Merge branch '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue

Tony Nguyen says:

====================
ice: use less resources in switchdev

Michal Swiatkowski says:

Switchdev is using one queue per created port representor. This can
quickly lead to Rx queue shortage, as with subfunction support user
can create high number of PRs.

Save one MSI-X and 'number of PRs' * 1 queues.
Refactor switchdev slow-path to use less resources (even no additional
resources). Do this by removing control plane VSI and move its
functionality to PF VSI. Even with current solution PF is acting like
uplink and can't be used simultaneously for other use cases (adding
filters can break slow-path).

In short, do Tx via PF VSI and Rx packets using PF resources. Rx needs
additional code in interrupt handler to choose correct PR netdev.
Previous solution had to queue filters, it was way more elegant but
needed one queue per PRs. Beside that this refactor mostly simplifies
switchdev configuration.

* '100GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/next-queue:
  ice: count representor stats
  ice: do switchdev slow-path Rx using PF VSI
  ice: change repr::id values
  ice: remove switchdev control plane VSI
  ice: control default Tx rule in lag
  ice: default Tx rule instead of to queue
  ice: do Tx through PF netdev in slow-path
  ice: remove eswitch changing queues algorithm
====================

Link: https://lore.kernel.org/r/20240325202623.1012287-1-anthony.l.nguyen@intel.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

drivers/net/ethernet/intel/ice/ice.h

@@ -522,17 +522,10 @@ enum ice_misc_thread_tasks {
 };
 
 struct ice_eswitch {
-	struct ice_vsi *control_vsi;
 	struct ice_vsi *uplink_vsi;
 	struct ice_esw_br_offloads *br_offloads;
 	struct xarray reprs;
 	bool is_running;
-	/* struct to allow cp queues management optimization */
-	struct {
-		int to_reach;
-		int value;
-		bool is_reaching;
-	} qs;
 };
 
 struct ice_agg_node {

drivers/net/ethernet/intel/ice/ice_base.c

@@ -263,30 +263,6 @@ static u16 ice_calc_txq_handle(struct ice_vsi *vsi, struct ice_tx_ring *ring, u8
 	return ring->q_index - vsi->tc_cfg.tc_info[tc].qoffset;
 }
 
-/**
- * ice_eswitch_calc_txq_handle
- * @ring: pointer to ring which unique index is needed
- *
- * To correctly work with many netdevs ring->q_index of Tx rings on switchdev
- * VSI can repeat. Hardware ring setup requires unique q_index. Calculate it
- * here by finding index in vsi->tx_rings of this ring.
- *
- * Return ICE_INVAL_Q_INDEX when index wasn't found. Should never happen,
- * because VSI is get from ring->vsi, so it has to be present in this VSI.
- */
-static u16 ice_eswitch_calc_txq_handle(struct ice_tx_ring *ring)
-{
-	const struct ice_vsi *vsi = ring->vsi;
-	int i;
-
-	ice_for_each_txq(vsi, i) {
-		if (vsi->tx_rings[i] == ring)
-			return i;
-	}
-
-	return ICE_INVAL_Q_INDEX;
-}
-
 /**
  * ice_cfg_xps_tx_ring - Configure XPS for a Tx ring
  * @ring: The Tx ring to configure
@@ -353,9 +329,6 @@ ice_setup_tx_ctx(struct ice_tx_ring *ring, struct ice_tlan_ctx *tlan_ctx, u16 pf
 		tlan_ctx->vmvf_num = hw->func_caps.vf_base_id + vsi->vf->vf_id;
 		tlan_ctx->vmvf_type = ICE_TLAN_CTX_VMVF_TYPE_VF;
 		break;
-	case ICE_VSI_SWITCHDEV_CTRL:
-		tlan_ctx->vmvf_type = ICE_TLAN_CTX_VMVF_TYPE_VMQ;
-		break;
 	default:
 		return;
 	}
@@ -479,6 +452,14 @@ static int ice_setup_rx_ctx(struct ice_rx_ring *ring)
 	/* Rx queue threshold in units of 64 */
 	rlan_ctx.lrxqthresh = 1;
 
+	/* PF acts as uplink for switchdev; set flex descriptor with src_vsi
+	 * metadata and flags to allow redirecting to PR netdev
+	 */
+	if (ice_is_eswitch_mode_switchdev(vsi->back)) {
+		ring->flags |= ICE_RX_FLAGS_MULTIDEV;
+		rxdid = ICE_RXDID_FLEX_NIC_2;
+	}
+
 	/* Enable Flexible Descriptors in the queue context which
 	 * allows this driver to select a specific receive descriptor format
 	 * increasing context priority to pick up profile ID; default is 0x01;
@@ -919,14 +900,7 @@ ice_vsi_cfg_txq(struct ice_vsi *vsi, struct ice_tx_ring *ring,
 	/* Add unique software queue handle of the Tx queue per
 	 * TC into the VSI Tx ring
 	 */
-	if (vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
-		ring->q_handle = ice_eswitch_calc_txq_handle(ring);
-
-		if (ring->q_handle == ICE_INVAL_Q_INDEX)
-			return -ENODEV;
-	} else {
-		ring->q_handle = ice_calc_txq_handle(vsi, ring, tc);
-	}
+	ring->q_handle = ice_calc_txq_handle(vsi, ring, tc);
 
 	if (ch)
 		status = ice_ena_vsi_txq(vsi->port_info, ch->ch_vsi->idx, 0,

drivers/net/ethernet/intel/ice/ice_dcb_lib.c

@@ -291,7 +291,6 @@ static void ice_dcb_ena_dis_vsi(struct ice_pf *pf, bool ena, bool locked)
 
 		switch (vsi->type) {
 		case ICE_VSI_CHNL:
-		case ICE_VSI_SWITCHDEV_CTRL:
 		case ICE_VSI_PF:
 			if (ena)
 				ice_ena_vsi(vsi, locked);
@@ -776,8 +775,7 @@ void ice_pf_dcb_recfg(struct ice_pf *pf, bool locked)
 		/* no need to proceed with remaining cfg if it is CHNL
 		 * or switchdev VSI
 		 */
-		if (vsi->type == ICE_VSI_CHNL ||
-		    vsi->type == ICE_VSI_SWITCHDEV_CTRL)
+		if (vsi->type == ICE_VSI_CHNL)
 			continue;
 
 		ice_vsi_map_rings_to_vectors(vsi);

drivers/net/ethernet/intel/ice/ice_eswitch.h

@@ -10,7 +10,7 @@
 void ice_eswitch_detach(struct ice_pf *pf, struct ice_vf *vf);
 int
 ice_eswitch_attach(struct ice_pf *pf, struct ice_vf *vf);
-int ice_eswitch_rebuild(struct ice_pf *pf);
+void ice_eswitch_rebuild(struct ice_pf *pf);
 
 int ice_eswitch_mode_get(struct devlink *devlink, u16 *mode);
 int
@@ -26,7 +26,8 @@ void ice_eswitch_set_target_vsi(struct sk_buff *skb,
 				struct ice_tx_offload_params *off);
 netdev_tx_t
 ice_eswitch_port_start_xmit(struct sk_buff *skb, struct net_device *netdev);
-void ice_eswitch_reserve_cp_queues(struct ice_pf *pf, int change);
+struct net_device *ice_eswitch_get_target(struct ice_rx_ring *rx_ring,
+					  union ice_32b_rx_flex_desc *rx_desc);
 #else /* CONFIG_ICE_SWITCHDEV */
 static inline void ice_eswitch_detach(struct ice_pf *pf, struct ice_vf *vf) { }
 
@@ -78,7 +79,11 @@ ice_eswitch_port_start_xmit(struct sk_buff *skb, struct net_device *netdev)
 	return NETDEV_TX_BUSY;
 }
 
-static inline void
-ice_eswitch_reserve_cp_queues(struct ice_pf *pf, int change) { }
+static inline struct net_device *
+ice_eswitch_get_target(struct ice_rx_ring *rx_ring,
+		       union ice_32b_rx_flex_desc *rx_desc)
+{
+	return rx_ring->netdev;
+}
 #endif /* CONFIG_ICE_SWITCHDEV */
 #endif /* _ICE_ESWITCH_H_ */

drivers/net/ethernet/intel/ice/ice_lag.c

@@ -202,11 +202,12 @@ static struct ice_lag *ice_lag_find_primary(struct ice_lag *lag)
  * @act: rule action
  * @recipe_id: recipe id for the new rule
  * @rule_idx: pointer to rule index
+ * @direction: ICE_FLTR_RX or ICE_FLTR_TX
  * @add: boolean on whether we are adding filters
  */
 static int
 ice_lag_cfg_fltr(struct ice_lag *lag, u32 act, u16 recipe_id, u16 *rule_idx,
-		 bool add)
+		 u8 direction, bool add)
 {
 	struct ice_sw_rule_lkup_rx_tx *s_rule;
 	u16 s_rule_sz, vsi_num;
@@ -231,9 +232,16 @@ ice_lag_cfg_fltr(struct ice_lag *lag, u32 act, u16 recipe_id, u16 *rule_idx,
 
 		act |= FIELD_PREP(ICE_SINGLE_ACT_VSI_ID_M, vsi_num);
 
-		s_rule->hdr.type = cpu_to_le16(ICE_AQC_SW_RULES_T_LKUP_RX);
 		s_rule->recipe_id = cpu_to_le16(recipe_id);
-		s_rule->src = cpu_to_le16(hw->port_info->lport);
+		if (direction == ICE_FLTR_RX) {
+			s_rule->hdr.type =
+				cpu_to_le16(ICE_AQC_SW_RULES_T_LKUP_RX);
+			s_rule->src = cpu_to_le16(hw->port_info->lport);
+		} else {
+			s_rule->hdr.type =
+				cpu_to_le16(ICE_AQC_SW_RULES_T_LKUP_TX);
+			s_rule->src = cpu_to_le16(vsi_num);
+		}
 		s_rule->act = cpu_to_le32(act);
 		s_rule->hdr_len = cpu_to_le16(DUMMY_ETH_HDR_LEN);
 		opc = ice_aqc_opc_add_sw_rules;
@@ -266,9 +274,27 @@ ice_lag_cfg_dflt_fltr(struct ice_lag *lag, bool add)
 {
 	u32 act = ICE_SINGLE_ACT_VSI_FORWARDING |
 		ICE_SINGLE_ACT_VALID_BIT | ICE_SINGLE_ACT_LAN_ENABLE;
+	int err;
+
+	err = ice_lag_cfg_fltr(lag, act, lag->pf_recipe, &lag->pf_rx_rule_id,
+			       ICE_FLTR_RX, add);
+	if (err)
+		goto err_rx;
 
-	return ice_lag_cfg_fltr(lag, act, lag->pf_recipe,
-				&lag->pf_rule_id, add);
+	act = ICE_SINGLE_ACT_VSI_FORWARDING | ICE_SINGLE_ACT_VALID_BIT |
+	      ICE_SINGLE_ACT_LB_ENABLE;
+	err = ice_lag_cfg_fltr(lag, act, lag->pf_recipe, &lag->pf_tx_rule_id,
+			       ICE_FLTR_TX, add);
+	if (err)
+		goto err_tx;
+
+	return 0;
+
+err_tx:
+	ice_lag_cfg_fltr(lag, act, lag->pf_recipe, &lag->pf_rx_rule_id,
+			 ICE_FLTR_RX, !add);
+err_rx:
+	return err;
 }
 
 /**
@@ -284,7 +310,7 @@ ice_lag_cfg_drop_fltr(struct ice_lag *lag, bool add)
 		  ICE_SINGLE_ACT_DROP;
 
 	return ice_lag_cfg_fltr(lag, act, lag->lport_recipe,
-				&lag->lport_rule_idx, add);
+				&lag->lport_rule_idx, ICE_FLTR_RX, add);
 }
 
 /**
@@ -310,7 +336,7 @@ ice_lag_cfg_pf_fltrs(struct ice_lag *lag, void *ptr)
 	dev = ice_pf_to_dev(lag->pf);
 
 	/* interface not active - remove old default VSI rule */
-	if (bonding_info->slave.state && lag->pf_rule_id) {
+	if (bonding_info->slave.state && lag->pf_rx_rule_id) {
 		if (ice_lag_cfg_dflt_fltr(lag, false))
 			dev_err(dev, "Error removing old default VSI filter\n");
 		if (ice_lag_cfg_drop_fltr(lag, true))
@@ -319,7 +345,7 @@ ice_lag_cfg_pf_fltrs(struct ice_lag *lag, void *ptr)
 	}
 
 	/* interface becoming active - add new default VSI rule */
-	if (!bonding_info->slave.state && !lag->pf_rule_id) {
+	if (!bonding_info->slave.state && !lag->pf_rx_rule_id) {
 		if (ice_lag_cfg_dflt_fltr(lag, true))
 			dev_err(dev, "Error adding new default VSI filter\n");
 		if (lag->lport_rule_idx && ice_lag_cfg_drop_fltr(lag, false))
@@ -714,8 +740,7 @@ static void ice_lag_move_vf_nodes(struct ice_lag *lag, u8 oldport, u8 newport)
 
 	pf = lag->pf;
 	ice_for_each_vsi(pf, i)
-		if (pf->vsi[i] && (pf->vsi[i]->type == ICE_VSI_VF ||
-				   pf->vsi[i]->type == ICE_VSI_SWITCHDEV_CTRL))
+		if (pf->vsi[i] && pf->vsi[i]->type == ICE_VSI_VF)
 			ice_lag_move_single_vf_nodes(lag, oldport, newport, i);
 }
 
@@ -953,8 +978,7 @@ ice_lag_reclaim_vf_nodes(struct ice_lag *lag, struct ice_hw *src_hw)
 
 	pf = lag->pf;
 	ice_for_each_vsi(pf, i)
-		if (pf->vsi[i] && (pf->vsi[i]->type == ICE_VSI_VF ||
-				   pf->vsi[i]->type == ICE_VSI_SWITCHDEV_CTRL))
+		if (pf->vsi[i] && pf->vsi[i]->type == ICE_VSI_VF)
 			ice_for_each_traffic_class(tc)
 				ice_lag_reclaim_vf_tc(lag, src_hw, i, tc);
 }
@@ -1976,8 +2000,7 @@ ice_lag_move_vf_nodes_sync(struct ice_lag *lag, struct ice_hw *dest_hw)
 
 	pf = lag->pf;
 	ice_for_each_vsi(pf, i)
-		if (pf->vsi[i] && (pf->vsi[i]->type == ICE_VSI_VF ||
-				   pf->vsi[i]->type == ICE_VSI_SWITCHDEV_CTRL))
+		if (pf->vsi[i] && pf->vsi[i]->type == ICE_VSI_VF)
 			ice_for_each_traffic_class(tc)
 				ice_lag_move_vf_nodes_tc_sync(lag, dest_hw, i,
 							      tc);
@@ -2149,7 +2172,7 @@ void ice_lag_rebuild(struct ice_pf *pf)
 
 	ice_lag_cfg_cp_fltr(lag, true);
 
-	if (lag->pf_rule_id)
+	if (lag->pf_rx_rule_id)
 		if (ice_lag_cfg_dflt_fltr(lag, true))
 			dev_err(ice_pf_to_dev(pf), "Error adding default VSI rule in rebuild\n");
 

drivers/net/ethernet/intel/ice/ice_lag.h

@@ -43,7 +43,8 @@ struct ice_lag {
 	u8 primary:1; /* this is primary */
 	u16 pf_recipe;
 	u16 lport_recipe;
-	u16 pf_rule_id;
+	u16 pf_rx_rule_id;
+	u16 pf_tx_rule_id;
 	u16 cp_rule_idx;
 	u16 lport_rule_idx;
 	u8 role;

drivers/net/ethernet/intel/ice/ice_lib.c

@@ -27,8 +27,6 @@ const char *ice_vsi_type_str(enum ice_vsi_type vsi_type)
 		return "ICE_VSI_CHNL";
 	case ICE_VSI_LB:
 		return "ICE_VSI_LB";
-	case ICE_VSI_SWITCHDEV_CTRL:
-		return "ICE_VSI_SWITCHDEV_CTRL";
 	default:
 		return "unknown";
 	}
@@ -144,7 +142,6 @@ static void ice_vsi_set_num_desc(struct ice_vsi *vsi)
 {
 	switch (vsi->type) {
 	case ICE_VSI_PF:
-	case ICE_VSI_SWITCHDEV_CTRL:
 	case ICE_VSI_CTRL:
 	case ICE_VSI_LB:
 		/* a user could change the values of num_[tr]x_desc using
@@ -211,21 +208,6 @@ static void ice_vsi_set_num_qs(struct ice_vsi *vsi)
 					   max_t(int, vsi->alloc_rxq,
 						 vsi->alloc_txq));
 		break;
-	case ICE_VSI_SWITCHDEV_CTRL:
-		/* The number of queues for ctrl VSI is equal to number of PRs
-		 * Each ring is associated to the corresponding VF_PR netdev.
-		 * Tx and Rx rings are always equal
-		 */
-		if (vsi->req_txq && vsi->req_rxq) {
-			vsi->alloc_txq = vsi->req_txq;
-			vsi->alloc_rxq = vsi->req_rxq;
-		} else {
-			vsi->alloc_txq = 1;
-			vsi->alloc_rxq = 1;
-		}
-
-		vsi->num_q_vectors = 1;
-		break;
 	case ICE_VSI_VF:
 		if (vf->num_req_qs)
 			vf->num_vf_qs = vf->num_req_qs;
@@ -522,22 +504,6 @@ static irqreturn_t ice_msix_clean_rings(int __always_unused irq, void *data)
 	return IRQ_HANDLED;
 }
 
-static irqreturn_t ice_eswitch_msix_clean_rings(int __always_unused irq, void *data)
-{
-	struct ice_q_vector *q_vector = (struct ice_q_vector *)data;
-	struct ice_pf *pf = q_vector->vsi->back;
-	struct ice_repr *repr;
-	unsigned long id;
-
-	if (!q_vector->tx.tx_ring && !q_vector->rx.rx_ring)
-		return IRQ_HANDLED;
-
-	xa_for_each(&pf->eswitch.reprs, id, repr)
-		napi_schedule(&repr->q_vector->napi);
-
-	return IRQ_HANDLED;
-}
-
 /**
  * ice_vsi_alloc_stat_arrays - Allocate statistics arrays
  * @vsi: VSI pointer
@@ -600,10 +566,6 @@ ice_vsi_alloc_def(struct ice_vsi *vsi, struct ice_channel *ch)
 	}
 
 	switch (vsi->type) {
-	case ICE_VSI_SWITCHDEV_CTRL:
-		/* Setup eswitch MSIX irq handler for VSI */
-		vsi->irq_handler = ice_eswitch_msix_clean_rings;
-		break;
 	case ICE_VSI_PF:
 		/* Setup default MSIX irq handler for VSI */
 		vsi->irq_handler = ice_msix_clean_rings;
@@ -933,11 +895,6 @@ static void ice_vsi_set_rss_params(struct ice_vsi *vsi)
 					      max_rss_size);
 		vsi->rss_lut_type = ICE_LUT_PF;
 		break;
-	case ICE_VSI_SWITCHDEV_CTRL:
-		vsi->rss_table_size = ICE_LUT_VSI_SIZE;
-		vsi->rss_size = min_t(u16, num_online_cpus(), max_rss_size);
-		vsi->rss_lut_type = ICE_LUT_VSI;
-		break;
 	case ICE_VSI_VF:
 		/* VF VSI will get a small RSS table.
 		 * For VSI_LUT, LUT size should be set to 64 bytes.
@@ -1263,7 +1220,6 @@ static int ice_vsi_init(struct ice_vsi *vsi, u32 vsi_flags)
 	case ICE_VSI_PF:
 		ctxt->flags = ICE_AQ_VSI_TYPE_PF;
 		break;
-	case ICE_VSI_SWITCHDEV_CTRL:
 	case ICE_VSI_CHNL:
 		ctxt->flags = ICE_AQ_VSI_TYPE_VMDQ2;
 		break;
@@ -2145,7 +2101,6 @@ static void ice_set_agg_vsi(struct ice_vsi *vsi)
 	case ICE_VSI_CHNL:
 	case ICE_VSI_LB:
 	case ICE_VSI_PF:
-	case ICE_VSI_SWITCHDEV_CTRL:
 		max_agg_nodes = ICE_MAX_PF_AGG_NODES;
 		agg_node_id_start = ICE_PF_AGG_NODE_ID_START;
 		agg_node_iter = &pf->pf_agg_node[0];
@@ -2317,7 +2272,6 @@ ice_vsi_cfg_def(struct ice_vsi *vsi, struct ice_vsi_cfg_params *params)
 
 	switch (vsi->type) {
 	case ICE_VSI_CTRL:
-	case ICE_VSI_SWITCHDEV_CTRL:
 	case ICE_VSI_PF:
 		ret = ice_vsi_alloc_q_vectors(vsi);
 		if (ret)
@@ -2750,8 +2704,7 @@ void ice_dis_vsi(struct ice_vsi *vsi, bool locked)
 		} else {
 			ice_vsi_close(vsi);
 		}
-	} else if (vsi->type == ICE_VSI_CTRL ||
-		   vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
+	} else if (vsi->type == ICE_VSI_CTRL) {
 		ice_vsi_close(vsi);
 	}
 }

drivers/net/ethernet/intel/ice/ice_main.c

@@ -7055,13 +7055,11 @@ int ice_down(struct ice_vsi *vsi)
 
 	WARN_ON(!test_bit(ICE_VSI_DOWN, vsi->state));
 
-	if (vsi->netdev && vsi->type == ICE_VSI_PF) {
+	if (vsi->netdev) {
 		vlan_err = ice_vsi_del_vlan_zero(vsi);
 		ice_ptp_link_change(vsi->back, vsi->back->hw.pf_id, false);
 		netif_carrier_off(vsi->netdev);
 		netif_tx_disable(vsi->netdev);
-	} else if (vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
-		ice_eswitch_stop_all_tx_queues(vsi->back);
 	}
 
 	ice_vsi_dis_irq(vsi);
@@ -7544,11 +7542,7 @@ static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
 		goto err_vsi_rebuild;
 	}
 
-	err = ice_eswitch_rebuild(pf);
-	if (err) {
-		dev_err(dev, "Switchdev rebuild failed: %d\n", err);
-		goto err_vsi_rebuild;
-	}
+	ice_eswitch_rebuild(pf);
 
 	if (reset_type == ICE_RESET_PFR) {
 		err = ice_rebuild_channels(pf);

drivers/net/ethernet/intel/ice/ice_repr.c

@@ -41,6 +41,47 @@ ice_repr_get_phys_port_name(struct net_device *netdev, char *buf, size_t len)
 	return 0;
 }
 
+/**
+ * ice_repr_inc_tx_stats - increment Tx statistic by one packet
+ * @repr: repr to increment stats on
+ * @len: length of the packet
+ * @xmit_status: value returned by xmit function
+ */
+void ice_repr_inc_tx_stats(struct ice_repr *repr, unsigned int len,
+			   int xmit_status)
+{
+	struct ice_repr_pcpu_stats *stats;
+
+	if (unlikely(xmit_status != NET_XMIT_SUCCESS &&
+		     xmit_status != NET_XMIT_CN)) {
+		this_cpu_inc(repr->stats->tx_drops);
+		return;
+	}
+
+	stats = this_cpu_ptr(repr->stats);
+	u64_stats_update_begin(&stats->syncp);
+	stats->tx_packets++;
+	stats->tx_bytes += len;
+	u64_stats_update_end(&stats->syncp);
+}
+
+/**
+ * ice_repr_inc_rx_stats - increment Rx statistic by one packet
+ * @netdev: repr netdev to increment stats on
+ * @len: length of the packet
+ */
+void ice_repr_inc_rx_stats(struct net_device *netdev, unsigned int len)
+{
+	struct ice_repr *repr = ice_netdev_to_repr(netdev);
+	struct ice_repr_pcpu_stats *stats;
+
+	stats = this_cpu_ptr(repr->stats);
+	u64_stats_update_begin(&stats->syncp);
+	stats->rx_packets++;
+	stats->rx_bytes += len;
+	u64_stats_update_end(&stats->syncp);
+}
+
 /**
  * ice_repr_get_stats64 - get VF stats for VFPR use
  * @netdev: pointer to port representor netdev
@@ -76,7 +117,7 @@ ice_repr_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
  * ice_netdev_to_repr - Get port representor for given netdevice
  * @netdev: pointer to port representor netdev
  */
-struct ice_repr *ice_netdev_to_repr(struct net_device *netdev)
+struct ice_repr *ice_netdev_to_repr(const struct net_device *netdev)
 {
 	struct ice_netdev_priv *np = netdev_priv(netdev);
 
@@ -139,38 +180,35 @@ static int ice_repr_stop(struct net_device *netdev)
  * ice_repr_sp_stats64 - get slow path stats for port representor
  * @dev: network interface device structure
  * @stats: netlink stats structure
- *
- * RX/TX stats are being swapped here to be consistent with VF stats. In slow
- * path, port representor receives data when the corresponding VF is sending it
- * (and vice versa), TX and RX bytes/packets are effectively swapped on port
- * representor.
  */
 static int
 ice_repr_sp_stats64(const struct net_device *dev,
 		    struct rtnl_link_stats64 *stats)
 {
-	struct ice_netdev_priv *np = netdev_priv(dev);
-	int vf_id = np->repr->vf->vf_id;
-	struct ice_tx_ring *tx_ring;
-	struct ice_rx_ring *rx_ring;
-	u64 pkts, bytes;
-
-	tx_ring = np->vsi->tx_rings[vf_id];
-	ice_fetch_u64_stats_per_ring(&tx_ring->ring_stats->syncp,
-				     tx_ring->ring_stats->stats,
-				     &pkts, &bytes);
-	stats->rx_packets = pkts;
-	stats->rx_bytes = bytes;
-
-	rx_ring = np->vsi->rx_rings[vf_id];
-	ice_fetch_u64_stats_per_ring(&rx_ring->ring_stats->syncp,
-				     rx_ring->ring_stats->stats,
-				     &pkts, &bytes);
-	stats->tx_packets = pkts;
-	stats->tx_bytes = bytes;
-	stats->tx_dropped = rx_ring->ring_stats->rx_stats.alloc_page_failed +
-			    rx_ring->ring_stats->rx_stats.alloc_buf_failed;
-
+	struct ice_repr *repr = ice_netdev_to_repr(dev);
+	int i;
+
+	for_each_possible_cpu(i) {
+		u64 tbytes, tpkts, tdrops, rbytes, rpkts;
+		struct ice_repr_pcpu_stats *repr_stats;
+		unsigned int start;
+
+		repr_stats = per_cpu_ptr(repr->stats, i);
+		do {
+			start = u64_stats_fetch_begin(&repr_stats->syncp);
+			tbytes = repr_stats->tx_bytes;
+			tpkts = repr_stats->tx_packets;
+			tdrops = repr_stats->tx_drops;
+			rbytes = repr_stats->rx_bytes;
+			rpkts = repr_stats->rx_packets;
+		} while (u64_stats_fetch_retry(&repr_stats->syncp, start));
+
+		stats->tx_bytes += tbytes;
+		stats->tx_packets += tpkts;
+		stats->tx_dropped += tdrops;
+		stats->rx_bytes += rbytes;
+		stats->rx_packets += rpkts;
+	}
 	return 0;
 }
 
@@ -291,7 +329,7 @@ static void ice_repr_remove_node(struct devlink_port *devlink_port)
  */
 static void ice_repr_rem(struct ice_repr *repr)
 {
-	kfree(repr->q_vector);
+	free_percpu(repr->stats);
 	free_netdev(repr->netdev);
 	kfree(repr);
 }
@@ -331,7 +369,6 @@ static void ice_repr_set_tx_topology(struct ice_pf *pf)
 static struct ice_repr *
 ice_repr_add(struct ice_pf *pf, struct ice_vsi *src_vsi, const u8 *parent_mac)
 {
-	struct ice_q_vector *q_vector;
 	struct ice_netdev_priv *np;
 	struct ice_repr *repr;
 	int err;
@@ -346,23 +383,22 @@ ice_repr_add(struct ice_pf *pf, struct ice_vsi *src_vsi, const u8 *parent_mac)
 		goto err_alloc;
 	}
 
+	repr->stats = netdev_alloc_pcpu_stats(struct ice_repr_pcpu_stats);
+	if (!repr->stats) {
+		err = -ENOMEM;
+		goto err_stats;
+	}
+
 	repr->src_vsi = src_vsi;
+	repr->id = src_vsi->vsi_num;
 	np = netdev_priv(repr->netdev);
 	np->repr = repr;
 
-	q_vector = kzalloc(sizeof(*q_vector), GFP_KERNEL);
-	if (!q_vector) {
-		err = -ENOMEM;
-		goto err_alloc_q_vector;
-	}
-	repr->q_vector = q_vector;
-	repr->q_id = repr->id;
-
 	ether_addr_copy(repr->parent_mac, parent_mac);
 
 	return repr;
 
-err_alloc_q_vector:
+err_stats:
 	free_netdev(repr->netdev);
 err_alloc:
 	kfree(repr);
@@ -439,15 +475,3 @@ void ice_repr_stop_tx_queues(struct ice_repr *repr)
 	netif_carrier_off(repr->netdev);
 	netif_tx_stop_all_queues(repr->netdev);
 }
-
-/**
- * ice_repr_set_traffic_vsi - set traffic VSI for port representor
- * @repr: repr on with VSI will be set
- * @vsi: pointer to VSI that will be used by port representor to pass traffic
- */
-void ice_repr_set_traffic_vsi(struct ice_repr *repr, struct ice_vsi *vsi)
-{
-	struct ice_netdev_priv *np = netdev_priv(repr->netdev);
-
-	np->vsi = vsi;
-}

drivers/net/ethernet/intel/ice/ice_repr.h

@@ -6,20 +6,24 @@
 
 #include <net/dst_metadata.h>
 
+struct ice_repr_pcpu_stats {
+	struct u64_stats_sync syncp;
+	u64 rx_packets;
+	u64 rx_bytes;
+	u64 tx_packets;
+	u64 tx_bytes;
+	u64 tx_drops;
+};
+
 struct ice_repr {
 	struct ice_vsi *src_vsi;
 	struct ice_vf *vf;
-	struct ice_q_vector *q_vector;
 	struct net_device *netdev;
 	struct metadata_dst *dst;
 	struct ice_esw_br_port *br_port;
-	int q_id;
+	struct ice_repr_pcpu_stats __percpu *stats;
 	u32 id;
 	u8 parent_mac[ETH_ALEN];
-#ifdef CONFIG_ICE_SWITCHDEV
-	/* info about slow path rule */
-	struct ice_rule_query_data sp_rule;
-#endif
 };
 
 struct ice_repr *ice_repr_add_vf(struct ice_vf *vf);
@@ -28,10 +32,12 @@ void ice_repr_rem_vf(struct ice_repr *repr);
 void ice_repr_start_tx_queues(struct ice_repr *repr);
 void ice_repr_stop_tx_queues(struct ice_repr *repr);
 
-void ice_repr_set_traffic_vsi(struct ice_repr *repr, struct ice_vsi *vsi);
-
-struct ice_repr *ice_netdev_to_repr(struct net_device *netdev);
+struct ice_repr *ice_netdev_to_repr(const struct net_device *netdev);
 bool ice_is_port_repr_netdev(const struct net_device *netdev);
 
 struct ice_repr *ice_repr_get_by_vsi(struct ice_vsi *vsi);
+
+void ice_repr_inc_tx_stats(struct ice_repr *repr, unsigned int len,
+			   int xmit_status);
+void ice_repr_inc_rx_stats(struct net_device *netdev, unsigned int len);
 #endif

drivers/net/ethernet/intel/ice/ice_sriov.c

@@ -170,8 +170,6 @@ void ice_free_vfs(struct ice_pf *pf)
 	else
 		dev_warn(dev, "VFs are assigned - not disabling SR-IOV\n");
 
-	ice_eswitch_reserve_cp_queues(pf, -ice_get_num_vfs(pf));
-
 	mutex_lock(&vfs->table_lock);
 
 	ice_for_each_vf(pf, bkt, vf) {
@@ -897,7 +895,6 @@ static int ice_ena_vfs(struct ice_pf *pf, u16 num_vfs)
 		goto err_unroll_sriov;
 	}
 
-	ice_eswitch_reserve_cp_queues(pf, num_vfs);
 	ret = ice_start_vfs(pf);
 	if (ret) {
 		dev_err(dev, "Failed to start %d VFs, err %d\n", num_vfs, ret);

drivers/net/ethernet/intel/ice/ice_switch.c

@@ -2446,6 +2446,9 @@ static void ice_fill_sw_info(struct ice_hw *hw, struct ice_fltr_info *fi)
 			fi->lan_en = true;
 		}
 	}
+
+	if (fi->flag & ICE_FLTR_TX_ONLY)
+		fi->lan_en = false;
 }
 
 /**
@@ -3821,6 +3824,7 @@ ice_cfg_dflt_vsi(struct ice_port_info *pi, u16 vsi_handle, bool set,
 	} else if (f_info.flag & ICE_FLTR_TX) {
 		f_info.src_id = ICE_SRC_ID_VSI;
 		f_info.src = hw_vsi_id;
+		f_info.flag |= ICE_FLTR_TX_ONLY;
 	}
 	f_list_entry.fltr_info = f_info;
 

drivers/net/ethernet/intel/ice/ice_switch.h

@@ -8,8 +8,9 @@
 
 #define ICE_SW_CFG_MAX_BUF_LEN 2048
 #define ICE_DFLT_VSI_INVAL 0xff
-#define ICE_FLTR_RX BIT(0)
-#define ICE_FLTR_TX BIT(1)
+#define ICE_FLTR_RX		BIT(0)
+#define ICE_FLTR_TX		BIT(1)
+#define ICE_FLTR_TX_ONLY	BIT(2)
 #define ICE_VSI_INVAL_ID 0xffff
 #define ICE_INVAL_Q_HANDLE 0xFFFF
 

drivers/net/ethernet/intel/ice/ice_txrx.h

@@ -365,6 +365,7 @@ struct ice_rx_ring {
 	u8 ptp_rx;
 #define ICE_RX_FLAGS_RING_BUILD_SKB	BIT(1)
 #define ICE_RX_FLAGS_CRC_STRIP_DIS	BIT(2)
+#define ICE_RX_FLAGS_MULTIDEV		BIT(3)
 	u8 flags;
 	/* CL5 - 5th cacheline starts here */
 	struct xdp_rxq_info xdp_rxq;

drivers/net/ethernet/intel/ice/ice_txrx_lib.c

@@ -236,7 +236,16 @@ ice_process_skb_fields(struct ice_rx_ring *rx_ring,
 	ice_rx_hash_to_skb(rx_ring, rx_desc, skb, ptype);
 
 	/* modifies the skb - consumes the enet header */
-	skb->protocol = eth_type_trans(skb, rx_ring->netdev);
+	if (unlikely(rx_ring->flags & ICE_RX_FLAGS_MULTIDEV)) {
+		struct net_device *netdev = ice_eswitch_get_target(rx_ring,
+								   rx_desc);
+
+		if (ice_is_port_repr_netdev(netdev))
+			ice_repr_inc_rx_stats(netdev, skb->len);
+		skb->protocol = eth_type_trans(skb, netdev);
+	} else {
+		skb->protocol = eth_type_trans(skb, rx_ring->netdev);
+	}
 
 	ice_rx_csum(rx_ring, skb, rx_desc, ptype);
 

drivers/net/ethernet/intel/ice/ice_type.h

@@ -150,7 +150,6 @@ enum ice_vsi_type {
 	ICE_VSI_CTRL = 3,	/* equates to ICE_VSI_PF with 1 queue pair */
 	ICE_VSI_CHNL = 4,
 	ICE_VSI_LB = 6,
-	ICE_VSI_SWITCHDEV_CTRL = 7,
 };
 
 struct ice_link_status {

drivers/net/ethernet/intel/ice/ice_vsi_vlan_ops.c

@@ -72,7 +72,6 @@ void ice_vsi_init_vlan_ops(struct ice_vsi *vsi)
 
 	switch (vsi->type) {
 	case ICE_VSI_PF:
-	case ICE_VSI_SWITCHDEV_CTRL:
 		ice_pf_vsi_init_vlan_ops(vsi);
 		break;
 	case ICE_VSI_VF: