[iwl-next v1 0/8] ice: use less resources in switchdev

[iwl-next v1 0/8] ice: use less resources in switchdev

[Michal Swiatkowski]

Hi,

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.

Michal Swiatkowski (8):
  ice: remove eswitch changing queues algorithm
  ice: do Tx through PF netdev in slow-path
  ice: default Tx rule instead of to queue
  ice: control default Tx rule in lag
  ice: remove switchdev control plane VSI
  ice: change repr::id values
  ice: do switchdev slow-path Rx using PF VSI
  ice: count representor stats

 drivers/net/ethernet/intel/ice/ice.h          |   7 -
 drivers/net/ethernet/intel/ice/ice_base.c     |  44 +--
 drivers/net/ethernet/intel/ice/ice_dcb_lib.c  |   4 +-
 drivers/net/ethernet/intel/ice/ice_eswitch.c  | 362 +++---------------
 drivers/net/ethernet/intel/ice/ice_eswitch.h  |  13 +-
 drivers/net/ethernet/intel/ice/ice_lag.c      |  48 ++-
 drivers/net/ethernet/intel/ice/ice_lag.h      |   3 +-
 drivers/net/ethernet/intel/ice/ice_lib.c      |  49 +--
 drivers/net/ethernet/intel/ice/ice_main.c     |  10 +-
 drivers/net/ethernet/intel/ice/ice_repr.c     | 126 +++---
 drivers/net/ethernet/intel/ice/ice_repr.h     |  24 +-
 drivers/net/ethernet/intel/ice/ice_sriov.c    |   3 -
 drivers/net/ethernet/intel/ice/ice_switch.c   |   4 +
 drivers/net/ethernet/intel/ice/ice_switch.h   |   5 +-
 drivers/net/ethernet/intel/ice/ice_txrx.h     |   1 +
 drivers/net/ethernet/intel/ice/ice_txrx_lib.c |  11 +-
 drivers/net/ethernet/intel/ice/ice_type.h     |   1 -
 .../net/ethernet/intel/ice/ice_vsi_vlan_ops.c |   1 -
 18 files changed, 227 insertions(+), 489 deletions(-)

-- 
2.42.0

[iwl-next v1 1/8] ice: remove eswitch changing queues algorithm

[Michal Swiatkowski]

Changing queues used by eswitch will be done through PF netdev.
There is no need to reserve queues if the number of used queues
is known.

Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Reviewed-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Signed-off-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
---
 drivers/net/ethernet/intel/ice/ice.h         |  6 ----
 drivers/net/ethernet/intel/ice/ice_eswitch.c | 34 --------------------
 drivers/net/ethernet/intel/ice/ice_eswitch.h |  4 ---
 drivers/net/ethernet/intel/ice/ice_sriov.c   |  3 --
 4 files changed, 47 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
index a4ba60e17d0b..e5241b9dc3c9 100644
--- a/drivers/net/ethernet/intel/ice/ice.h
+++ b/drivers/net/ethernet/intel/ice/ice.h
@@ -527,12 +527,6 @@ struct ice_eswitch {
 	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 {
diff --git a/drivers/net/ethernet/intel/ice/ice_eswitch.c b/drivers/net/ethernet/intel/ice/ice_eswitch.c
index 9069725c71b4..2e999f801c0a 100644
--- a/drivers/net/ethernet/intel/ice/ice_eswitch.c
+++ b/drivers/net/ethernet/intel/ice/ice_eswitch.c
@@ -455,8 +455,6 @@ static int ice_eswitch_enable_switchdev(struct ice_pf *pf)
 		return -ENODEV;
 
 	ctrl_vsi = pf->eswitch.control_vsi;
-	/* cp VSI is createad with 1 queue as default */
-	pf->eswitch.qs.value = 1;
 	pf->eswitch.uplink_vsi = uplink_vsi;
 
 	if (ice_eswitch_setup_env(pf))
@@ -489,7 +487,6 @@ static void ice_eswitch_disable_switchdev(struct ice_pf *pf)
 	ice_vsi_release(ctrl_vsi);
 
 	pf->eswitch.is_running = false;
-	pf->eswitch.qs.is_reaching = false;
 }
 
 /**
@@ -620,18 +617,6 @@ ice_eswitch_cp_change_queues(struct ice_eswitch *eswitch, int change)
 	struct ice_vsi *cp = eswitch->control_vsi;
 	int queues = 0;
 
-	if (eswitch->qs.is_reaching) {
-		if (eswitch->qs.to_reach >= eswitch->qs.value + change) {
-			queues = eswitch->qs.to_reach;
-			eswitch->qs.is_reaching = false;
-		} else {
-			queues = 0;
-		}
-	} else if ((change > 0 && cp->alloc_txq <= eswitch->qs.value) ||
-		   change < 0) {
-		queues = cp->alloc_txq + change;
-	}
-
 	if (queues) {
 		cp->req_txq = queues;
 		cp->req_rxq = queues;
@@ -643,7 +628,6 @@ ice_eswitch_cp_change_queues(struct ice_eswitch *eswitch, int change)
 		ice_vsi_open(cp);
 	}
 
-	eswitch->qs.value += change;
 	ice_eswitch_remap_rings_to_vectors(eswitch);
 }
 
@@ -661,8 +645,6 @@ ice_eswitch_attach(struct ice_pf *pf, struct ice_vf *vf)
 		err = ice_eswitch_enable_switchdev(pf);
 		if (err)
 			return err;
-		/* Control plane VSI is created with 1 queue as default */
-		pf->eswitch.qs.to_reach -= 1;
 		change = 0;
 	}
 
@@ -756,19 +738,3 @@ int ice_eswitch_rebuild(struct ice_pf *pf)
 
 	return 0;
 }
-
-/**
- * ice_eswitch_reserve_cp_queues - reserve control plane VSI queues
- * @pf: pointer to PF structure
- * @change: how many more (or less) queues is needed
- *
- * Remember to call ice_eswitch_attach/detach() the "change" times.
- */
-void ice_eswitch_reserve_cp_queues(struct ice_pf *pf, int change)
-{
-	if (pf->eswitch.qs.value + change < 0)
-		return;
-
-	pf->eswitch.qs.to_reach = pf->eswitch.qs.value + change;
-	pf->eswitch.qs.is_reaching = true;
-}
diff --git a/drivers/net/ethernet/intel/ice/ice_eswitch.h b/drivers/net/ethernet/intel/ice/ice_eswitch.h
index 1a288a03a79a..59d51c0d14e5 100644
--- a/drivers/net/ethernet/intel/ice/ice_eswitch.h
+++ b/drivers/net/ethernet/intel/ice/ice_eswitch.h
@@ -26,7 +26,6 @@ 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);
 #else /* CONFIG_ICE_SWITCHDEV */
 static inline void ice_eswitch_detach(struct ice_pf *pf, struct ice_vf *vf) { }
 
@@ -77,8 +76,5 @@ 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) { }
 #endif /* CONFIG_ICE_SWITCHDEV */
 #endif /* _ICE_ESWITCH_H_ */
diff --git a/drivers/net/ethernet/intel/ice/ice_sriov.c b/drivers/net/ethernet/intel/ice/ice_sriov.c
index a94a1c48c3de..706b5ee8ec89 100644
--- a/drivers/net/ethernet/intel/ice/ice_sriov.c
+++ b/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) {
@@ -898,7 +896,6 @@ 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);
-- 
2.42.0

[iwl-next v1 2/8] ice: do Tx through PF netdev in slow-path

[Michal Swiatkowski]

Tx can be done using PF netdev.

Checks before Tx are unnecessary. Checking if switchdev mode is set
seems too defensive (there is no PR netdev in legacy mode). If
corresponding VF is disabled or during reset, PR netdev also should be
down.

Reviewed-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Signed-off-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
---
 drivers/net/ethernet/intel/ice/ice_eswitch.c | 26 +++++---------------
 drivers/net/ethernet/intel/ice/ice_repr.c    | 12 ---------
 drivers/net/ethernet/intel/ice/ice_repr.h    |  2 --
 3 files changed, 6 insertions(+), 34 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_eswitch.c b/drivers/net/ethernet/intel/ice/ice_eswitch.c
index 2e999f801c0a..8ad271534d80 100644
--- a/drivers/net/ethernet/intel/ice/ice_eswitch.c
+++ b/drivers/net/ethernet/intel/ice/ice_eswitch.c
@@ -236,7 +236,7 @@ ice_eswitch_release_repr(struct ice_pf *pf, struct ice_repr *repr)
  */
 static int ice_eswitch_setup_repr(struct ice_pf *pf, struct ice_repr *repr)
 {
-	struct ice_vsi *ctrl_vsi = pf->eswitch.control_vsi;
+	struct ice_vsi *uplink_vsi = pf->eswitch.uplink_vsi;
 	struct ice_vsi *vsi = repr->src_vsi;
 	struct metadata_dst *dst;
 
@@ -255,12 +255,11 @@ static int ice_eswitch_setup_repr(struct ice_pf *pf, struct ice_repr *repr)
 	netif_napi_add(repr->netdev, &repr->q_vector->napi,
 		       ice_napi_poll);
 
-	netif_keep_dst(repr->netdev);
+	netif_keep_dst(uplink_vsi->netdev);
 
 	dst = repr->dst;
 	dst->u.port_info.port_id = vsi->vsi_num;
-	dst->u.port_info.lower_dev = repr->netdev;
-	ice_repr_set_traffic_vsi(repr, ctrl_vsi);
+	dst->u.port_info.lower_dev = uplink_vsi->netdev;
 
 	return 0;
 
@@ -318,27 +317,14 @@ void ice_eswitch_update_repr(unsigned long repr_id, struct ice_vsi *vsi)
 netdev_tx_t
 ice_eswitch_port_start_xmit(struct sk_buff *skb, struct net_device *netdev)
 {
-	struct ice_netdev_priv *np;
-	struct ice_repr *repr;
-	struct ice_vsi *vsi;
-
-	np = netdev_priv(netdev);
-	vsi = np->vsi;
-
-	if (!vsi || !ice_is_switchdev_running(vsi->back))
-		return NETDEV_TX_BUSY;
-
-	if (ice_is_reset_in_progress(vsi->back->state) ||
-	    test_bit(ICE_VF_DIS, vsi->back->state))
-		return NETDEV_TX_BUSY;
+	struct ice_repr *repr = ice_netdev_to_repr(netdev);
 
-	repr = ice_netdev_to_repr(netdev);
 	skb_dst_drop(skb);
 	dst_hold((struct dst_entry *)repr->dst);
 	skb_dst_set(skb, (struct dst_entry *)repr->dst);
-	skb->queue_mapping = repr->q_id;
+	skb->dev = repr->dst->u.port_info.lower_dev;
 
-	return ice_start_xmit(skb, netdev);
+	return dev_queue_xmit(skb);
 }
 
 /**
diff --git a/drivers/net/ethernet/intel/ice/ice_repr.c b/drivers/net/ethernet/intel/ice/ice_repr.c
index 5f30fb131f74..6191bee6c536 100644
--- a/drivers/net/ethernet/intel/ice/ice_repr.c
+++ b/drivers/net/ethernet/intel/ice/ice_repr.c
@@ -439,15 +439,3 @@ 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;
-}
diff --git a/drivers/net/ethernet/intel/ice/ice_repr.h b/drivers/net/ethernet/intel/ice/ice_repr.h
index f9aede315716..b107e058d538 100644
--- a/drivers/net/ethernet/intel/ice/ice_repr.h
+++ b/drivers/net/ethernet/intel/ice/ice_repr.h
@@ -28,8 +28,6 @@ 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);
 bool ice_is_port_repr_netdev(const struct net_device *netdev);
 
-- 
2.42.0

[iwl-next v1 3/8] ice: default Tx rule instead of to queue

[Michal Swiatkowski]

Steer all packets that miss other rules to PF VSI. Previously in
switchdev mode, PF VSI received missed packets, but only ones marked
as Rx. Now it is receiving all missed packets.

To queue rule per PR isn't needed, because we use PF VSI instead of
control VSI now, and it's already correctly configured.

Add flag to correctly set LAN_EN bit in default Tx rule. It shouldn't
allow packet to go outside when there is a match.

Reviewed-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Signed-off-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
---
 drivers/net/ethernet/intel/ice/ice_eswitch.c | 107 +++----------------
 drivers/net/ethernet/intel/ice/ice_repr.h    |   4 -
 drivers/net/ethernet/intel/ice/ice_switch.c  |   4 +
 drivers/net/ethernet/intel/ice/ice_switch.h  |   5 +-
 4 files changed, 23 insertions(+), 97 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_eswitch.c b/drivers/net/ethernet/intel/ice/ice_eswitch.c
index 8ad271534d80..50b3de700837 100644
--- a/drivers/net/ethernet/intel/ice/ice_eswitch.c
+++ b/drivers/net/ethernet/intel/ice/ice_eswitch.c
@@ -10,85 +10,6 @@
 #include "ice_devlink.h"
 #include "ice_tc_lib.h"
 
-/**
- * ice_eswitch_del_sp_rules - delete adv rules added on PRs
- * @pf: pointer to the PF struct
- *
- * Delete all advanced rules that were used to forward packets with the
- * device's VSI index to the corresponding eswitch ctrl VSI queue.
- */
-static void ice_eswitch_del_sp_rules(struct ice_pf *pf)
-{
-	struct ice_repr *repr;
-	unsigned long id;
-
-	xa_for_each(&pf->eswitch.reprs, id, repr) {
-		if (repr->sp_rule.rid)
-			ice_rem_adv_rule_by_id(&pf->hw, &repr->sp_rule);
-	}
-}
-
-/**
- * ice_eswitch_add_sp_rule - add adv rule with device's VSI index
- * @pf: pointer to PF struct
- * @repr: pointer to the repr struct
- *
- * This function adds advanced rule that forwards packets with
- * device's VSI index to the corresponding eswitch ctrl VSI queue.
- */
-static int ice_eswitch_add_sp_rule(struct ice_pf *pf, struct ice_repr *repr)
-{
-	struct ice_vsi *ctrl_vsi = pf->eswitch.control_vsi;
-	struct ice_adv_rule_info rule_info = { 0 };
-	struct ice_adv_lkup_elem *list;
-	struct ice_hw *hw = &pf->hw;
-	const u16 lkups_cnt = 1;
-	int err;
-
-	list = kcalloc(lkups_cnt, sizeof(*list), GFP_ATOMIC);
-	if (!list)
-		return -ENOMEM;
-
-	ice_rule_add_src_vsi_metadata(list);
-
-	rule_info.sw_act.flag = ICE_FLTR_TX;
-	rule_info.sw_act.vsi_handle = ctrl_vsi->idx;
-	rule_info.sw_act.fltr_act = ICE_FWD_TO_Q;
-	rule_info.sw_act.fwd_id.q_id = hw->func_caps.common_cap.rxq_first_id +
-				       ctrl_vsi->rxq_map[repr->q_id];
-	rule_info.flags_info.act |= ICE_SINGLE_ACT_LB_ENABLE;
-	rule_info.flags_info.act_valid = true;
-	rule_info.tun_type = ICE_SW_TUN_AND_NON_TUN;
-	rule_info.src_vsi = repr->src_vsi->idx;
-
-	err = ice_add_adv_rule(hw, list, lkups_cnt, &rule_info,
-			       &repr->sp_rule);
-	if (err)
-		dev_err(ice_pf_to_dev(pf), "Unable to add slow-path rule for eswitch for PR %d",
-			repr->id);
-
-	kfree(list);
-	return err;
-}
-
-static int
-ice_eswitch_add_sp_rules(struct ice_pf *pf)
-{
-	struct ice_repr *repr;
-	unsigned long id;
-	int err;
-
-	xa_for_each(&pf->eswitch.reprs, id, repr) {
-		err = ice_eswitch_add_sp_rule(pf, repr);
-		if (err) {
-			ice_eswitch_del_sp_rules(pf);
-			return err;
-		}
-	}
-
-	return 0;
-}
-
 /**
  * ice_eswitch_setup_env - configure eswitch HW filters
  * @pf: pointer to PF struct
@@ -102,7 +23,6 @@ static int ice_eswitch_setup_env(struct ice_pf *pf)
 	struct ice_vsi *ctrl_vsi = pf->eswitch.control_vsi;
 	struct net_device *netdev = uplink_vsi->netdev;
 	struct ice_vsi_vlan_ops *vlan_ops;
-	bool rule_added = false;
 
 	ice_remove_vsi_fltr(&pf->hw, uplink_vsi->idx);
 
@@ -112,17 +32,19 @@ static int ice_eswitch_setup_env(struct ice_pf *pf)
 	netif_addr_unlock_bh(netdev);
 
 	if (ice_vsi_add_vlan_zero(uplink_vsi))
+		goto err_vlan_zero;
+
+	if (ice_cfg_dflt_vsi(uplink_vsi->port_info, uplink_vsi->idx, true,
+			     ICE_FLTR_RX))
 		goto err_def_rx;
 
-	if (!ice_is_dflt_vsi_in_use(uplink_vsi->port_info)) {
-		if (ice_set_dflt_vsi(uplink_vsi))
-			goto err_def_rx;
-		rule_added = true;
-	}
+	if (ice_cfg_dflt_vsi(uplink_vsi->port_info, uplink_vsi->idx, true,
+			     ICE_FLTR_TX))
+		goto err_def_tx;
 
 	vlan_ops = ice_get_compat_vsi_vlan_ops(uplink_vsi);
 	if (vlan_ops->dis_rx_filtering(uplink_vsi))
-		goto err_dis_rx;
+		goto err_vlan_filtering;
 
 	if (ice_vsi_update_security(uplink_vsi, ice_vsi_ctx_set_allow_override))
 		goto err_override_uplink;
@@ -141,10 +63,15 @@ static int ice_eswitch_setup_env(struct ice_pf *pf)
 	ice_vsi_update_security(uplink_vsi, ice_vsi_ctx_clear_allow_override);
 err_override_uplink:
 	vlan_ops->ena_rx_filtering(uplink_vsi);
-err_dis_rx:
-	if (rule_added)
-		ice_clear_dflt_vsi(uplink_vsi);
+err_vlan_filtering:
+	ice_cfg_dflt_vsi(uplink_vsi->port_info, uplink_vsi->idx, false,
+			 ICE_FLTR_TX);
+err_def_tx:
+	ice_cfg_dflt_vsi(uplink_vsi->port_info, uplink_vsi->idx, false,
+			 ICE_FLTR_RX);
 err_def_rx:
+	ice_vsi_del_vlan_zero(uplink_vsi);
+err_vlan_zero:
 	ice_fltr_add_mac_and_broadcast(uplink_vsi,
 				       uplink_vsi->port_info->mac.perm_addr,
 				       ICE_FWD_TO_VSI);
@@ -585,7 +512,6 @@ void ice_eswitch_stop_all_tx_queues(struct ice_pf *pf)
 
 static void ice_eswitch_stop_reprs(struct ice_pf *pf)
 {
-	ice_eswitch_del_sp_rules(pf);
 	ice_eswitch_stop_all_tx_queues(pf);
 	ice_eswitch_napi_disable(&pf->eswitch.reprs);
 }
@@ -594,7 +520,6 @@ static void ice_eswitch_start_reprs(struct ice_pf *pf)
 {
 	ice_eswitch_napi_enable(&pf->eswitch.reprs);
 	ice_eswitch_start_all_tx_queues(pf);
-	ice_eswitch_add_sp_rules(pf);
 }
 
 static void
diff --git a/drivers/net/ethernet/intel/ice/ice_repr.h b/drivers/net/ethernet/intel/ice/ice_repr.h
index b107e058d538..34823f58cd23 100644
--- a/drivers/net/ethernet/intel/ice/ice_repr.h
+++ b/drivers/net/ethernet/intel/ice/ice_repr.h
@@ -16,10 +16,6 @@ struct ice_repr {
 	int q_id;
 	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);
diff --git a/drivers/net/ethernet/intel/ice/ice_switch.c b/drivers/net/ethernet/intel/ice/ice_switch.c
index f84bab80ca42..ce0b344dda99 100644
--- a/drivers/net/ethernet/intel/ice/ice_switch.c
+++ b/drivers/net/ethernet/intel/ice/ice_switch.c
@@ -2444,6 +2444,9 @@ 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;
 }
 
 /**
@@ -3819,6 +3822,7 @@ 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;
 
diff --git a/drivers/net/ethernet/intel/ice/ice_switch.h b/drivers/net/ethernet/intel/ice/ice_switch.h
index db7e501b7e0a..6f2cf9d6d01b 100644
--- a/drivers/net/ethernet/intel/ice/ice_switch.h
+++ b/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
 
-- 
2.42.0

[iwl-next v1 4/8] ice: control default Tx rule in lag

[Michal Swiatkowski]

Tx rule in switchdev was changed to use PF instead of additional control
plane VSI. Because of that during lag we should control it. Control
means to add and remove the default Tx rule during lag active/inactive
switching.

It can be done the same way as default Rx rule.

Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Reviewed-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Signed-off-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
---
 drivers/net/ethernet/intel/ice/ice_lag.c | 39 ++++++++++++++++++------
 drivers/net/ethernet/intel/ice/ice_lag.h |  3 +-
 2 files changed, 32 insertions(+), 10 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_lag.c b/drivers/net/ethernet/intel/ice/ice_lag.c
index 467372d541d2..efe882a46fd3 100644
--- a/drivers/net/ethernet/intel/ice/ice_lag.c
+++ b/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 +274,22 @@ 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)
+		return err;
+
+	err = ice_lag_cfg_fltr(lag, act, lag->pf_recipe, &lag->pf_tx_rule_id,
+			       ICE_FLTR_TX, add);
+	if (err) {
+		ice_lag_cfg_fltr(lag, act, lag->pf_recipe, &lag->pf_rx_rule_id,
+				 ICE_FLTR_RX, !add);
+		return err;
+	}
 
-	return ice_lag_cfg_fltr(lag, act, lag->pf_recipe,
-				&lag->pf_rule_id, add);
+	return 0;
 }
 
 /**
@@ -284,7 +305,7 @@ 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 +331,7 @@ 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 +340,7 @@ 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))
@@ -2149,7 +2170,7 @@ 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");
 
diff --git a/drivers/net/ethernet/intel/ice/ice_lag.h b/drivers/net/ethernet/intel/ice/ice_lag.h
index 183b38792ef2..bab2c83142a1 100644
--- a/drivers/net/ethernet/intel/ice/ice_lag.h
+++ b/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;
-- 
2.42.0

[iwl-next v1 5/8] ice: remove switchdev control plane VSI

[Michal Swiatkowski]

For slow-path Rx and Tx PF VSI is used. There is no need to have control
plane VSI. Remove all code related to it.

Eswitch rebuild can't fail without rebuilding control plane VSI. Return
void from ice_eswitch_rebuild().

Reviewed-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Signed-off-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
---
 drivers/net/ethernet/intel/ice/ice.h          |   1 -
 drivers/net/ethernet/intel/ice/ice_base.c     |  36 +---
 drivers/net/ethernet/intel/ice/ice_dcb_lib.c  |   4 +-
 drivers/net/ethernet/intel/ice/ice_eswitch.c  | 163 +-----------------
 drivers/net/ethernet/intel/ice/ice_eswitch.h  |   2 +-
 drivers/net/ethernet/intel/ice/ice_lag.c      |   9 +-
 drivers/net/ethernet/intel/ice/ice_lib.c      |  49 +-----
 drivers/net/ethernet/intel/ice/ice_main.c     |  10 +-
 drivers/net/ethernet/intel/ice/ice_repr.c     |  12 --
 drivers/net/ethernet/intel/ice/ice_repr.h     |   2 -
 drivers/net/ethernet/intel/ice/ice_type.h     |   1 -
 .../net/ethernet/intel/ice/ice_vsi_vlan_ops.c |   1 -
 12 files changed, 13 insertions(+), 277 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
index e5241b9dc3c9..c8b083dc78f0 100644
--- a/drivers/net/ethernet/intel/ice/ice.h
+++ b/drivers/net/ethernet/intel/ice/ice.h
@@ -522,7 +522,6 @@ 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;
diff --git a/drivers/net/ethernet/intel/ice/ice_base.c b/drivers/net/ethernet/intel/ice/ice_base.c
index d0d9d2f0beea..81081258bf13 100644
--- a/drivers/net/ethernet/intel/ice/ice_base.c
+++ b/drivers/net/ethernet/intel/ice/ice_base.c
@@ -265,30 +265,6 @@ 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
@@ -355,9 +331,6 @@ 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;
 	}
@@ -912,14 +885,7 @@ 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,
diff --git a/drivers/net/ethernet/intel/ice/ice_dcb_lib.c b/drivers/net/ethernet/intel/ice/ice_dcb_lib.c
index 6e20ee610022..ceb17c004d79 100644
--- a/drivers/net/ethernet/intel/ice/ice_dcb_lib.c
+++ b/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);
diff --git a/drivers/net/ethernet/intel/ice/ice_eswitch.c b/drivers/net/ethernet/intel/ice/ice_eswitch.c
index 50b3de700837..76fa114f22c9 100644
--- a/drivers/net/ethernet/intel/ice/ice_eswitch.c
+++ b/drivers/net/ethernet/intel/ice/ice_eswitch.c
@@ -20,7 +20,6 @@
 static int ice_eswitch_setup_env(struct ice_pf *pf)
 {
 	struct ice_vsi *uplink_vsi = pf->eswitch.uplink_vsi;
-	struct ice_vsi *ctrl_vsi = pf->eswitch.control_vsi;
 	struct net_device *netdev = uplink_vsi->netdev;
 	struct ice_vsi_vlan_ops *vlan_ops;
 
@@ -49,17 +48,12 @@ static int ice_eswitch_setup_env(struct ice_pf *pf)
 	if (ice_vsi_update_security(uplink_vsi, ice_vsi_ctx_set_allow_override))
 		goto err_override_uplink;
 
-	if (ice_vsi_update_security(ctrl_vsi, ice_vsi_ctx_set_allow_override))
-		goto err_override_control;
-
 	if (ice_vsi_update_local_lb(uplink_vsi, true))
 		goto err_override_local_lb;
 
 	return 0;
 
 err_override_local_lb:
-	ice_vsi_update_security(ctrl_vsi, ice_vsi_ctx_clear_allow_override);
-err_override_control:
 	ice_vsi_update_security(uplink_vsi, ice_vsi_ctx_clear_allow_override);
 err_override_uplink:
 	vlan_ops->ena_rx_filtering(uplink_vsi);
@@ -78,61 +72,6 @@ static int ice_eswitch_setup_env(struct ice_pf *pf)
 	return -ENODEV;
 }
 
-/**
- * ice_eswitch_remap_rings_to_vectors - reconfigure rings of eswitch ctrl VSI
- * @eswitch: pointer to eswitch struct
- *
- * In eswitch number of allocated Tx/Rx rings is equal.
- *
- * This function fills q_vectors structures associated with representor and
- * move each ring pairs to port representor netdevs. Each port representor
- * will have dedicated 1 Tx/Rx ring pair, so number of rings pair is equal to
- * number of VFs.
- */
-static void ice_eswitch_remap_rings_to_vectors(struct ice_eswitch *eswitch)
-{
-	struct ice_vsi *vsi = eswitch->control_vsi;
-	unsigned long repr_id = 0;
-	int q_id;
-
-	ice_for_each_txq(vsi, q_id) {
-		struct ice_q_vector *q_vector;
-		struct ice_tx_ring *tx_ring;
-		struct ice_rx_ring *rx_ring;
-		struct ice_repr *repr;
-
-		repr = xa_find(&eswitch->reprs, &repr_id, U32_MAX,
-			       XA_PRESENT);
-		if (!repr)
-			break;
-
-		repr_id += 1;
-		repr->q_id = q_id;
-		q_vector = repr->q_vector;
-		tx_ring = vsi->tx_rings[q_id];
-		rx_ring = vsi->rx_rings[q_id];
-
-		q_vector->vsi = vsi;
-		q_vector->reg_idx = vsi->q_vectors[0]->reg_idx;
-
-		q_vector->num_ring_tx = 1;
-		q_vector->tx.tx_ring = tx_ring;
-		tx_ring->q_vector = q_vector;
-		tx_ring->next = NULL;
-		tx_ring->netdev = repr->netdev;
-		/* In switchdev mode, from OS stack perspective, there is only
-		 * one queue for given netdev, so it needs to be indexed as 0.
-		 */
-		tx_ring->q_index = 0;
-
-		q_vector->num_ring_rx = 1;
-		q_vector->rx.rx_ring = rx_ring;
-		rx_ring->q_vector = q_vector;
-		rx_ring->next = NULL;
-		rx_ring->netdev = repr->netdev;
-	}
-}
-
 /**
  * ice_eswitch_release_repr - clear PR VSI configuration
  * @pf: poiner to PF struct
@@ -152,8 +91,6 @@ ice_eswitch_release_repr(struct ice_pf *pf, struct ice_repr *repr)
 	repr->dst = NULL;
 	ice_fltr_add_mac_and_broadcast(vsi, repr->parent_mac,
 				       ICE_FWD_TO_VSI);
-
-	netif_napi_del(&repr->q_vector->napi);
 }
 
 /**
@@ -179,9 +116,6 @@ static int ice_eswitch_setup_repr(struct ice_pf *pf, struct ice_repr *repr)
 	if (ice_vsi_add_vlan_zero(vsi))
 		goto err_update_security;
 
-	netif_napi_add(repr->netdev, &repr->q_vector->napi,
-		       ice_napi_poll);
-
 	netif_keep_dst(uplink_vsi->netdev);
 
 	dst = repr->dst;
@@ -287,13 +221,11 @@ ice_eswitch_set_target_vsi(struct sk_buff *skb,
 static void ice_eswitch_release_env(struct ice_pf *pf)
 {
 	struct ice_vsi *uplink_vsi = pf->eswitch.uplink_vsi;
-	struct ice_vsi *ctrl_vsi = pf->eswitch.control_vsi;
 	struct ice_vsi_vlan_ops *vlan_ops;
 
 	vlan_ops = ice_get_compat_vsi_vlan_ops(uplink_vsi);
 
 	ice_vsi_update_local_lb(uplink_vsi, false);
-	ice_vsi_update_security(ctrl_vsi, ice_vsi_ctx_clear_allow_override);
 	ice_vsi_update_security(uplink_vsi, ice_vsi_ctx_clear_allow_override);
 	vlan_ops->ena_rx_filtering(uplink_vsi);
 	ice_clear_dflt_vsi(uplink_vsi);
@@ -302,56 +234,13 @@ static void ice_eswitch_release_env(struct ice_pf *pf)
 				       ICE_FWD_TO_VSI);
 }
 
-/**
- * ice_eswitch_vsi_setup - configure eswitch control VSI
- * @pf: pointer to PF structure
- * @pi: pointer to port_info structure
- */
-static struct ice_vsi *
-ice_eswitch_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
-{
-	struct ice_vsi_cfg_params params = {};
-
-	params.type = ICE_VSI_SWITCHDEV_CTRL;
-	params.pi = pi;
-	params.flags = ICE_VSI_FLAG_INIT;
-
-	return ice_vsi_setup(pf, &params);
-}
-
-/**
- * ice_eswitch_napi_enable - enable NAPI for all port representors
- * @reprs: xarray of reprs
- */
-static void ice_eswitch_napi_enable(struct xarray *reprs)
-{
-	struct ice_repr *repr;
-	unsigned long id;
-
-	xa_for_each(reprs, id, repr)
-		napi_enable(&repr->q_vector->napi);
-}
-
-/**
- * ice_eswitch_napi_disable - disable NAPI for all port representors
- * @reprs: xarray of reprs
- */
-static void ice_eswitch_napi_disable(struct xarray *reprs)
-{
-	struct ice_repr *repr;
-	unsigned long id;
-
-	xa_for_each(reprs, id, repr)
-		napi_disable(&repr->q_vector->napi);
-}
-
 /**
  * ice_eswitch_enable_switchdev - configure eswitch in switchdev mode
  * @pf: pointer to PF structure
  */
 static int ice_eswitch_enable_switchdev(struct ice_pf *pf)
 {
-	struct ice_vsi *ctrl_vsi, *uplink_vsi;
+	struct ice_vsi *uplink_vsi;
 
 	uplink_vsi = ice_get_main_vsi(pf);
 	if (!uplink_vsi)
@@ -363,15 +252,10 @@ static int ice_eswitch_enable_switchdev(struct ice_pf *pf)
 		return -EINVAL;
 	}
 
-	pf->eswitch.control_vsi = ice_eswitch_vsi_setup(pf, pf->hw.port_info);
-	if (!pf->eswitch.control_vsi)
-		return -ENODEV;
-
-	ctrl_vsi = pf->eswitch.control_vsi;
 	pf->eswitch.uplink_vsi = uplink_vsi;
 
 	if (ice_eswitch_setup_env(pf))
-		goto err_vsi;
+		return -ENODEV;
 
 	if (ice_eswitch_br_offloads_init(pf))
 		goto err_br_offloads;
@@ -382,8 +266,6 @@ static int ice_eswitch_enable_switchdev(struct ice_pf *pf)
 
 err_br_offloads:
 	ice_eswitch_release_env(pf);
-err_vsi:
-	ice_vsi_release(ctrl_vsi);
 	return -ENODEV;
 }
 
@@ -393,11 +275,8 @@ static int ice_eswitch_enable_switchdev(struct ice_pf *pf)
  */
 static void ice_eswitch_disable_switchdev(struct ice_pf *pf)
 {
-	struct ice_vsi *ctrl_vsi = pf->eswitch.control_vsi;
-
 	ice_eswitch_br_offloads_deinit(pf);
 	ice_eswitch_release_env(pf);
-	ice_vsi_release(ctrl_vsi);
 
 	pf->eswitch.is_running = false;
 }
@@ -513,40 +392,17 @@ void ice_eswitch_stop_all_tx_queues(struct ice_pf *pf)
 static void ice_eswitch_stop_reprs(struct ice_pf *pf)
 {
 	ice_eswitch_stop_all_tx_queues(pf);
-	ice_eswitch_napi_disable(&pf->eswitch.reprs);
 }
 
 static void ice_eswitch_start_reprs(struct ice_pf *pf)
 {
-	ice_eswitch_napi_enable(&pf->eswitch.reprs);
 	ice_eswitch_start_all_tx_queues(pf);
 }
 
-static void
-ice_eswitch_cp_change_queues(struct ice_eswitch *eswitch, int change)
-{
-	struct ice_vsi *cp = eswitch->control_vsi;
-	int queues = 0;
-
-	if (queues) {
-		cp->req_txq = queues;
-		cp->req_rxq = queues;
-		ice_vsi_close(cp);
-		ice_vsi_rebuild(cp, ICE_VSI_FLAG_NO_INIT);
-		ice_vsi_open(cp);
-	} else if (!change) {
-		/* change == 0 means that VSI wasn't open, open it here */
-		ice_vsi_open(cp);
-	}
-
-	ice_eswitch_remap_rings_to_vectors(eswitch);
-}
-
 int
 ice_eswitch_attach(struct ice_pf *pf, struct ice_vf *vf)
 {
 	struct ice_repr *repr;
-	int change = 1;
 	int err;
 
 	if (pf->eswitch_mode == DEVLINK_ESWITCH_MODE_LEGACY)
@@ -556,7 +412,6 @@ ice_eswitch_attach(struct ice_pf *pf, struct ice_vf *vf)
 		err = ice_eswitch_enable_switchdev(pf);
 		if (err)
 			return err;
-		change = 0;
 	}
 
 	ice_eswitch_stop_reprs(pf);
@@ -578,7 +433,6 @@ ice_eswitch_attach(struct ice_pf *pf, struct ice_vf *vf)
 
 	vf->repr_id = repr->id;
 
-	ice_eswitch_cp_change_queues(&pf->eswitch, change);
 	ice_eswitch_start_reprs(pf);
 
 	return 0;
@@ -608,8 +462,6 @@ void ice_eswitch_detach(struct ice_pf *pf, struct ice_vf *vf)
 
 	if (xa_empty(&pf->eswitch.reprs))
 		ice_eswitch_disable_switchdev(pf);
-	else
-		ice_eswitch_cp_change_queues(&pf->eswitch, -1);
 
 	ice_eswitch_release_repr(pf, repr);
 	ice_repr_rem_vf(repr);
@@ -631,21 +483,14 @@ void ice_eswitch_detach(struct ice_pf *pf, struct ice_vf *vf)
  * ice_eswitch_rebuild - rebuild eswitch
  * @pf: pointer to PF structure
  */
-int ice_eswitch_rebuild(struct ice_pf *pf)
+void ice_eswitch_rebuild(struct ice_pf *pf)
 {
 	struct ice_repr *repr;
 	unsigned long id;
-	int err;
 
 	if (!ice_is_switchdev_running(pf))
-		return 0;
-
-	err = ice_vsi_rebuild(pf->eswitch.control_vsi, ICE_VSI_FLAG_INIT);
-	if (err)
-		return err;
+		return;
 
 	xa_for_each(&pf->eswitch.reprs, id, repr)
 		ice_eswitch_detach(pf, repr->vf);
-
-	return 0;
 }
diff --git a/drivers/net/ethernet/intel/ice/ice_eswitch.h b/drivers/net/ethernet/intel/ice/ice_eswitch.h
index 59d51c0d14e5..baad68507471 100644
--- a/drivers/net/ethernet/intel/ice/ice_eswitch.h
+++ b/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
diff --git a/drivers/net/ethernet/intel/ice/ice_lag.c b/drivers/net/ethernet/intel/ice/ice_lag.c
index efe882a46fd3..2756ed45941c 100644
--- a/drivers/net/ethernet/intel/ice/ice_lag.c
+++ b/drivers/net/ethernet/intel/ice/ice_lag.c
@@ -735,8 +735,7 @@ 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);
 }
 
@@ -974,8 +973,7 @@ 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);
 }
@@ -1997,8 +1995,7 @@ 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);
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index 60e0d824195e..718cb8df7853 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/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;
@@ -2108,7 +2064,6 @@ 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];
@@ -2280,7 +2235,6 @@ 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)
@@ -2713,8 +2667,7 @@ 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);
 	}
 }
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index 1fa3f40743f5..e1f62e8af14f 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -7077,13 +7077,11 @@ 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);
@@ -7566,11 +7564,7 @@ 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);
diff --git a/drivers/net/ethernet/intel/ice/ice_repr.c b/drivers/net/ethernet/intel/ice/ice_repr.c
index 6191bee6c536..a5c24da31b88 100644
--- a/drivers/net/ethernet/intel/ice/ice_repr.c
+++ b/drivers/net/ethernet/intel/ice/ice_repr.c
@@ -291,7 +291,6 @@ 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_netdev(repr->netdev);
 	kfree(repr);
 }
@@ -331,7 +330,6 @@ 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;
@@ -350,20 +348,10 @@ ice_repr_add(struct ice_pf *pf, struct ice_vsi *src_vsi, const u8 *parent_mac)
 	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:
-	free_netdev(repr->netdev);
 err_alloc:
 	kfree(repr);
 	return ERR_PTR(err);
diff --git a/drivers/net/ethernet/intel/ice/ice_repr.h b/drivers/net/ethernet/intel/ice/ice_repr.h
index 34823f58cd23..eb8dec1f7de4 100644
--- a/drivers/net/ethernet/intel/ice/ice_repr.h
+++ b/drivers/net/ethernet/intel/ice/ice_repr.h
@@ -9,11 +9,9 @@
 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;
 	u32 id;
 	u8 parent_mac[ETH_ALEN];
 };
diff --git a/drivers/net/ethernet/intel/ice/ice_type.h b/drivers/net/ethernet/intel/ice/ice_type.h
index 3ecfdebc0473..32386aecd6c5 100644
--- a/drivers/net/ethernet/intel/ice/ice_type.h
+++ b/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 {
diff --git a/drivers/net/ethernet/intel/ice/ice_vsi_vlan_ops.c b/drivers/net/ethernet/intel/ice/ice_vsi_vlan_ops.c
index 4a6c850d83ac..7aae7fdcfcdb 100644
--- a/drivers/net/ethernet/intel/ice/ice_vsi_vlan_ops.c
+++ b/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:
-- 
2.42.0

[iwl-next v1 6/8] ice: change repr::id values

[Michal Swiatkowski]

Instead of getting repr::id from xa_alloc() value, set it to the
src_vsi::num_vsi value. It is unique for each PR.

Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Signed-off-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
---
 drivers/net/ethernet/intel/ice/ice_eswitch.c | 5 ++---
 drivers/net/ethernet/intel/ice/ice_repr.c    | 1 +
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_eswitch.c b/drivers/net/ethernet/intel/ice/ice_eswitch.c
index 76fa114f22c9..5eba8dec9f94 100644
--- a/drivers/net/ethernet/intel/ice/ice_eswitch.c
+++ b/drivers/net/ethernet/intel/ice/ice_eswitch.c
@@ -319,7 +319,7 @@ ice_eswitch_mode_set(struct devlink *devlink, u16 mode,
 
 		dev_info(ice_pf_to_dev(pf), "PF %d changed eswitch mode to switchdev",
 			 pf->hw.pf_id);
-		xa_init_flags(&pf->eswitch.reprs, XA_FLAGS_ALLOC);
+		xa_init(&pf->eswitch.reprs);
 		NL_SET_ERR_MSG_MOD(extack, "Changed eswitch mode to switchdev");
 		break;
 	}
@@ -426,8 +426,7 @@ ice_eswitch_attach(struct ice_pf *pf, struct ice_vf *vf)
 	if (err)
 		goto err_setup_repr;
 
-	err = xa_alloc(&pf->eswitch.reprs, &repr->id, repr,
-		       XA_LIMIT(1, INT_MAX), GFP_KERNEL);
+	err = xa_insert(&pf->eswitch.reprs, repr->id, repr, GFP_KERNEL);
 	if (err)
 		goto err_xa_alloc;
 
diff --git a/drivers/net/ethernet/intel/ice/ice_repr.c b/drivers/net/ethernet/intel/ice/ice_repr.c
index a5c24da31b88..b4fb74271811 100644
--- a/drivers/net/ethernet/intel/ice/ice_repr.c
+++ b/drivers/net/ethernet/intel/ice/ice_repr.c
@@ -345,6 +345,7 @@ ice_repr_add(struct ice_pf *pf, struct ice_vsi *src_vsi, const u8 *parent_mac)
 	}
 
 	repr->src_vsi = src_vsi;
+	repr->id = src_vsi->vsi_num;
 	np = netdev_priv(repr->netdev);
 	np->repr = repr;
 
-- 
2.42.0

[iwl-next v1 7/8] ice: do switchdev slow-path Rx using PF VSI

[Michal Swiatkowski]

Add an ICE_RX_FLAG_MULTIDEV flag to Rx ring.

If it is set try to find correct port representor. Do it based on
src_vsi value stored in flex descriptor. Ids of representor pointers
stored in xarray are equal to corresponding src_vsi value. Thanks to
that we can directly get correct representor if we have src_vsi value.

Set multidev flag during ring configuration.

If the mode is switchdev, change the ring descriptor to the one that
contains src_vsi value.

PF netdev should be reconfigured, do it by calling ice_down() and
ice_up() if the netdev was up before configuring switchdev.

Reviewed-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Signed-off-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
---
 drivers/net/ethernet/intel/ice/ice_base.c     |  8 +++++
 drivers/net/ethernet/intel/ice/ice_eswitch.c  | 36 +++++++++++++++++++
 drivers/net/ethernet/intel/ice/ice_eswitch.h  |  9 +++++
 drivers/net/ethernet/intel/ice/ice_txrx.h     |  1 +
 drivers/net/ethernet/intel/ice/ice_txrx_lib.c |  8 ++++-
 5 files changed, 61 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_base.c b/drivers/net/ethernet/intel/ice/ice_base.c
index 81081258bf13..df2e3eb87b2e 100644
--- a/drivers/net/ethernet/intel/ice/ice_base.c
+++ b/drivers/net/ethernet/intel/ice/ice_base.c
@@ -454,6 +454,14 @@ 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;
diff --git a/drivers/net/ethernet/intel/ice/ice_eswitch.c b/drivers/net/ethernet/intel/ice/ice_eswitch.c
index 5eba8dec9f94..86a6d58ad3ec 100644
--- a/drivers/net/ethernet/intel/ice/ice_eswitch.c
+++ b/drivers/net/ethernet/intel/ice/ice_eswitch.c
@@ -21,8 +21,13 @@ static int ice_eswitch_setup_env(struct ice_pf *pf)
 {
 	struct ice_vsi *uplink_vsi = pf->eswitch.uplink_vsi;
 	struct net_device *netdev = uplink_vsi->netdev;
+	bool if_running = netif_running(netdev);
 	struct ice_vsi_vlan_ops *vlan_ops;
 
+	if (if_running && !test_and_set_bit(ICE_VSI_DOWN, uplink_vsi->state))
+		if (ice_down(uplink_vsi))
+			return -ENODEV;
+
 	ice_remove_vsi_fltr(&pf->hw, uplink_vsi->idx);
 
 	netif_addr_lock_bh(netdev);
@@ -51,8 +56,13 @@ static int ice_eswitch_setup_env(struct ice_pf *pf)
 	if (ice_vsi_update_local_lb(uplink_vsi, true))
 		goto err_override_local_lb;
 
+	if (if_running && ice_up(uplink_vsi))
+		goto err_up;
+
 	return 0;
 
+err_up:
+	ice_vsi_update_local_lb(uplink_vsi, false);
 err_override_local_lb:
 	ice_vsi_update_security(uplink_vsi, ice_vsi_ctx_clear_allow_override);
 err_override_uplink:
@@ -69,6 +79,9 @@ static int ice_eswitch_setup_env(struct ice_pf *pf)
 	ice_fltr_add_mac_and_broadcast(uplink_vsi,
 				       uplink_vsi->port_info->mac.perm_addr,
 				       ICE_FWD_TO_VSI);
+	if (if_running)
+		ice_up(uplink_vsi);
+
 	return -ENODEV;
 }
 
@@ -493,3 +506,26 @@ void ice_eswitch_rebuild(struct ice_pf *pf)
 	xa_for_each(&pf->eswitch.reprs, id, repr)
 		ice_eswitch_detach(pf, repr->vf);
 }
+
+/**
+ * ice_eswitch_get_target - get netdev based on src_vsi from descriptor
+ * @rx_ring: ring used to receive the packet
+ * @rx_desc: descriptor used to get src_vsi value
+ *
+ * Get src_vsi value from descriptor and load correct representor. If it isn't
+ * found return rx_ring->netdev.
+ */
+struct net_device *ice_eswitch_get_target(struct ice_rx_ring *rx_ring,
+					  union ice_32b_rx_flex_desc *rx_desc)
+{
+	struct ice_eswitch *eswitch = &rx_ring->vsi->back->eswitch;
+	struct ice_32b_rx_flex_desc_nic_2 *desc;
+	struct ice_repr *repr;
+
+	desc = (struct ice_32b_rx_flex_desc_nic_2 *)rx_desc;
+	repr = xa_load(&eswitch->reprs, le16_to_cpu(desc->src_vsi));
+	if (!repr)
+		return rx_ring->netdev;
+
+	return repr->netdev;
+}
diff --git a/drivers/net/ethernet/intel/ice/ice_eswitch.h b/drivers/net/ethernet/intel/ice/ice_eswitch.h
index baad68507471..e2e5c0c75e7d 100644
--- a/drivers/net/ethernet/intel/ice/ice_eswitch.h
+++ b/drivers/net/ethernet/intel/ice/ice_eswitch.h
@@ -26,6 +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);
+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) { }
 
@@ -76,5 +78,12 @@ ice_eswitch_port_start_xmit(struct sk_buff *skb, struct net_device *netdev)
 {
 	return NETDEV_TX_BUSY;
 }
+
+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_ */
diff --git a/drivers/net/ethernet/intel/ice/ice_txrx.h b/drivers/net/ethernet/intel/ice/ice_txrx.h
index b3379ff73674..4c57f5c0b979 100644
--- a/drivers/net/ethernet/intel/ice/ice_txrx.h
+++ b/drivers/net/ethernet/intel/ice/ice_txrx.h
@@ -364,6 +364,7 @@ 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;
diff --git a/drivers/net/ethernet/intel/ice/ice_txrx_lib.c b/drivers/net/ethernet/intel/ice/ice_txrx_lib.c
index f8f1d2bdc1be..0a6cdfd393b5 100644
--- a/drivers/net/ethernet/intel/ice/ice_txrx_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_txrx_lib.c
@@ -236,7 +236,13 @@ 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);
+		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);
 
-- 
2.42.0

[iwl-next v1 8/8] ice: count representor stats

[Michal Swiatkowski]

Removing control plane VSI result in no information about slow-path
statistic. In current solution statistics need to be counted in driver.

Patch is based on similar implementation done by Simon Horman in nfp:
commit eadfa4c3be99 ("nfp: add stats and xmit helpers for representors")

Add const modifier to netdev parameter in ice_netdev_to_repr(). It isn't
(and shouldn't be) modified in the function.

Reviewed-by: Marcin Szycik <marcin.szycik@linux.intel.com>
Signed-off-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
---
 drivers/net/ethernet/intel/ice/ice_eswitch.c  |   7 +-
 drivers/net/ethernet/intel/ice/ice_repr.c     | 103 +++++++++++++-----
 drivers/net/ethernet/intel/ice/ice_repr.h     |  16 ++-
 drivers/net/ethernet/intel/ice/ice_txrx_lib.c |   3 +
 4 files changed, 99 insertions(+), 30 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_eswitch.c b/drivers/net/ethernet/intel/ice/ice_eswitch.c
index 86a6d58ad3ec..af4e9530eb48 100644
--- a/drivers/net/ethernet/intel/ice/ice_eswitch.c
+++ b/drivers/net/ethernet/intel/ice/ice_eswitch.c
@@ -192,13 +192,18 @@ netdev_tx_t
 ice_eswitch_port_start_xmit(struct sk_buff *skb, struct net_device *netdev)
 {
 	struct ice_repr *repr = ice_netdev_to_repr(netdev);
+	unsigned int len = skb->len;
+	int ret;
 
 	skb_dst_drop(skb);
 	dst_hold((struct dst_entry *)repr->dst);
 	skb_dst_set(skb, (struct dst_entry *)repr->dst);
 	skb->dev = repr->dst->u.port_info.lower_dev;
 
-	return dev_queue_xmit(skb);
+	ret = dev_queue_xmit(skb);
+	ice_repr_inc_tx_stats(repr, len, ret);
+
+	return ret;
 }
 
 /**
diff --git a/drivers/net/ethernet/intel/ice/ice_repr.c b/drivers/net/ethernet/intel/ice/ice_repr.c
index b4fb74271811..2429727d5562 100644
--- a/drivers/net/ethernet/intel/ice/ice_repr.c
+++ b/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;
 }
 
@@ -291,6 +329,7 @@ static void ice_repr_remove_node(struct devlink_port *devlink_port)
  */
 static void ice_repr_rem(struct ice_repr *repr)
 {
+	free_percpu(repr->stats);
 	free_netdev(repr->netdev);
 	kfree(repr);
 }
@@ -344,6 +383,12 @@ 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);
@@ -353,6 +398,8 @@ ice_repr_add(struct ice_pf *pf, struct ice_vsi *src_vsi, const u8 *parent_mac)
 
 	return repr;
 
+err_stats:
+	free_netdev(repr->netdev);
 err_alloc:
 	kfree(repr);
 	return ERR_PTR(err);
diff --git a/drivers/net/ethernet/intel/ice/ice_repr.h b/drivers/net/ethernet/intel/ice/ice_repr.h
index eb8dec1f7de4..cff730b15ca0 100644
--- a/drivers/net/ethernet/intel/ice/ice_repr.h
+++ b/drivers/net/ethernet/intel/ice/ice_repr.h
@@ -6,12 +6,22 @@
 
 #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 net_device *netdev;
 	struct metadata_dst *dst;
 	struct ice_esw_br_port *br_port;
+	struct ice_repr_pcpu_stats __percpu *stats;
 	u32 id;
 	u8 parent_mac[ETH_ALEN];
 };
@@ -22,8 +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);
 
-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
diff --git a/drivers/net/ethernet/intel/ice/ice_txrx_lib.c b/drivers/net/ethernet/intel/ice/ice_txrx_lib.c
index 0a6cdfd393b5..df072ce767b1 100644
--- a/drivers/net/ethernet/intel/ice/ice_txrx_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_txrx_lib.c
@@ -239,6 +239,9 @@ ice_process_skb_fields(struct ice_rx_ring *rx_ring,
 	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);
-- 
2.42.0

Re: [iwl-next v1 4/8] ice: control default Tx rule in lag

[Simon Horman]

On Thu, Jan 25, 2024 at 01:53:10PM +0100, Michal Swiatkowski wrote:
> Tx rule in switchdev was changed to use PF instead of additional control
> plane VSI. Because of that during lag we should control it. Control
> means to add and remove the default Tx rule during lag active/inactive
> switching.
> 
> It can be done the same way as default Rx rule.

Hi Michal,

Can I confirm that LAG TX/RX works both before and after this patch?

> 
> Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
> Reviewed-by: Marcin Szycik <marcin.szycik@linux.intel.com>
> Signed-off-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
> ---
>  drivers/net/ethernet/intel/ice/ice_lag.c | 39 ++++++++++++++++++------
>  drivers/net/ethernet/intel/ice/ice_lag.h |  3 +-
>  2 files changed, 32 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/net/ethernet/intel/ice/ice_lag.c b/drivers/net/ethernet/intel/ice/ice_lag.c

...

> @@ -266,9 +274,22 @@ 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)
> +		return err;
> +
> +	err = ice_lag_cfg_fltr(lag, act, lag->pf_recipe, &lag->pf_tx_rule_id,
> +			       ICE_FLTR_TX, add);
> +	if (err) {
> +		ice_lag_cfg_fltr(lag, act, lag->pf_recipe, &lag->pf_rx_rule_id,
> +				 ICE_FLTR_RX, !add);
> +		return err;
> +	}
>  
> -	return ice_lag_cfg_fltr(lag, act, lag->pf_recipe,
> -				&lag->pf_rule_id, add);
> +	return 0;
>  }

nit: perhaps this could be more idiomatically written using a
     goto to unwind on error.

...

Re: [iwl-next v1 4/8] ice: control default Tx rule in lag

[Michal Swiatkowski]

On Mon, Jan 29, 2024 at 10:55:41AM +0000, Simon Horman wrote:
> On Thu, Jan 25, 2024 at 01:53:10PM +0100, Michal Swiatkowski wrote:
> > Tx rule in switchdev was changed to use PF instead of additional control
> > plane VSI. Because of that during lag we should control it. Control
> > means to add and remove the default Tx rule during lag active/inactive
> > switching.
> > 
> > It can be done the same way as default Rx rule.
> 
> Hi Michal,
> 
> Can I confirm that LAG TX/RX works both before and after this patch?
> 

Hi Simon,

This part of LAG code is related to the LAG + switchdev feature (it
isn't chaning LAG core code). Hope that normal LAG also works well. This
is the scenario when you have PF in switchdev, bond created of two PFs
connected to the bridge with representors. Switching between interfaces
from bond needs to add default Rx rule, and after my changes also
default Tx rule.

Do you think I should add this description to commit message?

Thanks,
Michal

> > 
> > Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
> > Reviewed-by: Marcin Szycik <marcin.szycik@linux.intel.com>
> > Signed-off-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
> > ---
> >  drivers/net/ethernet/intel/ice/ice_lag.c | 39 ++++++++++++++++++------
> >  drivers/net/ethernet/intel/ice/ice_lag.h |  3 +-
> >  2 files changed, 32 insertions(+), 10 deletions(-)
> > 
> > diff --git a/drivers/net/ethernet/intel/ice/ice_lag.c b/drivers/net/ethernet/intel/ice/ice_lag.c
> 
> ...
> 
> > @@ -266,9 +274,22 @@ 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)
> > +		return err;
> > +
> > +	err = ice_lag_cfg_fltr(lag, act, lag->pf_recipe, &lag->pf_tx_rule_id,
> > +			       ICE_FLTR_TX, add);
> > +	if (err) {
> > +		ice_lag_cfg_fltr(lag, act, lag->pf_recipe, &lag->pf_rx_rule_id,
> > +				 ICE_FLTR_RX, !add);
> > +		return err;
> > +	}
> >  
> > -	return ice_lag_cfg_fltr(lag, act, lag->pf_recipe,
> > -				&lag->pf_rule_id, add);
> > +	return 0;
> >  }
> 
> nit: perhaps this could be more idiomatically written using a
>      goto to unwind on error.
>
Thanks, I will rewrite it.

> ...

Re: [iwl-next v1 4/8] ice: control default Tx rule in lag

[Simon Horman]

On Thu, Feb 01, 2024 at 08:05:06AM +0100, Michal Swiatkowski wrote:
> On Mon, Jan 29, 2024 at 10:55:41AM +0000, Simon Horman wrote:
> > On Thu, Jan 25, 2024 at 01:53:10PM +0100, Michal Swiatkowski wrote:
> > > Tx rule in switchdev was changed to use PF instead of additional control
> > > plane VSI. Because of that during lag we should control it. Control
> > > means to add and remove the default Tx rule during lag active/inactive
> > > switching.
> > > 
> > > It can be done the same way as default Rx rule.
> > 
> > Hi Michal,
> > 
> > Can I confirm that LAG TX/RX works both before and after this patch?
> > 
> 
> Hi Simon,
> 
> This part of LAG code is related to the LAG + switchdev feature (it
> isn't chaning LAG core code). Hope that normal LAG also works well. This
> is the scenario when you have PF in switchdev, bond created of two PFs
> connected to the bridge with representors. Switching between interfaces
> from bond needs to add default Rx rule, and after my changes also
> default Tx rule.
> 
> Do you think I should add this description to commit message?

Thanks Michal,

I think that might be a good idea.

But my question was a bit different: I'm asking if this patch addresses
a regression introduced earlier in the series. Sorry for not being clearer
the first time around.