[dpdk-stable] patch 'net/mlx5: fix drop action for Direct Rules/Verbs' has been queued to stable release 20.11.2

Xueming Li xuemingl at nvidia.com
Mon May 10 18:01:35 CEST 2021
Previous message (by thread): [dpdk-stable] patch 'net/mlx5: fix missing shared RSS hash types' has been queued to stable release 20.11.2
Next message (by thread): [dpdk-stable] patch 'net/bnxt: fix double free in port start failure' has been queued to stable release 20.11.2
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Hi,

FYI, your patch has been queued to stable release 20.11.2

Note it hasn't been pushed to http://dpdk.org/browse/dpdk-stable yet.
It will be pushed if I get no objections before 05/12/21. So please
shout if anyone has objections.

Also note that after the patch there's a diff of the upstream commit vs the
patch applied to the branch. This will indicate if there was any rebasing
needed to apply to the stable branch. If there were code changes for rebasing
(ie: not only metadata diffs), please double check that the rebase was
correctly done.

Queued patches are on a temporary branch at:
https://github.com/steevenlee/dpdk

This queued commit can be viewed at:
https://github.com/steevenlee/dpdk/commit/f30dc57e7857c39d34d1255a589f11ee499b9d17

Thanks.

Xueming Li <xuemingl at nvidia.com>

---
>From f30dc57e7857c39d34d1255a589f11ee499b9d17 Mon Sep 17 00:00:00 2001
From: Viacheslav Ovsiienko <viacheslavo at nvidia.com>
Date: Mon, 5 Apr 2021 09:59:01 +0000
Subject: [PATCH] net/mlx5: fix drop action for Direct Rules/Verbs
Cc: Luca Boccassi <bluca at debian.org>

[ upstream commit da845ae9d7c1499cbf76e766604cc981ddd9eb17 ]

There are multiple branches in rdma-core library backing
the rte flows:
  - Verbs
  - Direct Verbs (DV)
  - Direct Rules (DR)

The Verbs API always requires the specifying the queue even
if there is the drop action in the flow, though the kernel
optimizes out the actual queue usage for the flows containing
the drop action. The PMD handles the dedicated Rx queue to
provide Verbs API compatibility.

The DV/DR API does not require explicit specifying the queue
at the flow creation, but PMD still specified the dedicated
drop queue as action. It performed the packet forwarding to
the dummy queue (that was not polled at all) causing the
steering pipeline resources usage and degrading the overall
packet processing rate. For example, with inserted flow to
drop all the ingress packets the statistics reported only
15Mpps of 64B packets were received over 100Gbps line.

Since the Direct Rule API for E-Switch was introduced the
rdma-core supports the dedicated drop action, that is recognized
both for DV and DR and can be used for the entire device in
unified fashion, regardless of steering domain. The similar drop
action was introduced for E-Switch, the usage of this one can be
extended for other steering domains, not for E-Switch's one only.

This patch:
  - renames esw_drop_action to dr_drop_action to emphasize
    the global nature of the variable (not only E-Switch domain)
  - specifies this global drop action instead of dedicated
    drop queue for the DR/DV flows

Fixes: 34fa7c0268e7 ("net/mlx5: add drop action to Direct Verbs E-Switch")
Fixes: 65b3cd0dc39b ("net/mlx5: create global drop action")

Signed-off-by: Viacheslav Ovsiienko <viacheslavo at nvidia.com>
Acked-by: Matan Azrad <matan at nvidia.com>
---
 drivers/net/mlx5/linux/mlx5_os.c | 24 +++++++++++++++++-------
 drivers/net/mlx5/mlx5.h          |  2 +-
 drivers/net/mlx5/mlx5_flow_dv.c  | 13 +++++++++++--
 3 files changed, 29 insertions(+), 10 deletions(-)

diff --git a/drivers/net/mlx5/linux/mlx5_os.c b/drivers/net/mlx5/linux/mlx5_os.c
index 91001473b0..fb385460e5 100644
--- a/drivers/net/mlx5/linux/mlx5_os.c
+++ b/drivers/net/mlx5/linux/mlx5_os.c
@@ -319,7 +319,17 @@ mlx5_alloc_shared_dr(struct mlx5_priv *priv)
 			goto error;
 		}
 		sh->fdb_domain = domain;
-		sh->esw_drop_action = mlx5_glue->dr_create_flow_action_drop();
+	}
+	/*
+	 * The drop action is just some dummy placeholder in rdma-core. It
+	 * does not belong to domains and has no any attributes, and, can be
+	 * shared by the entire device.
+	 */
+	sh->dr_drop_action = mlx5_glue->dr_create_flow_action_drop();
+	if (!sh->dr_drop_action) {
+		DRV_LOG(ERR, "FDB mlx5dv_dr_create_flow_action_drop");
+		err = errno;
+		goto error;
 	}
 #endif
 	if (!sh->tunnel_hub)
@@ -355,9 +365,9 @@ error:
 		mlx5_glue->dr_destroy_domain(sh->fdb_domain);
 		sh->fdb_domain = NULL;
 	}
-	if (sh->esw_drop_action) {
-		mlx5_glue->destroy_flow_action(sh->esw_drop_action);
-		sh->esw_drop_action = NULL;
+	if (sh->dr_drop_action) {
+		mlx5_glue->destroy_flow_action(sh->dr_drop_action);
+		sh->dr_drop_action = NULL;
 	}
 	if (sh->pop_vlan_action) {
 		mlx5_glue->destroy_flow_action(sh->pop_vlan_action);
@@ -412,9 +422,9 @@ mlx5_os_free_shared_dr(struct mlx5_priv *priv)
 		mlx5_glue->dr_destroy_domain(sh->fdb_domain);
 		sh->fdb_domain = NULL;
 	}
-	if (sh->esw_drop_action) {
-		mlx5_glue->destroy_flow_action(sh->esw_drop_action);
-		sh->esw_drop_action = NULL;
+	if (sh->dr_drop_action) {
+		mlx5_glue->destroy_flow_action(sh->dr_drop_action);
+		sh->dr_drop_action = NULL;
 	}
 #endif
 	if (sh->pop_vlan_action) {
diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
index 9bf1bf3146..97f8a016b4 100644
--- a/drivers/net/mlx5/mlx5.h
+++ b/drivers/net/mlx5/mlx5.h
@@ -719,7 +719,7 @@ struct mlx5_dev_ctx_shared {
 	struct mlx5_hlist *flow_tbls;
 	struct mlx5_flow_tunnel_hub *tunnel_hub;
 	/* Direct Rules tables for FDB, NIC TX+RX */
-	void *esw_drop_action; /* Pointer to DR E-Switch drop action. */
+	void *dr_drop_action; /* Pointer to DR drop action, any domain. */
 	void *pop_vlan_action; /* Pointer to DR pop VLAN action. */
 	struct mlx5_hlist *encaps_decaps; /* Encap/decap action hash list. */
 	struct mlx5_hlist *modify_cmds;
diff --git a/drivers/net/mlx5/mlx5_flow_dv.c b/drivers/net/mlx5/mlx5_flow_dv.c
index 94550e5a73..b1c8a95e8a 100644
--- a/drivers/net/mlx5/mlx5_flow_dv.c
+++ b/drivers/net/mlx5/mlx5_flow_dv.c
@@ -10839,11 +10839,19 @@ flow_dv_apply(struct rte_eth_dev *dev, struct rte_flow *flow,
 		n = dv->actions_n;
 		if (dh->fate_action == MLX5_FLOW_FATE_DROP) {
 			if (dv->transfer) {
-				dv->actions[n++] = priv->sh->esw_drop_action;
+				MLX5_ASSERT(priv->sh->dr_drop_action);
+				dv->actions[n++] = priv->sh->dr_drop_action;
 			} else {
+#ifdef HAVE_MLX5DV_DR
+				/* DR supports drop action placeholder. */
+				MLX5_ASSERT(priv->sh->dr_drop_action);
+				dv->actions[n++] = priv->sh->dr_drop_action;
+#else
+				/* For DV we use the explicit drop queue. */
 				MLX5_ASSERT(priv->drop_queue.hrxq);
 				dv->actions[n++] =
 						priv->drop_queue.hrxq->action;
+#endif
 			}
 		} else if ((dh->fate_action == MLX5_FLOW_FATE_QUEUE &&
 			   !dv_h->rix_sample && !dv_h->rix_dest_array)) {
@@ -12610,7 +12618,8 @@ mlx5_flow_dv_discover_counter_offset_support(struct rte_eth_dev *dev)
 						    &actions[0]);
 	if (ret)
 		goto err;
-	actions[1] = priv->drop_queue.hrxq->action;
+	actions[1] = sh->dr_drop_action ? sh->dr_drop_action :
+					  priv->drop_queue.hrxq->action;
 	dv_attr.match_criteria_enable = flow_dv_matcher_enable(mask.buf);
 	ret = mlx5_flow_os_create_flow_matcher(sh->ctx, &dv_attr, tbl->obj,
 					       &matcher);
-- 
2.25.1

---
  Diff of the applied patch vs upstream commit (please double-check if non-empty:
---
--- -	2021-05-10 23:59:30.377593400 +0800
+++ 0147-net-mlx5-fix-drop-action-for-Direct-Rules-Verbs.patch	2021-05-10 23:59:26.570000000 +0800
@@ -1 +1 @@
-From da845ae9d7c1499cbf76e766604cc981ddd9eb17 Mon Sep 17 00:00:00 2001
+From f30dc57e7857c39d34d1255a589f11ee499b9d17 Mon Sep 17 00:00:00 2001
@@ -4,0 +5,3 @@
+Cc: Luca Boccassi <bluca at debian.org>
+
+[ upstream commit da845ae9d7c1499cbf76e766604cc981ddd9eb17 ]
@@ -42 +44,0 @@
-Cc: stable at dpdk.org
@@ -53 +55 @@
-index 2d5bcab4cf..6ac334263e 100644
+index 91001473b0..fb385460e5 100644
@@ -56 +58 @@
-@@ -325,7 +325,17 @@ mlx5_alloc_shared_dr(struct mlx5_priv *priv)
+@@ -319,7 +319,17 @@ mlx5_alloc_shared_dr(struct mlx5_priv *priv)
@@ -75 +77 @@
-@@ -361,9 +371,9 @@ error:
+@@ -355,9 +365,9 @@ error:
@@ -88 +90 @@
-@@ -418,9 +428,9 @@ mlx5_os_free_shared_dr(struct mlx5_priv *priv)
+@@ -412,9 +422,9 @@ mlx5_os_free_shared_dr(struct mlx5_priv *priv)
@@ -102 +104 @@
-index 6faba4fbb1..0f69f9d125 100644
+index 9bf1bf3146..97f8a016b4 100644
@@ -105 +107 @@
-@@ -729,7 +729,7 @@ struct mlx5_dev_ctx_shared {
+@@ -719,7 +719,7 @@ struct mlx5_dev_ctx_shared {
@@ -115 +117 @@
-index 533dadf07b..691595942f 100644
+index 94550e5a73..b1c8a95e8a 100644
@@ -118 +120 @@
-@@ -12059,11 +12059,19 @@ flow_dv_apply(struct rte_eth_dev *dev, struct rte_flow *flow,
+@@ -10839,11 +10839,19 @@ flow_dv_apply(struct rte_eth_dev *dev, struct rte_flow *flow,
@@ -139 +141 @@
-@@ -13849,7 +13857,8 @@ mlx5_flow_dv_discover_counter_offset_support(struct rte_eth_dev *dev)
+@@ -12610,7 +12618,8 @@ mlx5_flow_dv_discover_counter_offset_support(struct rte_eth_dev *dev)
Previous message (by thread): [dpdk-stable] patch 'net/mlx5: fix missing shared RSS hash types' has been queued to stable release 20.11.2
Next message (by thread): [dpdk-stable] patch 'net/bnxt: fix double free in port start failure' has been queued to stable release 20.11.2
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the stable mailing list