DPDK patches and discussions
 help / color / mirror / Atom feed
From: Suanming Mou <suanmingm@nvidia.com>
To: Matan Azrad <matan@nvidia.com>,
	Viacheslav Ovsiienko <viacheslavo@nvidia.com>,
	Ori Kam <orika@nvidia.com>
Cc: <dev@dpdk.org>, <rasland@nvidia.com>,
	Dariusz Sosnowski <dsosnowski@nvidia.com>, <stable@dpdk.org>
Subject: [PATCH 1/2] net/mlx5: fix missing flow rules for external SQ
Date: Thu, 9 Nov 2023 16:55:46 +0800	[thread overview]
Message-ID: <20231109085547.1313003-2-suanmingm@nvidia.com> (raw)
In-Reply-To: <20231109085547.1313003-1-suanmingm@nvidia.com>

From: Dariusz Sosnowski <dsosnowski@nvidia.com>

mlx5 PMD exposes a capability to register externally created SQs
as if it was an SQ of a given representor port. Registration would
cause a creation of control flow rules in FDB domain used to
forward traffic betwen SQ and destination represented port.

Before this patch, if representor matching was enabled (device argument
repr_matching_en is equal to 1, default configuration), then during
registration of external SQs, mlx5 PMD would not create control flow
rules in NIC Tx domain. This caused an issue with packet metadata.
If a packet sent on external SQ had packet metadata attached, then
it would be lost when it would go from NIC Tx to FDB domain.

With representor matching disabled everything is working correctly,
because in that mode there is a single global flow rule for preserving
packet metadata. This flow rule matches whole traffic on NIC Tx domain.
With representor matching enabled, NIC Tx flow rules are created per SQ.

This patch fixes that behavior. If representor matching is enabled, then
NIC Tx flow rules are created for each external SQ registered in
rte_pmd_mlx5_external_sq_enable().

This patch also adds an ability to destroy SQ miss flow rules for a
given port and SQ number. This is required for error rollback flow in
rte_pmd_mlx5_external_sq_enable().

Fixes: 26e1eaf2dac4 ("net/mlx5: support device control for E-Switch default rule")

Cc: stable@dpdk.org

Signed-off-by: Dariusz Sosnowski <dsosnowski@nvidia.com>
---
 drivers/net/mlx5/mlx5.h         |  40 ++++++++++++
 drivers/net/mlx5/mlx5_flow.h    |   2 +
 drivers/net/mlx5/mlx5_flow_hw.c | 107 +++++++++++++++++++++++++++++---
 drivers/net/mlx5/mlx5_txq.c     |  12 +++-
 4 files changed, 149 insertions(+), 12 deletions(-)

diff --git a/drivers/net/mlx5/mlx5.h b/drivers/net/mlx5/mlx5.h
index f5eacb2c67..45ad0701f1 100644
--- a/drivers/net/mlx5/mlx5.h
+++ b/drivers/net/mlx5/mlx5.h
@@ -1705,10 +1705,50 @@ struct mlx5_obj_ops {
 
 #define MLX5_RSS_HASH_FIELDS_LEN RTE_DIM(mlx5_rss_hash_fields)
 
+enum mlx5_hw_ctrl_flow_type {
+	MLX5_HW_CTRL_FLOW_TYPE_GENERAL,
+	MLX5_HW_CTRL_FLOW_TYPE_SQ_MISS_ROOT,
+	MLX5_HW_CTRL_FLOW_TYPE_SQ_MISS,
+	MLX5_HW_CTRL_FLOW_TYPE_DEFAULT_JUMP,
+	MLX5_HW_CTRL_FLOW_TYPE_TX_META_COPY,
+	MLX5_HW_CTRL_FLOW_TYPE_TX_REPR_MATCH,
+	MLX5_HW_CTRL_FLOW_TYPE_LACP_RX,
+	MLX5_HW_CTRL_FLOW_TYPE_DEFAULT_RX_RSS,
+};
+
+/** Additional info about control flow rule. */
+struct mlx5_hw_ctrl_flow_info {
+	/** Determines the kind of control flow rule. */
+	enum mlx5_hw_ctrl_flow_type type;
+	union {
+		/**
+		 * If control flow is a SQ miss flow (root or not),
+		 * then fields contains matching SQ number.
+		 */
+		uint32_t esw_mgr_sq;
+		/**
+		 * If control flow is a Tx representor matching,
+		 * then fields contains matching SQ number.
+		 */
+		uint32_t tx_repr_sq;
+	};
+};
+
+/** Entry for tracking control flow rules in HWS. */
 struct mlx5_hw_ctrl_flow {
 	LIST_ENTRY(mlx5_hw_ctrl_flow) next;
+	/**
+	 * Owner device is a port on behalf of which flow rule was created.
+	 *
+	 * It's different from the port which really created the flow rule
+	 * if and only if flow rule is created on transfer proxy port
+	 * on behalf of representor port.
+	 */
 	struct rte_eth_dev *owner_dev;
+	/** Pointer to flow rule handle. */
 	struct rte_flow *flow;
+	/** Additional information about the control flow rule. */
+	struct mlx5_hw_ctrl_flow_info info;
 };
 
 /*
diff --git a/drivers/net/mlx5/mlx5_flow.h b/drivers/net/mlx5/mlx5_flow.h
index 094be12715..d57b3b5465 100644
--- a/drivers/net/mlx5/mlx5_flow.h
+++ b/drivers/net/mlx5/mlx5_flow.h
@@ -2875,6 +2875,8 @@ int mlx5_flow_hw_flush_ctrl_flows(struct rte_eth_dev *dev);
 
 int mlx5_flow_hw_esw_create_sq_miss_flow(struct rte_eth_dev *dev,
 					 uint32_t sqn);
+int mlx5_flow_hw_esw_destroy_sq_miss_flow(struct rte_eth_dev *dev,
+					  uint32_t sqn);
 int mlx5_flow_hw_esw_create_default_jump_flow(struct rte_eth_dev *dev);
 int mlx5_flow_hw_create_tx_default_mreg_copy_flow(struct rte_eth_dev *dev);
 int mlx5_flow_hw_tx_repr_matching_flow(struct rte_eth_dev *dev, uint32_t sqn);
diff --git a/drivers/net/mlx5/mlx5_flow_hw.c b/drivers/net/mlx5/mlx5_flow_hw.c
index f57126e2ff..d512889682 100644
--- a/drivers/net/mlx5/mlx5_flow_hw.c
+++ b/drivers/net/mlx5/mlx5_flow_hw.c
@@ -11341,6 +11341,8 @@ const struct mlx5_flow_driver_ops mlx5_flow_hw_drv_ops = {
  *   Pointer to flow rule actions.
  * @param action_template_idx
  *   Index of an action template associated with @p table.
+ * @param info
+ *   Additional info about control flow rule.
  *
  * @return
  *   0 on success, negative errno value otherwise and rte_errno set.
@@ -11352,7 +11354,8 @@ flow_hw_create_ctrl_flow(struct rte_eth_dev *owner_dev,
 			 struct rte_flow_item items[],
 			 uint8_t item_template_idx,
 			 struct rte_flow_action actions[],
-			 uint8_t action_template_idx)
+			 uint8_t action_template_idx,
+			 struct mlx5_hw_ctrl_flow_info *info)
 {
 	struct mlx5_priv *priv = proxy_dev->data->dev_private;
 	uint32_t queue = CTRL_QUEUE_ID(priv);
@@ -11399,6 +11402,10 @@ flow_hw_create_ctrl_flow(struct rte_eth_dev *owner_dev,
 	}
 	entry->owner_dev = owner_dev;
 	entry->flow = flow;
+	if (info)
+		entry->info = *info;
+	else
+		entry->info.type = MLX5_HW_CTRL_FLOW_TYPE_GENERAL;
 	LIST_INSERT_HEAD(&priv->hw_ctrl_flows, entry, next);
 	rte_spinlock_unlock(&priv->hw_ctrl_lock);
 	return 0;
@@ -11602,6 +11609,10 @@ mlx5_flow_hw_esw_create_sq_miss_flow(struct rte_eth_dev *dev, uint32_t sqn)
 	};
 	struct rte_flow_item items[3] = { { 0 } };
 	struct rte_flow_action actions[3] = { { 0 } };
+	struct mlx5_hw_ctrl_flow_info flow_info = {
+		.type = MLX5_HW_CTRL_FLOW_TYPE_SQ_MISS_ROOT,
+		.esw_mgr_sq = sqn,
+	};
 	struct rte_eth_dev *proxy_dev;
 	struct mlx5_priv *proxy_priv;
 	uint16_t proxy_port_id = dev->data->port_id;
@@ -11657,7 +11668,7 @@ mlx5_flow_hw_esw_create_sq_miss_flow(struct rte_eth_dev *dev, uint32_t sqn)
 		.type = RTE_FLOW_ACTION_TYPE_END,
 	};
 	ret = flow_hw_create_ctrl_flow(dev, proxy_dev, proxy_priv->hw_esw_sq_miss_root_tbl,
-				       items, 0, actions, 0);
+				       items, 0, actions, 0, &flow_info);
 	if (ret) {
 		DRV_LOG(ERR, "Port %u failed to create root SQ miss flow rule for SQ %u, ret %d",
 			port_id, sqn, ret);
@@ -11686,8 +11697,9 @@ mlx5_flow_hw_esw_create_sq_miss_flow(struct rte_eth_dev *dev, uint32_t sqn)
 	actions[1] = (struct rte_flow_action){
 		.type = RTE_FLOW_ACTION_TYPE_END,
 	};
+	flow_info.type = MLX5_HW_CTRL_FLOW_TYPE_SQ_MISS;
 	ret = flow_hw_create_ctrl_flow(dev, proxy_dev, proxy_priv->hw_esw_sq_miss_tbl,
-				       items, 0, actions, 0);
+				       items, 0, actions, 0, &flow_info);
 	if (ret) {
 		DRV_LOG(ERR, "Port %u failed to create HWS SQ miss flow rule for SQ %u, ret %d",
 			port_id, sqn, ret);
@@ -11696,6 +11708,58 @@ mlx5_flow_hw_esw_create_sq_miss_flow(struct rte_eth_dev *dev, uint32_t sqn)
 	return 0;
 }
 
+static bool
+flow_hw_is_matching_sq_miss_flow(struct mlx5_hw_ctrl_flow *cf,
+				 struct rte_eth_dev *dev,
+				 uint32_t sqn)
+{
+	if (cf->owner_dev != dev)
+		return false;
+	if (cf->info.type == MLX5_HW_CTRL_FLOW_TYPE_SQ_MISS_ROOT && cf->info.esw_mgr_sq == sqn)
+		return true;
+	if (cf->info.type == MLX5_HW_CTRL_FLOW_TYPE_SQ_MISS && cf->info.esw_mgr_sq == sqn)
+		return true;
+	return false;
+}
+
+int
+mlx5_flow_hw_esw_destroy_sq_miss_flow(struct rte_eth_dev *dev, uint32_t sqn)
+{
+	uint16_t port_id = dev->data->port_id;
+	uint16_t proxy_port_id = dev->data->port_id;
+	struct rte_eth_dev *proxy_dev;
+	struct mlx5_priv *proxy_priv;
+	struct mlx5_hw_ctrl_flow *cf;
+	struct mlx5_hw_ctrl_flow *cf_next;
+	int ret;
+
+	ret = rte_flow_pick_transfer_proxy(port_id, &proxy_port_id, NULL);
+	if (ret) {
+		DRV_LOG(ERR, "Unable to pick transfer proxy port for port %u. Transfer proxy "
+			     "port must be present for default SQ miss flow rules to exist.",
+			     port_id);
+		return ret;
+	}
+	proxy_dev = &rte_eth_devices[proxy_port_id];
+	proxy_priv = proxy_dev->data->dev_private;
+	if (!proxy_priv->dr_ctx)
+		return 0;
+	if (!proxy_priv->hw_esw_sq_miss_root_tbl ||
+	    !proxy_priv->hw_esw_sq_miss_tbl)
+		return 0;
+	cf = LIST_FIRST(&proxy_priv->hw_ctrl_flows);
+	while (cf != NULL) {
+		cf_next = LIST_NEXT(cf, next);
+		if (flow_hw_is_matching_sq_miss_flow(cf, dev, sqn)) {
+			claim_zero(flow_hw_destroy_ctrl_flow(proxy_dev, cf->flow));
+			LIST_REMOVE(cf, next);
+			mlx5_free(cf);
+		}
+		cf = cf_next;
+	}
+	return 0;
+}
+
 int
 mlx5_flow_hw_esw_create_default_jump_flow(struct rte_eth_dev *dev)
 {
@@ -11724,6 +11788,9 @@ mlx5_flow_hw_esw_create_default_jump_flow(struct rte_eth_dev *dev)
 			.type = RTE_FLOW_ACTION_TYPE_END,
 		}
 	};
+	struct mlx5_hw_ctrl_flow_info flow_info = {
+		.type = MLX5_HW_CTRL_FLOW_TYPE_DEFAULT_JUMP,
+	};
 	struct rte_eth_dev *proxy_dev;
 	struct mlx5_priv *proxy_priv;
 	uint16_t proxy_port_id = dev->data->port_id;
@@ -11754,7 +11821,7 @@ mlx5_flow_hw_esw_create_default_jump_flow(struct rte_eth_dev *dev)
 	}
 	return flow_hw_create_ctrl_flow(dev, proxy_dev,
 					proxy_priv->hw_esw_zero_tbl,
-					items, 0, actions, 0);
+					items, 0, actions, 0, &flow_info);
 }
 
 int
@@ -11800,13 +11867,16 @@ mlx5_flow_hw_create_tx_default_mreg_copy_flow(struct rte_eth_dev *dev)
 			.type = RTE_FLOW_ACTION_TYPE_END,
 		},
 	};
+	struct mlx5_hw_ctrl_flow_info flow_info = {
+		.type = MLX5_HW_CTRL_FLOW_TYPE_TX_META_COPY,
+	};
 
 	MLX5_ASSERT(priv->master);
 	if (!priv->dr_ctx || !priv->hw_tx_meta_cpy_tbl)
 		return 0;
 	return flow_hw_create_ctrl_flow(dev, dev,
 					priv->hw_tx_meta_cpy_tbl,
-					eth_all, 0, copy_reg_action, 0);
+					eth_all, 0, copy_reg_action, 0, &flow_info);
 }
 
 int
@@ -11835,6 +11905,10 @@ mlx5_flow_hw_tx_repr_matching_flow(struct rte_eth_dev *dev, uint32_t sqn)
 		{ .type = RTE_FLOW_ACTION_TYPE_END },
 		{ .type = RTE_FLOW_ACTION_TYPE_END },
 	};
+	struct mlx5_hw_ctrl_flow_info flow_info = {
+		.type = MLX5_HW_CTRL_FLOW_TYPE_TX_REPR_MATCH,
+		.tx_repr_sq = sqn,
+	};
 
 	/* It is assumed that caller checked for representor matching. */
 	MLX5_ASSERT(priv->sh->config.repr_matching);
@@ -11860,7 +11934,7 @@ mlx5_flow_hw_tx_repr_matching_flow(struct rte_eth_dev *dev, uint32_t sqn)
 		actions[2].type = RTE_FLOW_ACTION_TYPE_JUMP;
 	}
 	return flow_hw_create_ctrl_flow(dev, dev, priv->hw_tx_repr_tagging_tbl,
-					items, 0, actions, 0);
+					items, 0, actions, 0, &flow_info);
 }
 
 static uint32_t
@@ -11975,6 +12049,9 @@ __flow_hw_ctrl_flows_single(struct rte_eth_dev *dev,
 		{ .type = RTE_FLOW_ACTION_TYPE_RSS },
 		{ .type = RTE_FLOW_ACTION_TYPE_END },
 	};
+	struct mlx5_hw_ctrl_flow_info flow_info = {
+		.type = MLX5_HW_CTRL_FLOW_TYPE_DEFAULT_RX_RSS,
+	};
 
 	if (!eth_spec)
 		return -EINVAL;
@@ -11988,7 +12065,7 @@ __flow_hw_ctrl_flows_single(struct rte_eth_dev *dev,
 	items[3] = flow_hw_get_ctrl_rx_l4_item(rss_type);
 	items[4] = (struct rte_flow_item){ .type = RTE_FLOW_ITEM_TYPE_END };
 	/* Without VLAN filtering, only a single flow rule must be created. */
-	return flow_hw_create_ctrl_flow(dev, dev, tbl, items, 0, actions, 0);
+	return flow_hw_create_ctrl_flow(dev, dev, tbl, items, 0, actions, 0, &flow_info);
 }
 
 static int
@@ -12004,6 +12081,9 @@ __flow_hw_ctrl_flows_single_vlan(struct rte_eth_dev *dev,
 		{ .type = RTE_FLOW_ACTION_TYPE_RSS },
 		{ .type = RTE_FLOW_ACTION_TYPE_END },
 	};
+	struct mlx5_hw_ctrl_flow_info flow_info = {
+		.type = MLX5_HW_CTRL_FLOW_TYPE_DEFAULT_RX_RSS,
+	};
 	unsigned int i;
 
 	if (!eth_spec)
@@ -12026,7 +12106,7 @@ __flow_hw_ctrl_flows_single_vlan(struct rte_eth_dev *dev,
 		};
 
 		items[1].spec = &vlan_spec;
-		if (flow_hw_create_ctrl_flow(dev, dev, tbl, items, 0, actions, 0))
+		if (flow_hw_create_ctrl_flow(dev, dev, tbl, items, 0, actions, 0, &flow_info))
 			return -rte_errno;
 	}
 	return 0;
@@ -12044,6 +12124,9 @@ __flow_hw_ctrl_flows_unicast(struct rte_eth_dev *dev,
 		{ .type = RTE_FLOW_ACTION_TYPE_RSS },
 		{ .type = RTE_FLOW_ACTION_TYPE_END },
 	};
+	struct mlx5_hw_ctrl_flow_info flow_info = {
+		.type = MLX5_HW_CTRL_FLOW_TYPE_DEFAULT_RX_RSS,
+	};
 	const struct rte_ether_addr cmp = {
 		.addr_bytes = "\x00\x00\x00\x00\x00\x00",
 	};
@@ -12067,7 +12150,7 @@ __flow_hw_ctrl_flows_unicast(struct rte_eth_dev *dev,
 		if (!memcmp(mac, &cmp, sizeof(*mac)))
 			continue;
 		memcpy(&eth_spec.hdr.dst_addr.addr_bytes, mac->addr_bytes, RTE_ETHER_ADDR_LEN);
-		if (flow_hw_create_ctrl_flow(dev, dev, tbl, items, 0, actions, 0))
+		if (flow_hw_create_ctrl_flow(dev, dev, tbl, items, 0, actions, 0, &flow_info))
 			return -rte_errno;
 	}
 	return 0;
@@ -12086,6 +12169,9 @@ __flow_hw_ctrl_flows_unicast_vlan(struct rte_eth_dev *dev,
 		{ .type = RTE_FLOW_ACTION_TYPE_RSS },
 		{ .type = RTE_FLOW_ACTION_TYPE_END },
 	};
+	struct mlx5_hw_ctrl_flow_info flow_info = {
+		.type = MLX5_HW_CTRL_FLOW_TYPE_DEFAULT_RX_RSS,
+	};
 	const struct rte_ether_addr cmp = {
 		.addr_bytes = "\x00\x00\x00\x00\x00\x00",
 	};
@@ -12117,7 +12203,8 @@ __flow_hw_ctrl_flows_unicast_vlan(struct rte_eth_dev *dev,
 			};
 
 			items[1].spec = &vlan_spec;
-			if (flow_hw_create_ctrl_flow(dev, dev, tbl, items, 0, actions, 0))
+			if (flow_hw_create_ctrl_flow(dev, dev, tbl, items, 0, actions, 0,
+						     &flow_info))
 				return -rte_errno;
 		}
 	}
diff --git a/drivers/net/mlx5/mlx5_txq.c b/drivers/net/mlx5/mlx5_txq.c
index b584055fa8..ccdf2ffb14 100644
--- a/drivers/net/mlx5/mlx5_txq.c
+++ b/drivers/net/mlx5/mlx5_txq.c
@@ -1310,8 +1310,16 @@ rte_pmd_mlx5_external_sq_enable(uint16_t port_id, uint32_t sq_num)
 		return -rte_errno;
 	}
 #ifdef HAVE_MLX5_HWS_SUPPORT
-	if (priv->sh->config.dv_flow_en == 2)
-		return mlx5_flow_hw_esw_create_sq_miss_flow(dev, sq_num);
+	if (priv->sh->config.dv_flow_en == 2) {
+		if (mlx5_flow_hw_esw_create_sq_miss_flow(dev, sq_num))
+			return -rte_errno;
+		if (priv->sh->config.repr_matching &&
+		    mlx5_flow_hw_tx_repr_matching_flow(dev, sq_num)) {
+			mlx5_flow_hw_esw_destroy_sq_miss_flow(dev, sq_num);
+			return -rte_errno;
+		}
+		return 0;
+	}
 #endif
 	flow = mlx5_flow_create_devx_sq_miss_flow(dev, sq_num);
 	if (flow > 0)
-- 
2.34.1


  reply	other threads:[~2023-11-09  8:56 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-11-09  8:55 [PATCH 0/2] net/mlx5: fix " Suanming Mou
2023-11-09  8:55 ` Suanming Mou [this message]
2023-11-09  8:55 ` [PATCH 2/2] net/mlx5: fix destroying external representor matched flows Suanming Mou
2023-11-12 14:27 ` [PATCH 0/2] net/mlx5: fix flow rules for external SQ Raslan Darawsheh

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20231109085547.1313003-2-suanmingm@nvidia.com \
    --to=suanmingm@nvidia.com \
    --cc=dev@dpdk.org \
    --cc=dsosnowski@nvidia.com \
    --cc=matan@nvidia.com \
    --cc=orika@nvidia.com \
    --cc=rasland@nvidia.com \
    --cc=stable@dpdk.org \
    --cc=viacheslavo@nvidia.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).