_sample; + + if (unlikely(!MLX5E_TEST_BIT(rq->state, MLX5E_RQ_STATE_AM))) + return; + + net_dim_sample(rq->cq.event_ctr, rq->stats.packets, rq->stats.bytes, + &dim_sample); + net_dim(&rq->dim, dim_sample); +} + int mlx5e_napi_poll(struct napi_struct *napi, int budget) { struct mlx5e_channel *c = container_of(napi, struct mlx5e_channel, @@ -75,18 +99,13 @@ int mlx5e_napi_poll(struct napi_struct *napi, int budget) if (unlikely(!napi_complete_done(napi, work_done))) return work_done; - for (i = 0; i < c->num_tc; i++) + for (i = 0; i < c->num_tc; i++) { + mlx5e_handle_tx_dim(&c->sq[i]); mlx5e_cq_arm(&c->sq[i].cq); - - if (MLX5E_TEST_BIT(c->rq.state, MLX5E_RQ_STATE_AM)) { - struct net_dim_sample dim_sample; - net_dim_sample(c->rq.cq.event_ctr, - c->rq.stats.packets, - c->rq.stats.bytes, - &dim_sample); - net_dim(&c->rq.dim, dim_sample); } + mlx5e_handle_rx_dim(&c->rq); + mlx5e_cq_arm(&c->rq.cq); mlx5e_cq_arm(&c->icosq.cq);