OSDN Git Service

ice: Enable RX queue selection using skbedit action
authorAmritha Nambiar <amritha.nambiar@intel.com>
Fri, 21 Oct 2022 07:58:45 +0000 (00:58 -0700)
committerPaolo Abeni <pabeni@redhat.com>
Tue, 25 Oct 2022 08:32:40 +0000 (10:32 +0200)
This patch uses TC skbedit queue_mapping action to support
forwarding packets to a device queue. Such filters with action
forward to queue will be the highest priority switch filter in
HW.
Example:
$ tc filter add dev ens4f0 protocol ip ingress flower\
  dst_ip 192.168.1.12 ip_proto tcp dst_port 5001\
  action skbedit queue_mapping 5 skip_sw

The above command adds an ingress filter, incoming packets
qualifying the match will be accepted into queue 5. The queue
number is in decimal format.

Refactored ice_add_tc_flower_adv_fltr() to consolidate code with
action FWD_TO_VSI and FWD_TO QUEUE.

Reviewed-by: Sridhar Samudrala <sridhar.samudrala@intel.com>
Reviewed-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: Amritha Nambiar <amritha.nambiar@intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
drivers/net/ethernet/intel/ice/ice.h
drivers/net/ethernet/intel/ice/ice_main.c
drivers/net/ethernet/intel/ice/ice_tc_lib.c
drivers/net/ethernet/intel/ice/ice_tc_lib.h

index 001500a..d75c8d1 100644 (file)
  */
 #define ICE_BW_KBPS_DIVISOR            125
 
+/* Default recipes have priority 4 and below, hence priority values between 5..7
+ * can be used as filter priority for advanced switch filter (advanced switch
+ * filters need new recipe to be created for specified extraction sequence
+ * because default recipe extraction sequence does not represent custom
+ * extraction)
+ */
+#define ICE_SWITCH_FLTR_PRIO_QUEUE     7
+/* prio 6 is reserved for future use (e.g. switch filter with L3 fields +
+ * (Optional: IP TOS/TTL) + L4 fields + (optionally: TCP fields such as
+ * SYN/FIN/RST))
+ */
+#define ICE_SWITCH_FLTR_PRIO_RSVD      6
+#define ICE_SWITCH_FLTR_PRIO_VSI       5
+#define ICE_SWITCH_FLTR_PRIO_QGRP      ICE_SWITCH_FLTR_PRIO_VSI
+
 /* Macro for each VSI in a PF */
 #define ice_for_each_vsi(pf, i) \
        for ((i) = 0; (i) < (pf)->num_alloc_vsi; (i)++)
index 0f67187..df65e82 100644 (file)
@@ -8283,7 +8283,7 @@ static void ice_rem_all_chnl_fltrs(struct ice_pf *pf)
 
                rule.rid = fltr->rid;
                rule.rule_id = fltr->rule_id;
-               rule.vsi_handle = fltr->dest_id;
+               rule.vsi_handle = fltr->dest_vsi_handle;
                status = ice_rem_adv_rule_by_id(&pf->hw, &rule);
                if (status) {
                        if (status == -ENOENT)
index f68c555..faba0f8 100644 (file)
@@ -724,7 +724,7 @@ ice_eswitch_add_tc_fltr(struct ice_vsi *vsi, struct ice_tc_flower_fltr *fltr)
         */
        fltr->rid = rule_added.rid;
        fltr->rule_id = rule_added.rule_id;
-       fltr->dest_id = rule_added.vsi_handle;
+       fltr->dest_vsi_handle = rule_added.vsi_handle;
 
 exit:
        kfree(list);
@@ -732,6 +732,116 @@ exit:
 }
 
 /**
+ * ice_locate_vsi_using_queue - locate VSI using queue (forward to queue action)
+ * @vsi: Pointer to VSI
+ * @tc_fltr: Pointer to tc_flower_filter
+ *
+ * Locate the VSI using specified queue. When ADQ is not enabled, always
+ * return input VSI, otherwise locate corresponding VSI based on per channel
+ * offset and qcount
+ */
+static struct ice_vsi *
+ice_locate_vsi_using_queue(struct ice_vsi *vsi,
+                          struct ice_tc_flower_fltr *tc_fltr)
+{
+       int num_tc, tc, queue;
+
+       /* if ADQ is not active, passed VSI is the candidate VSI */
+       if (!ice_is_adq_active(vsi->back))
+               return vsi;
+
+       /* Locate the VSI (it could still be main PF VSI or CHNL_VSI depending
+        * upon queue number)
+        */
+       num_tc = vsi->mqprio_qopt.qopt.num_tc;
+       queue = tc_fltr->action.fwd.q.queue;
+
+       for (tc = 0; tc < num_tc; tc++) {
+               int qcount = vsi->mqprio_qopt.qopt.count[tc];
+               int offset = vsi->mqprio_qopt.qopt.offset[tc];
+
+               if (queue >= offset && queue < offset + qcount) {
+                       /* for non-ADQ TCs, passed VSI is the candidate VSI */
+                       if (tc < ICE_CHNL_START_TC)
+                               return vsi;
+                       else
+                               return vsi->tc_map_vsi[tc];
+               }
+       }
+       return NULL;
+}
+
+static struct ice_rx_ring *
+ice_locate_rx_ring_using_queue(struct ice_vsi *vsi,
+                              struct ice_tc_flower_fltr *tc_fltr)
+{
+       u16 queue = tc_fltr->action.fwd.q.queue;
+
+       return queue < vsi->num_rxq ? vsi->rx_rings[queue] : NULL;
+}
+
+/**
+ * ice_tc_forward_action - Determine destination VSI and queue for the action
+ * @vsi: Pointer to VSI
+ * @tc_fltr: Pointer to TC flower filter structure
+ *
+ * Validates the tc forward action and determines the destination VSI and queue
+ * for the forward action.
+ */
+static struct ice_vsi *
+ice_tc_forward_action(struct ice_vsi *vsi, struct ice_tc_flower_fltr *tc_fltr)
+{
+       struct ice_rx_ring *ring = NULL;
+       struct ice_vsi *ch_vsi = NULL;
+       struct ice_pf *pf = vsi->back;
+       struct device *dev;
+       u32 tc_class;
+
+       dev = ice_pf_to_dev(pf);
+
+       /* Get the destination VSI and/or destination queue and validate them */
+       switch (tc_fltr->action.fltr_act) {
+       case ICE_FWD_TO_VSI:
+               tc_class = tc_fltr->action.fwd.tc.tc_class;
+               /* Select the destination VSI */
+               if (tc_class < ICE_CHNL_START_TC) {
+                       NL_SET_ERR_MSG_MOD(tc_fltr->extack,
+                                          "Unable to add filter because of unsupported destination");
+                       return ERR_PTR(-EOPNOTSUPP);
+               }
+               /* Locate ADQ VSI depending on hw_tc number */
+               ch_vsi = vsi->tc_map_vsi[tc_class];
+               break;
+       case ICE_FWD_TO_Q:
+               /* Locate the Rx queue */
+               ring = ice_locate_rx_ring_using_queue(vsi, tc_fltr);
+               if (!ring) {
+                       dev_err(dev,
+                               "Unable to locate Rx queue for action fwd_to_queue: %u\n",
+                               tc_fltr->action.fwd.q.queue);
+                       return ERR_PTR(-EINVAL);
+               }
+               /* Determine destination VSI even though the action is
+                * FWD_TO_QUEUE, because QUEUE is associated with VSI
+                */
+               ch_vsi = tc_fltr->dest_vsi;
+               break;
+       default:
+               dev_err(dev,
+                       "Unable to add filter because of unsupported action %u (supported actions: fwd to tc, fwd to queue)\n",
+                       tc_fltr->action.fltr_act);
+               return ERR_PTR(-EINVAL);
+       }
+       /* Must have valid ch_vsi (it could be main VSI or ADQ VSI) */
+       if (!ch_vsi) {
+               dev_err(dev,
+                       "Unable to add filter because specified destination VSI doesn't exist\n");
+               return ERR_PTR(-EINVAL);
+       }
+       return ch_vsi;
+}
+
+/**
  * ice_add_tc_flower_adv_fltr - add appropriate filter rules
  * @vsi: Pointer to VSI
  * @tc_fltr: Pointer to TC flower filter structure
@@ -772,11 +882,10 @@ ice_add_tc_flower_adv_fltr(struct ice_vsi *vsi,
                return -EOPNOTSUPP;
        }
 
-       /* get the channel (aka ADQ VSI) */
-       if (tc_fltr->dest_vsi)
-               ch_vsi = tc_fltr->dest_vsi;
-       else
-               ch_vsi = vsi->tc_map_vsi[tc_fltr->action.tc_class];
+       /* validate forwarding action VSI and queue */
+       ch_vsi = ice_tc_forward_action(vsi, tc_fltr);
+       if (IS_ERR(ch_vsi))
+               return PTR_ERR(ch_vsi);
 
        lkups_cnt = ice_tc_count_lkups(flags, headers, tc_fltr);
        list = kcalloc(lkups_cnt, sizeof(*list), GFP_ATOMIC);
@@ -790,30 +899,40 @@ ice_add_tc_flower_adv_fltr(struct ice_vsi *vsi,
        }
 
        rule_info.sw_act.fltr_act = tc_fltr->action.fltr_act;
-       if (tc_fltr->action.tc_class >= ICE_CHNL_START_TC) {
-               if (!ch_vsi) {
-                       NL_SET_ERR_MSG_MOD(tc_fltr->extack, "Unable to add filter because specified destination doesn't exist");
-                       ret = -EINVAL;
-                       goto exit;
-               }
+       /* specify the cookie as filter_rule_id */
+       rule_info.fltr_rule_id = tc_fltr->cookie;
 
-               rule_info.sw_act.fltr_act = ICE_FWD_TO_VSI;
+       switch (tc_fltr->action.fltr_act) {
+       case ICE_FWD_TO_VSI:
                rule_info.sw_act.vsi_handle = ch_vsi->idx;
-               rule_info.priority = 7;
+               rule_info.priority = ICE_SWITCH_FLTR_PRIO_VSI;
                rule_info.sw_act.src = hw->pf_id;
                rule_info.rx = true;
                dev_dbg(dev, "add switch rule for TC:%u vsi_idx:%u, lkups_cnt:%u\n",
-                       tc_fltr->action.tc_class,
+                       tc_fltr->action.fwd.tc.tc_class,
                        rule_info.sw_act.vsi_handle, lkups_cnt);
-       } else {
+               break;
+       case ICE_FWD_TO_Q:
+               /* HW queue number in global space */
+               rule_info.sw_act.fwd_id.q_id = tc_fltr->action.fwd.q.hw_queue;
+               rule_info.sw_act.vsi_handle = ch_vsi->idx;
+               rule_info.priority = ICE_SWITCH_FLTR_PRIO_QUEUE;
+               rule_info.sw_act.src = hw->pf_id;
+               rule_info.rx = true;
+               dev_dbg(dev, "add switch rule action to forward to queue:%u (HW queue %u), lkups_cnt:%u\n",
+                       tc_fltr->action.fwd.q.queue,
+                       tc_fltr->action.fwd.q.hw_queue, lkups_cnt);
+               break;
+       default:
                rule_info.sw_act.flag |= ICE_FLTR_TX;
+               /* In case of Tx (LOOKUP_TX), src needs to be src VSI */
                rule_info.sw_act.src = vsi->idx;
+               /* 'Rx' is false, direction of rule(LOOKUPTRX) */
                rule_info.rx = false;
+               rule_info.priority = ICE_SWITCH_FLTR_PRIO_VSI;
+               break;
        }
 
-       /* specify the cookie as filter_rule_id */
-       rule_info.fltr_rule_id = tc_fltr->cookie;
-
        ret = ice_add_adv_rule(hw, list, lkups_cnt, &rule_info, &rule_added);
        if (ret == -EEXIST) {
                NL_SET_ERR_MSG_MOD(tc_fltr->extack,
@@ -831,19 +950,14 @@ ice_add_tc_flower_adv_fltr(struct ice_vsi *vsi,
         */
        tc_fltr->rid = rule_added.rid;
        tc_fltr->rule_id = rule_added.rule_id;
-       if (tc_fltr->action.tc_class > 0 && ch_vsi) {
-               /* For PF ADQ, VSI type is set as ICE_VSI_CHNL, and
-                * for PF ADQ filter, it is not yet set in tc_fltr,
-                * hence store the dest_vsi ptr in tc_fltr
-                */
-               if (ch_vsi->type == ICE_VSI_CHNL)
-                       tc_fltr->dest_vsi = ch_vsi;
+       tc_fltr->dest_vsi_handle = rule_added.vsi_handle;
+       if (tc_fltr->action.fltr_act == ICE_FWD_TO_VSI ||
+           tc_fltr->action.fltr_act == ICE_FWD_TO_Q) {
+               tc_fltr->dest_vsi = ch_vsi;
                /* keep track of advanced switch filter for
-                * destination VSI (channel VSI)
+                * destination VSI
                 */
                ch_vsi->num_chnl_fltr++;
-               /* in this case, dest_id is VSI handle (sw handle) */
-               tc_fltr->dest_id = rule_added.vsi_handle;
 
                /* keeps track of channel filters for PF VSI */
                if (vsi->type == ICE_VSI_PF &&
@@ -851,10 +965,22 @@ ice_add_tc_flower_adv_fltr(struct ice_vsi *vsi,
                              ICE_TC_FLWR_FIELD_ENC_DST_MAC)))
                        pf->num_dmac_chnl_fltrs++;
        }
-       dev_dbg(dev, "added switch rule (lkups_cnt %u, flags 0x%x) for TC %u, rid %u, rule_id %u, vsi_idx %u\n",
-               lkups_cnt, flags,
-               tc_fltr->action.tc_class, rule_added.rid,
-               rule_added.rule_id, rule_added.vsi_handle);
+       switch (tc_fltr->action.fltr_act) {
+       case ICE_FWD_TO_VSI:
+               dev_dbg(dev, "added switch rule (lkups_cnt %u, flags 0x%x), action is forward to TC %u, rid %u, rule_id %u, vsi_idx %u\n",
+                       lkups_cnt, flags,
+                       tc_fltr->action.fwd.tc.tc_class, rule_added.rid,
+                       rule_added.rule_id, rule_added.vsi_handle);
+               break;
+       case ICE_FWD_TO_Q:
+               dev_dbg(dev, "added switch rule (lkups_cnt %u, flags 0x%x), action is forward to queue: %u (HW queue %u)     , rid %u, rule_id %u\n",
+                       lkups_cnt, flags, tc_fltr->action.fwd.q.queue,
+                       tc_fltr->action.fwd.q.hw_queue, rule_added.rid,
+                       rule_added.rule_id);
+               break;
+       default:
+               break;
+       }
 exit:
        kfree(list);
        return ret;
@@ -1455,43 +1581,15 @@ ice_add_switch_fltr(struct ice_vsi *vsi, struct ice_tc_flower_fltr *fltr)
 }
 
 /**
- * ice_handle_tclass_action - Support directing to a traffic class
+ * ice_prep_adq_filter - Prepare ADQ filter with the required additional headers
  * @vsi: Pointer to VSI
- * @cls_flower: Pointer to TC flower offload structure
  * @fltr: Pointer to TC flower filter structure
  *
- * Support directing traffic to a traffic class
+ * Prepare ADQ filter with the required additional header fields
  */
 static int
-ice_handle_tclass_action(struct ice_vsi *vsi,
-                        struct flow_cls_offload *cls_flower,
-                        struct ice_tc_flower_fltr *fltr)
+ice_prep_adq_filter(struct ice_vsi *vsi, struct ice_tc_flower_fltr *fltr)
 {
-       int tc = tc_classid_to_hwtc(vsi->netdev, cls_flower->classid);
-       struct ice_vsi *main_vsi;
-
-       if (tc < 0) {
-               NL_SET_ERR_MSG_MOD(fltr->extack, "Unable to add filter because specified destination is invalid");
-               return -EINVAL;
-       }
-       if (!tc) {
-               NL_SET_ERR_MSG_MOD(fltr->extack, "Unable to add filter because of invalid destination");
-               return -EINVAL;
-       }
-
-       if (!(vsi->all_enatc & BIT(tc))) {
-               NL_SET_ERR_MSG_MOD(fltr->extack, "Unable to add filter because of non-existence destination");
-               return -EINVAL;
-       }
-
-       /* Redirect to a TC class or Queue Group */
-       main_vsi = ice_get_main_vsi(vsi->back);
-       if (!main_vsi || !main_vsi->netdev) {
-               NL_SET_ERR_MSG_MOD(fltr->extack,
-                                  "Unable to add filter because of invalid netdevice");
-               return -EINVAL;
-       }
-
        if ((fltr->flags & ICE_TC_FLWR_FIELD_TENANT_ID) &&
            (fltr->flags & (ICE_TC_FLWR_FIELD_DST_MAC |
                           ICE_TC_FLWR_FIELD_SRC_MAC))) {
@@ -1503,9 +1601,8 @@ ice_handle_tclass_action(struct ice_vsi *vsi,
        /* For ADQ, filter must include dest MAC address, otherwise unwanted
         * packets with unrelated MAC address get delivered to ADQ VSIs as long
         * as remaining filter criteria is satisfied such as dest IP address
-        * and dest/src L4 port. Following code is trying to handle:
-        * 1. For non-tunnel, if user specify MAC addresses, use them (means
-        * this code won't do anything
+        * and dest/src L4 port. Below code handles the following cases:
+        * 1. For non-tunnel, if user specify MAC addresses, use them.
         * 2. For non-tunnel, if user didn't specify MAC address, add implicit
         * dest MAC to be lower netdev's active unicast MAC address
         * 3. For tunnel,  as of now TC-filter through flower classifier doesn't
@@ -1528,35 +1625,97 @@ ice_handle_tclass_action(struct ice_vsi *vsi,
                eth_broadcast_addr(fltr->outer_headers.l2_mask.dst_mac);
        }
 
-       /* validate specified dest MAC address, make sure either it belongs to
-        * lower netdev or any of MACVLAN. MACVLANs MAC address are added as
-        * unicast MAC filter destined to main VSI.
-        */
-       if (!ice_mac_fltr_exist(&main_vsi->back->hw,
-                               fltr->outer_headers.l2_key.dst_mac,
-                               main_vsi->idx)) {
-               NL_SET_ERR_MSG_MOD(fltr->extack,
-                                  "Unable to add filter because legacy MAC filter for specified destination doesn't exist");
-               return -EINVAL;
-       }
-
        /* Make sure VLAN is already added to main VSI, before allowing ADQ to
         * add a VLAN based filter such as MAC + VLAN + L4 port.
         */
        if (fltr->flags & ICE_TC_FLWR_FIELD_VLAN) {
                u16 vlan_id = be16_to_cpu(fltr->outer_headers.vlan_hdr.vlan_id);
 
-               if (!ice_vlan_fltr_exist(&main_vsi->back->hw, vlan_id,
-                                        main_vsi->idx)) {
+               if (!ice_vlan_fltr_exist(&vsi->back->hw, vlan_id, vsi->idx)) {
                        NL_SET_ERR_MSG_MOD(fltr->extack,
                                           "Unable to add filter because legacy VLAN filter for specified destination doesn't exist");
                        return -EINVAL;
                }
        }
+       return 0;
+}
+
+/**
+ * ice_handle_tclass_action - Support directing to a traffic class
+ * @vsi: Pointer to VSI
+ * @cls_flower: Pointer to TC flower offload structure
+ * @fltr: Pointer to TC flower filter structure
+ *
+ * Support directing traffic to a traffic class/queue-set
+ */
+static int
+ice_handle_tclass_action(struct ice_vsi *vsi,
+                        struct flow_cls_offload *cls_flower,
+                        struct ice_tc_flower_fltr *fltr)
+{
+       int tc = tc_classid_to_hwtc(vsi->netdev, cls_flower->classid);
+
+       /* user specified hw_tc (must be non-zero for ADQ TC), action is forward
+        * to hw_tc (i.e. ADQ channel number)
+        */
+       if (tc < ICE_CHNL_START_TC) {
+               NL_SET_ERR_MSG_MOD(fltr->extack,
+                                  "Unable to add filter because of unsupported destination");
+               return -EOPNOTSUPP;
+       }
+       if (!(vsi->all_enatc & BIT(tc))) {
+               NL_SET_ERR_MSG_MOD(fltr->extack,
+                                  "Unable to add filter because of non-existence destination");
+               return -EINVAL;
+       }
        fltr->action.fltr_act = ICE_FWD_TO_VSI;
-       fltr->action.tc_class = tc;
+       fltr->action.fwd.tc.tc_class = tc;
 
-       return 0;
+       return ice_prep_adq_filter(vsi, fltr);
+}
+
+static int
+ice_tc_forward_to_queue(struct ice_vsi *vsi, struct ice_tc_flower_fltr *fltr,
+                       struct flow_action_entry *act)
+{
+       struct ice_vsi *ch_vsi = NULL;
+       u16 queue = act->rx_queue;
+
+       if (queue > vsi->num_rxq) {
+               NL_SET_ERR_MSG_MOD(fltr->extack,
+                                  "Unable to add filter because specified queue is invalid");
+               return -EINVAL;
+       }
+       fltr->action.fltr_act = ICE_FWD_TO_Q;
+       fltr->action.fwd.q.queue = queue;
+       /* determine corresponding HW queue */
+       fltr->action.fwd.q.hw_queue = vsi->rxq_map[queue];
+
+       /* If ADQ is configured, and the queue belongs to ADQ VSI, then prepare
+        * ADQ switch filter
+        */
+       ch_vsi = ice_locate_vsi_using_queue(vsi, fltr);
+       if (!ch_vsi)
+               return -EINVAL;
+       fltr->dest_vsi = ch_vsi;
+       if (!ice_is_chnl_fltr(fltr))
+               return 0;
+
+       return ice_prep_adq_filter(vsi, fltr);
+}
+
+static int
+ice_tc_parse_action(struct ice_vsi *vsi, struct ice_tc_flower_fltr *fltr,
+                   struct flow_action_entry *act)
+{
+       switch (act->id) {
+       case FLOW_ACTION_RX_QUEUE_MAPPING:
+               /* forward to queue */
+               return ice_tc_forward_to_queue(vsi, fltr, act);
+       default:
+               NL_SET_ERR_MSG_MOD(fltr->extack, "Unsupported TC action");
+               return -EOPNOTSUPP;
+       }
 }
 
 /**
@@ -1575,7 +1734,7 @@ ice_parse_tc_flower_actions(struct ice_vsi *vsi,
        struct flow_rule *rule = flow_cls_offload_flow_rule(cls_flower);
        struct flow_action *flow_action = &rule->action;
        struct flow_action_entry *act;
-       int i;
+       int i, err;
 
        if (cls_flower->classid)
                return ice_handle_tclass_action(vsi, cls_flower, fltr);
@@ -1584,21 +1743,13 @@ ice_parse_tc_flower_actions(struct ice_vsi *vsi,
                return -EINVAL;
 
        flow_action_for_each(i, act, flow_action) {
-               if (ice_is_eswitch_mode_switchdev(vsi->back)) {
-                       int err = ice_eswitch_tc_parse_action(fltr, act);
-
-                       if (err)
-                               return err;
-                       continue;
-               }
-               /* Allow only one rule per filter */
-
-               /* Drop action */
-               if (act->id == FLOW_ACTION_DROP) {
-                       NL_SET_ERR_MSG_MOD(fltr->extack, "Unsupported action DROP");
-                       return -EINVAL;
-               }
-               fltr->action.fltr_act = ICE_FWD_TO_VSI;
+               if (ice_is_eswitch_mode_switchdev(vsi->back))
+                       err = ice_eswitch_tc_parse_action(fltr, act);
+               else
+                       err = ice_tc_parse_action(vsi, fltr, act);
+               if (err)
+                       return err;
+               continue;
        }
        return 0;
 }
@@ -1618,7 +1769,7 @@ static int ice_del_tc_fltr(struct ice_vsi *vsi, struct ice_tc_flower_fltr *fltr)
 
        rule_rem.rid = fltr->rid;
        rule_rem.rule_id = fltr->rule_id;
-       rule_rem.vsi_handle = fltr->dest_id;
+       rule_rem.vsi_handle = fltr->dest_vsi_handle;
        err = ice_rem_adv_rule_by_id(&pf->hw, &rule_rem);
        if (err) {
                if (err == -ENOENT) {
index 92642fa..d916d1e 100644 (file)
@@ -45,7 +45,20 @@ struct ice_indr_block_priv {
 };
 
 struct ice_tc_flower_action {
-       u32 tc_class;
+       /* forward action specific params */
+       union {
+               struct {
+                       u32 tc_class; /* forward to hw_tc */
+                       u32 rsvd;
+               } tc;
+               struct {
+                       u16 queue; /* forward to queue */
+                       /* To add filter in HW, absolute queue number in global
+                        * space of queues (between 0...N) is needed
+                        */
+                       u16 hw_queue;
+               } q;
+       } fwd;
        enum ice_sw_fwd_act_type fltr_act;
 };
 
@@ -131,11 +144,11 @@ struct ice_tc_flower_fltr {
         */
        u16 rid;
        u16 rule_id;
-       /* this could be queue/vsi_idx (sw handle)/queue_group, depending upon
-        * destination type
+       /* VSI handle of the destination VSI (it could be main PF VSI, CHNL_VSI,
+        * VF VSI)
         */
-       u16 dest_id;
-       /* if dest_id is vsi_idx, then need to store destination VSI ptr */
+       u16 dest_vsi_handle;
+       /* ptr to destination VSI */
        struct ice_vsi *dest_vsi;
        /* direction of fltr for eswitch use case */
        enum ice_eswitch_fltr_direction direction;
@@ -162,12 +175,23 @@ struct ice_tc_flower_fltr {
  * @f: Pointer to tc-flower filter
  *
  * Criteria to determine of given filter is valid channel filter
- * or not is based on its "destination". If destination is hw_tc (aka tc_class)
- * and it is non-zero, then it is valid channel (aka ADQ) filter
+ * or not is based on its destination.
+ * For forward to VSI action, if destination is valid hw_tc (aka tc_class)
+ * and in supported range of TCs for ADQ, then return true.
+ * For forward to queue, as long as dest_vsi is valid and it is of type
+ * VSI_CHNL (PF ADQ VSI is of type VSI_CHNL), return true.
+ * NOTE: For forward to queue, correct dest_vsi is still set in tc_fltr based
+ * on destination queue specified.
  */
 static inline bool ice_is_chnl_fltr(struct ice_tc_flower_fltr *f)
 {
-       return !!f->action.tc_class;
+       if (f->action.fltr_act == ICE_FWD_TO_VSI)
+               return f->action.fwd.tc.tc_class >= ICE_CHNL_START_TC &&
+                      f->action.fwd.tc.tc_class < ICE_CHNL_MAX_TC;
+       else if (f->action.fltr_act == ICE_FWD_TO_Q)
+               return f->dest_vsi && f->dest_vsi->type == ICE_VSI_CHNL;
+
+       return false;
 }
 
 /**