net/ice: support L4 for QinQ switch filter
[dpdk.git] / drivers / net / sfc / sfc_mae.c
index 2de72fa..4b520bc 100644 (file)
@@ -19,6 +19,7 @@
 #include "sfc_mae_counter.h"
 #include "sfc_log.h"
 #include "sfc_switch.h"
+#include "sfc_service.h"
 
 static int
 sfc_mae_assign_entity_mport(struct sfc_adapter *sa,
@@ -30,6 +31,19 @@ sfc_mae_assign_entity_mport(struct sfc_adapter *sa,
                                              mportp);
 }
 
+static int
+sfc_mae_counter_registry_init(struct sfc_mae_counter_registry *registry,
+                             uint32_t nb_counters_max)
+{
+       return sfc_mae_counters_init(&registry->counters, nb_counters_max);
+}
+
+static void
+sfc_mae_counter_registry_fini(struct sfc_mae_counter_registry *registry)
+{
+       sfc_mae_counters_fini(&registry->counters);
+}
+
 int
 sfc_mae_attach(struct sfc_adapter *sa)
 {
@@ -59,6 +73,15 @@ sfc_mae_attach(struct sfc_adapter *sa)
        if (rc != 0)
                goto fail_mae_get_limits;
 
+       sfc_log_init(sa, "init MAE counter registry");
+       rc = sfc_mae_counter_registry_init(&mae->counter_registry,
+                                          limits.eml_max_n_counters);
+       if (rc != 0) {
+               sfc_err(sa, "failed to init MAE counters registry for %u entries: %s",
+                       limits.eml_max_n_counters, rte_strerror(rc));
+               goto fail_counter_registry_init;
+       }
+
        sfc_log_init(sa, "assign entity MPORT");
        rc = sfc_mae_assign_entity_mport(sa, &entity_mport);
        if (rc != 0)
@@ -107,6 +130,9 @@ fail_mae_alloc_bounce_eh:
 fail_mae_assign_switch_port:
 fail_mae_assign_switch_domain:
 fail_mae_assign_entity_mport:
+       sfc_mae_counter_registry_fini(&mae->counter_registry);
+
+fail_counter_registry_init:
 fail_mae_get_limits:
        efx_mae_fini(sa->nic);
 
@@ -131,6 +157,7 @@ sfc_mae_detach(struct sfc_adapter *sa)
                return;
 
        rte_free(mae->bounce_eh.buf);
+       sfc_mae_counter_registry_fini(&mae->counter_registry);
 
        efx_mae_fini(sa->nic);
 
@@ -480,9 +507,72 @@ sfc_mae_encap_header_disable(struct sfc_adapter *sa,
        --(fw_rsrc->refcnt);
 }
 
+static int
+sfc_mae_counters_enable(struct sfc_adapter *sa,
+                       struct sfc_mae_counter_id *counters,
+                       unsigned int n_counters,
+                       efx_mae_actions_t *action_set_spec)
+{
+       int rc;
+
+       sfc_log_init(sa, "entry");
+
+       if (n_counters == 0) {
+               sfc_log_init(sa, "no counters - skip");
+               return 0;
+       }
+
+       SFC_ASSERT(sfc_adapter_is_locked(sa));
+       SFC_ASSERT(n_counters == 1);
+
+       rc = sfc_mae_counter_enable(sa, &counters[0]);
+       if (rc != 0) {
+               sfc_err(sa, "failed to enable MAE counter %u: %s",
+                       counters[0].mae_id.id, rte_strerror(rc));
+               goto fail_counter_add;
+       }
+
+       rc = efx_mae_action_set_fill_in_counter_id(action_set_spec,
+                                                  &counters[0].mae_id);
+       if (rc != 0) {
+               sfc_err(sa, "failed to fill in MAE counter %u in action set: %s",
+                       counters[0].mae_id.id, rte_strerror(rc));
+               goto fail_fill_in_id;
+       }
+
+       return 0;
+
+fail_fill_in_id:
+       (void)sfc_mae_counter_disable(sa, &counters[0]);
+
+fail_counter_add:
+       sfc_log_init(sa, "failed: %s", rte_strerror(rc));
+       return rc;
+}
+
+static int
+sfc_mae_counters_disable(struct sfc_adapter *sa,
+                        struct sfc_mae_counter_id *counters,
+                        unsigned int n_counters)
+{
+       if (n_counters == 0)
+               return 0;
+
+       SFC_ASSERT(sfc_adapter_is_locked(sa));
+       SFC_ASSERT(n_counters == 1);
+
+       if (counters[0].mae_id.id == EFX_MAE_RSRC_ID_INVALID) {
+               sfc_err(sa, "failed to disable: already disabled");
+               return EALREADY;
+       }
+
+       return sfc_mae_counter_disable(sa, &counters[0]);
+}
+
 static struct sfc_mae_action_set *
 sfc_mae_action_set_attach(struct sfc_adapter *sa,
                          const struct sfc_mae_encap_header *encap_header,
+                         unsigned int n_count,
                          const efx_mae_actions_t *spec)
 {
        struct sfc_mae_action_set *action_set;
@@ -491,7 +581,12 @@ sfc_mae_action_set_attach(struct sfc_adapter *sa,
        SFC_ASSERT(sfc_adapter_is_locked(sa));
 
        TAILQ_FOREACH(action_set, &mae->action_sets, entries) {
+               /*
+                * Shared counters are not supported, hence action sets with
+                * COUNT are not attachable.
+                */
                if (action_set->encap_header == encap_header &&
+                   n_count == 0 &&
                    efx_mae_action_set_specs_equal(action_set->spec, spec)) {
                        sfc_dbg(sa, "attaching to action_set=%p", action_set);
                        ++(action_set->refcnt);
@@ -504,18 +599,52 @@ sfc_mae_action_set_attach(struct sfc_adapter *sa,
 
 static int
 sfc_mae_action_set_add(struct sfc_adapter *sa,
+                      const struct rte_flow_action actions[],
                       efx_mae_actions_t *spec,
                       struct sfc_mae_encap_header *encap_header,
+                      unsigned int n_counters,
                       struct sfc_mae_action_set **action_setp)
 {
        struct sfc_mae_action_set *action_set;
        struct sfc_mae *mae = &sa->mae;
+       unsigned int i;
 
        SFC_ASSERT(sfc_adapter_is_locked(sa));
 
        action_set = rte_zmalloc("sfc_mae_action_set", sizeof(*action_set), 0);
-       if (action_set == NULL)
+       if (action_set == NULL) {
+               sfc_err(sa, "failed to alloc action set");
                return ENOMEM;
+       }
+
+       if (n_counters > 0) {
+               const struct rte_flow_action *action;
+
+               action_set->counters = rte_malloc("sfc_mae_counter_ids",
+                       sizeof(action_set->counters[0]) * n_counters, 0);
+               if (action_set->counters == NULL) {
+                       rte_free(action_set);
+                       sfc_err(sa, "failed to alloc counters");
+                       return ENOMEM;
+               }
+
+               for (action = actions, i = 0;
+                    action->type != RTE_FLOW_ACTION_TYPE_END && i < n_counters;
+                    ++action) {
+                       const struct rte_flow_action_count *conf;
+
+                       if (action->type != RTE_FLOW_ACTION_TYPE_COUNT)
+                               continue;
+
+                       conf = action->conf;
+
+                       action_set->counters[i].mae_id.id =
+                               EFX_MAE_RSRC_ID_INVALID;
+                       action_set->counters[i].rte_id = conf->id;
+                       i++;
+               }
+               action_set->n_counters = n_counters;
+       }
 
        action_set->refcnt = 1;
        action_set->spec = spec;
@@ -555,6 +684,12 @@ sfc_mae_action_set_del(struct sfc_adapter *sa,
 
        efx_mae_action_set_spec_fini(sa->nic, action_set->spec);
        sfc_mae_encap_header_del(sa, action_set->encap_header);
+       if (action_set->n_counters > 0) {
+               SFC_ASSERT(action_set->n_counters == 1);
+               SFC_ASSERT(action_set->counters[0].mae_id.id ==
+                          EFX_MAE_RSRC_ID_INVALID);
+               rte_free(action_set->counters);
+       }
        TAILQ_REMOVE(&mae->action_sets, action_set, entries);
        rte_free(action_set);
 
@@ -566,6 +701,7 @@ sfc_mae_action_set_enable(struct sfc_adapter *sa,
                          struct sfc_mae_action_set *action_set)
 {
        struct sfc_mae_encap_header *encap_header = action_set->encap_header;
+       struct sfc_mae_counter_id *counters = action_set->counters;
        struct sfc_mae_fw_rsrc *fw_rsrc = &action_set->fw_rsrc;
        int rc;
 
@@ -580,14 +716,26 @@ sfc_mae_action_set_enable(struct sfc_adapter *sa,
                if (rc != 0)
                        return rc;
 
-               rc = efx_mae_action_set_alloc(sa->nic, action_set->spec,
-                                             &fw_rsrc->aset_id);
+               rc = sfc_mae_counters_enable(sa, counters,
+                                            action_set->n_counters,
+                                            action_set->spec);
                if (rc != 0) {
+                       sfc_err(sa, "failed to enable %u MAE counters: %s",
+                               action_set->n_counters, rte_strerror(rc));
+
                        sfc_mae_encap_header_disable(sa, encap_header);
+                       return rc;
+               }
 
+               rc = efx_mae_action_set_alloc(sa->nic, action_set->spec,
+                                             &fw_rsrc->aset_id);
+               if (rc != 0) {
                        sfc_err(sa, "failed to enable action_set=%p: %s",
                                action_set, strerror(rc));
 
+                       (void)sfc_mae_counters_disable(sa, counters,
+                                                      action_set->n_counters);
+                       sfc_mae_encap_header_disable(sa, encap_header);
                        return rc;
                }
 
@@ -627,6 +775,13 @@ sfc_mae_action_set_disable(struct sfc_adapter *sa,
                }
                fw_rsrc->aset_id.id = EFX_MAE_RSRC_ID_INVALID;
 
+               rc = sfc_mae_counters_disable(sa, action_set->counters,
+                                             action_set->n_counters);
+               if (rc != 0) {
+                       sfc_err(sa, "failed to disable %u MAE counters: %s",
+                               action_set->n_counters, rte_strerror(rc));
+               }
+
                sfc_mae_encap_header_disable(sa, action_set->encap_header);
        }
 
@@ -753,6 +908,8 @@ sfc_mae_rule_process_pattern_data(struct sfc_mae_parse_ctx *ctx,
 
        for (ethertype_idx = 0;
             ethertype_idx < pdata->nb_vlan_tags; ++ethertype_idx) {
+               rte_be16_t tpid_v = ethertypes[ethertype_idx].value;
+               rte_be16_t tpid_m = ethertypes[ethertype_idx].mask;
                unsigned int tpid_idx;
 
                /*
@@ -762,7 +919,7 @@ sfc_mae_rule_process_pattern_data(struct sfc_mae_parse_ctx *ctx,
                 */
                enforce_tag_presence[0] = B_FALSE;
 
-               if (ethertypes[ethertype_idx].mask == RTE_BE16(0)) {
+               if (tpid_m == RTE_BE16(0)) {
                        if (pdata->tci_masks[ethertype_idx] == RTE_BE16(0))
                                enforce_tag_presence[ethertype_idx] = B_TRUE;
 
@@ -772,19 +929,22 @@ sfc_mae_rule_process_pattern_data(struct sfc_mae_parse_ctx *ctx,
                }
 
                /* Exact match is supported only. */
-               if (ethertypes[ethertype_idx].mask != RTE_BE16(0xffff)) {
+               if (tpid_m != RTE_BE16(0xffff)) {
+                       sfc_err(ctx->sa, "TPID mask must be 0x0 or 0xffff; got 0x%04x",
+                               rte_be_to_cpu_16(tpid_m));
                        rc = EINVAL;
                        goto fail;
                }
 
                for (tpid_idx = pdata->nb_vlan_tags - ethertype_idx - 1;
                     tpid_idx < nb_supported_tpids; ++tpid_idx) {
-                       if (ethertypes[ethertype_idx].value ==
-                           supported_tpids[tpid_idx])
+                       if (tpid_v == supported_tpids[tpid_idx])
                                break;
                }
 
                if (tpid_idx == nb_supported_tpids) {
+                       sfc_err(ctx->sa, "TPID 0x%04x is unsupported",
+                               rte_be_to_cpu_16(tpid_v));
                        rc = EINVAL;
                        goto fail;
                }
@@ -794,14 +954,19 @@ sfc_mae_rule_process_pattern_data(struct sfc_mae_parse_ctx *ctx,
 
        if (pdata->innermost_ethertype_restriction.mask == RTE_BE16(0xffff)) {
                struct sfc_mae_ethertype *et = &ethertypes[ethertype_idx];
+               rte_be16_t enforced_et;
+
+               enforced_et = pdata->innermost_ethertype_restriction.value;
 
                if (et->mask == 0) {
                        et->mask = RTE_BE16(0xffff);
-                       et->value =
-                           pdata->innermost_ethertype_restriction.value;
+                       et->value = enforced_et;
                } else if (et->mask != RTE_BE16(0xffff) ||
-                          et->value !=
-                          pdata->innermost_ethertype_restriction.value) {
+                          et->value != enforced_et) {
+                       sfc_err(ctx->sa, "L3 EtherType must be 0x0/0x0 or 0x%04x/0xffff; got 0x%04x/0x%04x",
+                               rte_be_to_cpu_16(enforced_et),
+                               rte_be_to_cpu_16(et->value),
+                               rte_be_to_cpu_16(et->mask));
                        rc = EINVAL;
                        goto fail;
                }
@@ -821,10 +986,14 @@ sfc_mae_rule_process_pattern_data(struct sfc_mae_parse_ctx *ctx,
                if (pdata->l3_next_proto_mask == 0) {
                        pdata->l3_next_proto_mask = 0xff;
                        pdata->l3_next_proto_value =
-                           pdata->l3_next_proto_restriction_value;
+                               pdata->l3_next_proto_restriction_value;
                } else if (pdata->l3_next_proto_mask != 0xff ||
                           pdata->l3_next_proto_value !=
                           pdata->l3_next_proto_restriction_value) {
+                       sfc_err(ctx->sa, "L3 next protocol must be 0x0/0x0 or 0x%02x/0xff; got 0x%02x/0x%02x",
+                               pdata->l3_next_proto_restriction_value,
+                               pdata->l3_next_proto_value,
+                               pdata->l3_next_proto_mask);
                        rc = EINVAL;
                        goto fail;
                }
@@ -1816,6 +1985,7 @@ sfc_mae_rule_parse_item_tunnel(const struct rte_flow_item *item,
 static const struct sfc_flow_item sfc_flow_items[] = {
        {
                .type = RTE_FLOW_ITEM_TYPE_PORT_ID,
+               .name = "PORT_ID",
                /*
                 * In terms of RTE flow, this item is a META one,
                 * and its position in the pattern is don't care.
@@ -1827,6 +1997,7 @@ static const struct sfc_flow_item sfc_flow_items[] = {
        },
        {
                .type = RTE_FLOW_ITEM_TYPE_PHY_PORT,
+               .name = "PHY_PORT",
                /*
                 * In terms of RTE flow, this item is a META one,
                 * and its position in the pattern is don't care.
@@ -1838,6 +2009,7 @@ static const struct sfc_flow_item sfc_flow_items[] = {
        },
        {
                .type = RTE_FLOW_ITEM_TYPE_PF,
+               .name = "PF",
                /*
                 * In terms of RTE flow, this item is a META one,
                 * and its position in the pattern is don't care.
@@ -1849,6 +2021,7 @@ static const struct sfc_flow_item sfc_flow_items[] = {
        },
        {
                .type = RTE_FLOW_ITEM_TYPE_VF,
+               .name = "VF",
                /*
                 * In terms of RTE flow, this item is a META one,
                 * and its position in the pattern is don't care.
@@ -1860,6 +2033,7 @@ static const struct sfc_flow_item sfc_flow_items[] = {
        },
        {
                .type = RTE_FLOW_ITEM_TYPE_ETH,
+               .name = "ETH",
                .prev_layer = SFC_FLOW_ITEM_START_LAYER,
                .layer = SFC_FLOW_ITEM_L2,
                .ctx_type = SFC_FLOW_PARSE_CTX_MAE,
@@ -1867,6 +2041,7 @@ static const struct sfc_flow_item sfc_flow_items[] = {
        },
        {
                .type = RTE_FLOW_ITEM_TYPE_VLAN,
+               .name = "VLAN",
                .prev_layer = SFC_FLOW_ITEM_L2,
                .layer = SFC_FLOW_ITEM_L2,
                .ctx_type = SFC_FLOW_PARSE_CTX_MAE,
@@ -1874,6 +2049,7 @@ static const struct sfc_flow_item sfc_flow_items[] = {
        },
        {
                .type = RTE_FLOW_ITEM_TYPE_IPV4,
+               .name = "IPV4",
                .prev_layer = SFC_FLOW_ITEM_L2,
                .layer = SFC_FLOW_ITEM_L3,
                .ctx_type = SFC_FLOW_PARSE_CTX_MAE,
@@ -1881,6 +2057,7 @@ static const struct sfc_flow_item sfc_flow_items[] = {
        },
        {
                .type = RTE_FLOW_ITEM_TYPE_IPV6,
+               .name = "IPV6",
                .prev_layer = SFC_FLOW_ITEM_L2,
                .layer = SFC_FLOW_ITEM_L3,
                .ctx_type = SFC_FLOW_PARSE_CTX_MAE,
@@ -1888,6 +2065,7 @@ static const struct sfc_flow_item sfc_flow_items[] = {
        },
        {
                .type = RTE_FLOW_ITEM_TYPE_TCP,
+               .name = "TCP",
                .prev_layer = SFC_FLOW_ITEM_L3,
                .layer = SFC_FLOW_ITEM_L4,
                .ctx_type = SFC_FLOW_PARSE_CTX_MAE,
@@ -1895,6 +2073,7 @@ static const struct sfc_flow_item sfc_flow_items[] = {
        },
        {
                .type = RTE_FLOW_ITEM_TYPE_UDP,
+               .name = "UDP",
                .prev_layer = SFC_FLOW_ITEM_L3,
                .layer = SFC_FLOW_ITEM_L4,
                .ctx_type = SFC_FLOW_PARSE_CTX_MAE,
@@ -1902,6 +2081,7 @@ static const struct sfc_flow_item sfc_flow_items[] = {
        },
        {
                .type = RTE_FLOW_ITEM_TYPE_VXLAN,
+               .name = "VXLAN",
                .prev_layer = SFC_FLOW_ITEM_L4,
                .layer = SFC_FLOW_ITEM_START_LAYER,
                .ctx_type = SFC_FLOW_PARSE_CTX_MAE,
@@ -1909,6 +2089,7 @@ static const struct sfc_flow_item sfc_flow_items[] = {
        },
        {
                .type = RTE_FLOW_ITEM_TYPE_GENEVE,
+               .name = "GENEVE",
                .prev_layer = SFC_FLOW_ITEM_L4,
                .layer = SFC_FLOW_ITEM_START_LAYER,
                .ctx_type = SFC_FLOW_PARSE_CTX_MAE,
@@ -1916,6 +2097,7 @@ static const struct sfc_flow_item sfc_flow_items[] = {
        },
        {
                .type = RTE_FLOW_ITEM_TYPE_NVGRE,
+               .name = "NVGRE",
                .prev_layer = SFC_FLOW_ITEM_L3,
                .layer = SFC_FLOW_ITEM_START_LAYER,
                .ctx_type = SFC_FLOW_PARSE_CTX_MAE,
@@ -2124,7 +2306,7 @@ sfc_mae_rule_parse_pattern(struct sfc_adapter *sa,
        if (rc != 0)
                goto fail_encap_parse_init;
 
-       rc = sfc_flow_parse_pattern(sfc_flow_items, RTE_DIM(sfc_flow_items),
+       rc = sfc_flow_parse_pattern(sa, sfc_flow_items, RTE_DIM(sfc_flow_items),
                                    pattern, &ctx, error);
        if (rc != 0)
                goto fail_parse_pattern;
@@ -2375,6 +2557,7 @@ sfc_mae_rule_parse_action_vxlan_encap(
        size_t next_proto_ofst = 0;
        size_t ethertype_ofst = 0;
        uint64_t exp_items;
+       int rc;
 
        if (pattern == NULL) {
                return rte_flow_error_set(error, EINVAL,
@@ -2594,14 +2777,69 @@ sfc_mae_rule_parse_action_vxlan_encap(
        /* Take care of the masks. */
        sfc_mae_header_force_item_masks(buf, parsed_items, nb_parsed_items);
 
-       return (spec != NULL) ? efx_mae_action_set_populate_encap(spec) : 0;
+       rc = efx_mae_action_set_populate_encap(spec);
+       if (rc != 0) {
+               rc = rte_flow_error_set(error, rc, RTE_FLOW_ERROR_TYPE_ACTION,
+                               NULL, "failed to request action ENCAP");
+       }
+
+       return rc;
 }
 
 static int
-sfc_mae_rule_parse_action_mark(const struct rte_flow_action_mark *conf,
+sfc_mae_rule_parse_action_mark(struct sfc_adapter *sa,
+                              const struct rte_flow_action_mark *conf,
                               efx_mae_actions_t *spec)
 {
-       return efx_mae_action_set_populate_mark(spec, conf->id);
+       int rc;
+
+       rc = efx_mae_action_set_populate_mark(spec, conf->id);
+       if (rc != 0)
+               sfc_err(sa, "failed to request action MARK: %s", strerror(rc));
+
+       return rc;
+}
+
+static int
+sfc_mae_rule_parse_action_count(struct sfc_adapter *sa,
+                               const struct rte_flow_action_count *conf,
+                               efx_mae_actions_t *spec)
+{
+       int rc;
+
+       if (conf->shared) {
+               rc = ENOTSUP;
+               goto fail_counter_shared;
+       }
+
+       if ((sa->counter_rxq.state & SFC_COUNTER_RXQ_INITIALIZED) == 0) {
+               sfc_err(sa,
+                       "counter queue is not configured for COUNT action");
+               rc = EINVAL;
+               goto fail_counter_queue_uninit;
+       }
+
+       if (sfc_get_service_lcore(SOCKET_ID_ANY) == RTE_MAX_LCORE) {
+               rc = EINVAL;
+               goto fail_no_service_core;
+       }
+
+       rc = efx_mae_action_set_populate_count(spec);
+       if (rc != 0) {
+               sfc_err(sa,
+                       "failed to populate counters in MAE action set: %s",
+                       rte_strerror(rc));
+               goto fail_populate_count;
+       }
+
+       return 0;
+
+fail_populate_count:
+fail_no_service_core:
+fail_counter_queue_uninit:
+fail_counter_shared:
+
+       return rc;
 }
 
 static int
@@ -2619,10 +2857,19 @@ sfc_mae_rule_parse_action_phy_port(struct sfc_adapter *sa,
                phy_port = conf->index;
 
        rc = efx_mae_mport_by_phy_port(phy_port, &mport);
-       if (rc != 0)
+       if (rc != 0) {
+               sfc_err(sa, "failed to convert phys. port ID %u to m-port selector: %s",
+                       phy_port, strerror(rc));
                return rc;
+       }
 
-       return efx_mae_action_set_populate_deliver(spec, &mport);
+       rc = efx_mae_action_set_populate_deliver(spec, &mport);
+       if (rc != 0) {
+               sfc_err(sa, "failed to request action DELIVER with m-port selector 0x%08x: %s",
+                       mport.sel, strerror(rc));
+       }
+
+       return rc;
 }
 
 static int
@@ -2643,10 +2890,20 @@ sfc_mae_rule_parse_action_pf_vf(struct sfc_adapter *sa,
                vf = vf_conf->id;
 
        rc = efx_mae_mport_by_pcie_function(encp->enc_pf, vf, &mport);
-       if (rc != 0)
+       if (rc != 0) {
+               sfc_err(sa, "failed to convert PF %u VF %d to m-port: %s",
+                       encp->enc_pf, (vf != EFX_PCI_VF_INVALID) ? (int)vf : -1,
+                       strerror(rc));
                return rc;
+       }
 
-       return efx_mae_action_set_populate_deliver(spec, &mport);
+       rc = efx_mae_action_set_populate_deliver(spec, &mport);
+       if (rc != 0) {
+               sfc_err(sa, "failed to request action DELIVER with m-port selector 0x%08x: %s",
+                       mport.sel, strerror(rc));
+       }
+
+       return rc;
 }
 
 static int
@@ -2667,12 +2924,37 @@ sfc_mae_rule_parse_action_port_id(struct sfc_adapter *sa,
 
        rc = sfc_mae_switch_port_by_ethdev(mae->switch_domain_id,
                                           port_id, &mport);
-       if (rc != 0)
+       if (rc != 0) {
+               sfc_err(sa, "failed to find MAE switch port SW entry for RTE ethdev port %u: %s",
+                       port_id, strerror(rc));
                return rc;
+       }
 
-       return efx_mae_action_set_populate_deliver(spec, &mport);
+       rc = efx_mae_action_set_populate_deliver(spec, &mport);
+       if (rc != 0) {
+               sfc_err(sa, "failed to request action DELIVER with m-port selector 0x%08x: %s",
+                       mport.sel, strerror(rc));
+       }
+
+       return rc;
 }
 
+static const char * const action_names[] = {
+       [RTE_FLOW_ACTION_TYPE_VXLAN_DECAP] = "VXLAN_DECAP",
+       [RTE_FLOW_ACTION_TYPE_OF_POP_VLAN] = "OF_POP_VLAN",
+       [RTE_FLOW_ACTION_TYPE_OF_PUSH_VLAN] = "OF_PUSH_VLAN",
+       [RTE_FLOW_ACTION_TYPE_OF_SET_VLAN_VID] = "OF_SET_VLAN_VID",
+       [RTE_FLOW_ACTION_TYPE_OF_SET_VLAN_PCP] = "OF_SET_VLAN_PCP",
+       [RTE_FLOW_ACTION_TYPE_VXLAN_ENCAP] = "VXLAN_ENCAP",
+       [RTE_FLOW_ACTION_TYPE_FLAG] = "FLAG",
+       [RTE_FLOW_ACTION_TYPE_MARK] = "MARK",
+       [RTE_FLOW_ACTION_TYPE_PHY_PORT] = "PHY_PORT",
+       [RTE_FLOW_ACTION_TYPE_PF] = "PF",
+       [RTE_FLOW_ACTION_TYPE_VF] = "VF",
+       [RTE_FLOW_ACTION_TYPE_PORT_ID] = "PORT_ID",
+       [RTE_FLOW_ACTION_TYPE_DROP] = "DROP",
+};
+
 static int
 sfc_mae_rule_parse_action(struct sfc_adapter *sa,
                          const struct rte_flow_action *action,
@@ -2722,6 +3004,11 @@ sfc_mae_rule_parse_action(struct sfc_adapter *sa,
                                                           spec, error);
                custom_error = B_TRUE;
                break;
+       case RTE_FLOW_ACTION_TYPE_COUNT:
+               SFC_BUILD_SET_OVERFLOW(RTE_FLOW_ACTION_TYPE_COUNT,
+                                      bundle->actions_mask);
+               rc = sfc_mae_rule_parse_action_count(sa, action->conf, spec);
+               break;
        case RTE_FLOW_ACTION_TYPE_FLAG:
                SFC_BUILD_SET_OVERFLOW(RTE_FLOW_ACTION_TYPE_FLAG,
                                       bundle->actions_mask);
@@ -2730,7 +3017,7 @@ sfc_mae_rule_parse_action(struct sfc_adapter *sa,
        case RTE_FLOW_ACTION_TYPE_MARK:
                SFC_BUILD_SET_OVERFLOW(RTE_FLOW_ACTION_TYPE_MARK,
                                       bundle->actions_mask);
-               rc = sfc_mae_rule_parse_action_mark(action->conf, spec);
+               rc = sfc_mae_rule_parse_action_mark(sa, action->conf, spec);
                break;
        case RTE_FLOW_ACTION_TYPE_PHY_PORT:
                SFC_BUILD_SET_OVERFLOW(RTE_FLOW_ACTION_TYPE_PHY_PORT,
@@ -2766,6 +3053,14 @@ sfc_mae_rule_parse_action(struct sfc_adapter *sa,
        if (rc == 0) {
                bundle->actions_mask |= (1ULL << action->type);
        } else if (!custom_error) {
+               if (action->type < RTE_DIM(action_names)) {
+                       const char *action_name = action_names[action->type];
+
+                       if (action_name != NULL) {
+                               sfc_err(sa, "action %s was rejected: %s",
+                                       action_name, strerror(rc));
+                       }
+               }
                rc = rte_flow_error_set(error, rc, RTE_FLOW_ERROR_TYPE_ACTION,
                                NULL, "Failed to request the action");
        }
@@ -2807,6 +3102,7 @@ sfc_mae_rule_parse_actions(struct sfc_adapter *sa,
        const struct rte_flow_action *action;
        struct sfc_mae *mae = &sa->mae;
        efx_mae_actions_t *spec;
+       unsigned int n_count;
        int rc;
 
        rte_errno = 0;
@@ -2844,15 +3140,22 @@ sfc_mae_rule_parse_actions(struct sfc_adapter *sa,
        if (rc != 0)
                goto fail_process_encap_header;
 
+       n_count = efx_mae_action_set_get_nb_count(spec);
+       if (n_count > 1) {
+               rc = ENOTSUP;
+               sfc_err(sa, "too many count actions requested: %u", n_count);
+               goto fail_nb_count;
+       }
+
        spec_mae->action_set = sfc_mae_action_set_attach(sa, encap_header,
-                                                        spec);
+                                                        n_count, spec);
        if (spec_mae->action_set != NULL) {
                sfc_mae_encap_header_del(sa, encap_header);
                efx_mae_action_set_spec_fini(sa->nic, spec);
                return 0;
        }
 
-       rc = sfc_mae_action_set_add(sa, spec, encap_header,
+       rc = sfc_mae_action_set_add(sa, actions, spec, encap_header, n_count,
                                    &spec_mae->action_set);
        if (rc != 0)
                goto fail_action_set_add;
@@ -2860,6 +3163,7 @@ sfc_mae_rule_parse_actions(struct sfc_adapter *sa,
        return 0;
 
 fail_action_set_add:
+fail_nb_count:
        sfc_mae_encap_header_del(sa, encap_header);
 
 fail_process_encap_header:
@@ -3014,6 +3318,15 @@ sfc_mae_flow_insert(struct sfc_adapter *sa,
        if (rc != 0)
                goto fail_action_set_enable;
 
+       if (action_set->n_counters > 0) {
+               rc = sfc_mae_counter_start(sa);
+               if (rc != 0) {
+                       sfc_err(sa, "failed to start MAE counters support: %s",
+                               rte_strerror(rc));
+                       goto fail_mae_counter_start;
+               }
+       }
+
        rc = efx_mae_action_rule_insert(sa->nic, spec_mae->match_spec,
                                        NULL, &fw_rsrc->aset_id,
                                        &spec_mae->rule_id);
@@ -3026,6 +3339,7 @@ sfc_mae_flow_insert(struct sfc_adapter *sa,
        return 0;
 
 fail_action_rule_insert:
+fail_mae_counter_start:
        sfc_mae_action_set_disable(sa, action_set);
 
 fail_action_set_enable:
@@ -3065,3 +3379,67 @@ sfc_mae_flow_remove(struct sfc_adapter *sa,
 
        return 0;
 }
+
+static int
+sfc_mae_query_counter(struct sfc_adapter *sa,
+                     struct sfc_flow_spec_mae *spec,
+                     const struct rte_flow_action *action,
+                     struct rte_flow_query_count *data,
+                     struct rte_flow_error *error)
+{
+       struct sfc_mae_action_set *action_set = spec->action_set;
+       const struct rte_flow_action_count *conf = action->conf;
+       unsigned int i;
+       int rc;
+
+       if (action_set->n_counters == 0) {
+               return rte_flow_error_set(error, EINVAL,
+                       RTE_FLOW_ERROR_TYPE_ACTION, action,
+                       "Queried flow rule does not have count actions");
+       }
+
+       for (i = 0; i < action_set->n_counters; i++) {
+               /*
+                * Get the first available counter of the flow rule if
+                * counter ID is not specified.
+                */
+               if (conf != NULL && action_set->counters[i].rte_id != conf->id)
+                       continue;
+
+               rc = sfc_mae_counter_get(&sa->mae.counter_registry.counters,
+                                        &action_set->counters[i], data);
+               if (rc != 0) {
+                       return rte_flow_error_set(error, EINVAL,
+                               RTE_FLOW_ERROR_TYPE_ACTION, action,
+                               "Queried flow rule counter action is invalid");
+               }
+
+               return 0;
+       }
+
+       return rte_flow_error_set(error, ENOENT,
+                                 RTE_FLOW_ERROR_TYPE_ACTION, action,
+                                 "No such flow rule action count ID");
+}
+
+int
+sfc_mae_flow_query(struct rte_eth_dev *dev,
+                  struct rte_flow *flow,
+                  const struct rte_flow_action *action,
+                  void *data,
+                  struct rte_flow_error *error)
+{
+       struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev);
+       struct sfc_flow_spec *spec = &flow->spec;
+       struct sfc_flow_spec_mae *spec_mae = &spec->mae;
+
+       switch (action->type) {
+       case RTE_FLOW_ACTION_TYPE_COUNT:
+               return sfc_mae_query_counter(sa, spec_mae, action,
+                                            data, error);
+       default:
+               return rte_flow_error_set(error, ENOTSUP,
+                       RTE_FLOW_ERROR_TYPE_ACTION, NULL,
+                       "Query for action of this type is not supported");
+       }
+}