net/mlx5: support Sub-Function
[dpdk.git] / drivers / net / mlx5 / mlx5_flow_meter.c
index b11962f..78eb2a6 100644 (file)
@@ -4,13 +4,79 @@
  */
 #include <math.h>
 
+#include <rte_tailq.h>
 #include <rte_malloc.h>
 #include <rte_mtr.h>
 #include <rte_mtr_driver.h>
 
+#include <mlx5_devx_cmds.h>
+#include <mlx5_malloc.h>
+
 #include "mlx5.h"
 #include "mlx5_flow.h"
 
+/**
+ * Create the meter action.
+ *
+ * @param priv
+ *   Pointer to mlx5_priv.
+ * @param[in] fm
+ *   Pointer to flow meter to be converted.
+ *
+ * @return
+ *   Pointer to the meter action on success, NULL otherwise.
+ */
+static void *
+mlx5_flow_meter_action_create(struct mlx5_priv *priv,
+                             struct mlx5_flow_meter_info *fm)
+{
+#ifdef HAVE_MLX5_DR_CREATE_ACTION_FLOW_METER
+       struct mlx5dv_dr_flow_meter_attr mtr_init;
+       uint32_t fmp[MLX5_ST_SZ_DW(flow_meter_parameters)];
+       struct mlx5_flow_meter_srtcm_rfc2697_prm *srtcm =
+                                                    &fm->profile->srtcm_prm;
+       uint32_t cbs_cir = rte_be_to_cpu_32(srtcm->cbs_cir);
+       uint32_t ebs_eir = rte_be_to_cpu_32(srtcm->ebs_eir);
+       uint32_t val;
+       enum mlx5_meter_domain domain =
+               fm->transfer ? MLX5_MTR_DOMAIN_TRANSFER :
+                       fm->egress ? MLX5_MTR_DOMAIN_EGRESS :
+                               MLX5_MTR_DOMAIN_INGRESS;
+       struct mlx5_flow_meter_def_policy *def_policy =
+               priv->sh->mtrmng->def_policy[domain];
+
+       memset(fmp, 0, MLX5_ST_SZ_BYTES(flow_meter_parameters));
+       MLX5_SET(flow_meter_parameters, fmp, valid, 1);
+       MLX5_SET(flow_meter_parameters, fmp, bucket_overflow, 1);
+       MLX5_SET(flow_meter_parameters, fmp,
+               start_color, MLX5_FLOW_COLOR_GREEN);
+       MLX5_SET(flow_meter_parameters, fmp, both_buckets_on_green, 0);
+       val = (cbs_cir >> ASO_DSEG_CBS_EXP_OFFSET) & ASO_DSEG_EXP_MASK;
+       MLX5_SET(flow_meter_parameters, fmp, cbs_exponent, val);
+       val = (cbs_cir >> ASO_DSEG_CBS_MAN_OFFSET) & ASO_DSEG_MAN_MASK;
+       MLX5_SET(flow_meter_parameters, fmp, cbs_mantissa, val);
+       val = (cbs_cir >> ASO_DSEG_CIR_EXP_OFFSET) & ASO_DSEG_EXP_MASK;
+       MLX5_SET(flow_meter_parameters, fmp, cir_exponent, val);
+       val = (cbs_cir & ASO_DSEG_MAN_MASK);
+       MLX5_SET(flow_meter_parameters, fmp, cir_mantissa, val);
+       val = (ebs_eir >> ASO_DSEG_EBS_EXP_OFFSET) & ASO_DSEG_EXP_MASK;
+       MLX5_SET(flow_meter_parameters, fmp, ebs_exponent, val);
+       val = (ebs_eir >> ASO_DSEG_EBS_MAN_OFFSET) & ASO_DSEG_MAN_MASK;
+       MLX5_SET(flow_meter_parameters, fmp, ebs_mantissa, val);
+       mtr_init.next_table = def_policy->sub_policy.tbl_rsc->obj;
+       mtr_init.reg_c_index = priv->mtr_color_reg - REG_C_0;
+       mtr_init.flow_meter_parameter = fmp;
+       mtr_init.flow_meter_parameter_sz =
+               MLX5_ST_SZ_BYTES(flow_meter_parameters);
+       mtr_init.active = fm->active_state;
+       return mlx5_glue->dv_create_flow_action_meter(&mtr_init);
+#else
+       (void)priv;
+       (void)fm;
+       return NULL;
+#endif
+}
+
 /**
  * Find meter profile by id.
  *
 static struct mlx5_flow_meter_profile *
 mlx5_flow_meter_profile_find(struct mlx5_priv *priv, uint32_t meter_profile_id)
 {
-       struct mlx5_mtr_profiles *fmps = &priv->flow_meter_profiles;
        struct mlx5_flow_meter_profile *fmp;
+       union mlx5_l3t_data data;
+       int32_t ret;
 
-       TAILQ_FOREACH(fmp, fmps, next)
-               if (meter_profile_id == fmp->meter_profile_id)
-                       return fmp;
-       return NULL;
+       if (mlx5_l3t_get_entry(priv->mtr_profile_tbl,
+                              meter_profile_id, &data) || !data.ptr)
+               return NULL;
+       fmp = data.ptr;
+       /* Remove reference taken by the mlx5_l3t_get_entry. */
+       ret = mlx5_l3t_clear_entry(priv->mtr_profile_tbl,
+                                  meter_profile_id);
+       if (!ret || ret == -1)
+               return NULL;
+       return fmp;
+}
+
+/**
+ * Validate the MTR profile.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[in] meter_profile_id
+ *   Meter profile id.
+ * @param[in] profile
+ *   Pointer to meter profile detail.
+ * @param[out] error
+ *   Pointer to the error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_profile_validate(struct rte_eth_dev *dev,
+                                uint32_t meter_profile_id,
+                                struct rte_mtr_meter_profile *profile,
+                                struct rte_mtr_error *error)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_flow_meter_profile *fmp;
+
+       /* Profile must not be NULL. */
+       if (profile == NULL)
+               return -rte_mtr_error_set(error, EINVAL,
+                                         RTE_MTR_ERROR_TYPE_METER_PROFILE,
+                                         NULL, "Meter profile is null.");
+       /* Meter profile ID must be valid. */
+       if (meter_profile_id == UINT32_MAX)
+               return -rte_mtr_error_set(error, EINVAL,
+                                         RTE_MTR_ERROR_TYPE_METER_PROFILE_ID,
+                                         NULL, "Meter profile id not valid.");
+       /* Meter profile must not exist. */
+       fmp = mlx5_flow_meter_profile_find(priv, meter_profile_id);
+       if (fmp)
+               return -rte_mtr_error_set(error, EEXIST,
+                                         RTE_MTR_ERROR_TYPE_METER_PROFILE_ID,
+                                         NULL,
+                                         "Meter profile already exists.");
+       if (profile->alg == RTE_MTR_SRTCM_RFC2697) {
+               if (priv->config.hca_attr.qos.flow_meter_old) {
+                       /* Verify support for flow meter parameters. */
+                       if (priv->sh->meter_aso_en && profile->packet_mode) {
+                               if (profile->srtcm_rfc2697.cir > 0 &&
+                                       (profile->srtcm_rfc2697.cir <<
+                                       MLX5_MTRS_PPS_MAP_BPS_SHIFT)
+                                       <= MLX5_SRTCM_CIR_MAX &&
+                                       profile->srtcm_rfc2697.cbs > 0 &&
+                                       (profile->srtcm_rfc2697.cbs <<
+                                       MLX5_MTRS_PPS_MAP_BPS_SHIFT)
+                                       <= MLX5_SRTCM_CBS_MAX &&
+                                       (profile->srtcm_rfc2697.ebs <<
+                                       MLX5_MTRS_PPS_MAP_BPS_SHIFT)
+                                       <= MLX5_SRTCM_EBS_MAX)
+                                       return 0;
+                               return -rte_mtr_error_set
+                                            (error, ENOTSUP,
+                                             RTE_MTR_ERROR_TYPE_MTR_PARAMS,
+                                             NULL,
+                                             profile->srtcm_rfc2697.ebs ?
+                                             "Metering value ebs must be 0." :
+                                             "Invalid metering parameters.");
+                       }
+                       if (profile->srtcm_rfc2697.cir > 0 &&
+                               profile->srtcm_rfc2697.cir <=
+                                               MLX5_SRTCM_CIR_MAX &&
+                               profile->srtcm_rfc2697.cbs > 0 &&
+                               profile->srtcm_rfc2697.cbs <=
+                                               MLX5_SRTCM_CBS_MAX &&
+                               profile->srtcm_rfc2697.ebs <=
+                                               MLX5_SRTCM_EBS_MAX)
+                               return 0;
+                       return -rte_mtr_error_set(error, ENOTSUP,
+                                       RTE_MTR_ERROR_TYPE_MTR_PARAMS,
+                                       NULL,
+                                       profile->srtcm_rfc2697.ebs ?
+                                       "Metering value ebs must be 0." :
+                                       "Invalid metering parameters.");
+               }
+       }
+       return -rte_mtr_error_set(error, ENOTSUP,
+                                 RTE_MTR_ERROR_TYPE_METER_PROFILE,
+                                 NULL, "Metering algorithm not supported.");
 }
 
 /**
@@ -110,48 +270,69 @@ mlx5_flow_meter_xbs_man_exp_calc(uint64_t xbs, uint8_t *man, uint8_t *exp)
  */
 static int
 mlx5_flow_meter_param_fill(struct mlx5_flow_meter_profile *fmp,
-                         struct rte_mtr_error *error)
+                       struct mlx5_priv *priv, struct rte_mtr_error *error)
 {
        struct mlx5_flow_meter_srtcm_rfc2697_prm *srtcm = &fmp->srtcm_prm;
        uint8_t man, exp;
+       uint32_t cbs_exp, cbs_man, cir_exp, cir_man;
+       uint32_t ebs_exp, ebs_man;
+       uint64_t cir, cbs, ebs;
 
        if (fmp->profile.alg != RTE_MTR_SRTCM_RFC2697)
                return -rte_mtr_error_set(error, ENOTSUP,
                                RTE_MTR_ERROR_TYPE_METER_PROFILE,
                                NULL, "Metering algorithm not supported.");
+       if (!priv->sh->meter_aso_en && fmp->profile.packet_mode)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                       RTE_MTR_ERROR_TYPE_METER_PROFILE,
+                       NULL, "Metering algorithm packet mode not supported.");
+       if (priv->sh->meter_aso_en && fmp->profile.packet_mode) {
+               cir = fmp->profile.srtcm_rfc2697.cir <<
+                               MLX5_MTRS_PPS_MAP_BPS_SHIFT;
+               cbs = fmp->profile.srtcm_rfc2697.cbs <<
+                               MLX5_MTRS_PPS_MAP_BPS_SHIFT;
+               ebs = fmp->profile.srtcm_rfc2697.ebs <<
+                               MLX5_MTRS_PPS_MAP_BPS_SHIFT;
+       } else {
+               cir = fmp->profile.srtcm_rfc2697.cir;
+               cbs = fmp->profile.srtcm_rfc2697.cbs;
+               ebs = fmp->profile.srtcm_rfc2697.ebs;
+       }
+       /* cir = 8G * cir_mantissa * 1/(2^cir_exponent)) Bytes/Sec */
+       mlx5_flow_meter_cir_man_exp_calc(cir, &man, &exp);
+       /* Check if cir mantissa is too large. */
+       if (exp > ASO_DSEG_CIR_EXP_MASK)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                                         RTE_MTR_ERROR_TYPE_MTR_PARAMS, NULL,
+                                         "meter profile parameter cir is"
+                                         " not supported.");
+       cir_man = man;
+       cir_exp = exp;
         /* cbs = cbs_mantissa * 2^cbs_exponent */
-       mlx5_flow_meter_xbs_man_exp_calc(fmp->profile.srtcm_rfc2697.cbs,
-                                   &man, &exp);
-       srtcm->cbs_mantissa = man;
-       srtcm->cbs_exponent = exp;
+       mlx5_flow_meter_xbs_man_exp_calc(cbs, &man, &exp);
        /* Check if cbs mantissa is too large. */
-       if (srtcm->cbs_exponent != exp)
-               return -rte_mtr_error_set(error, EINVAL,
+       if (exp > ASO_DSEG_EXP_MASK)
+               return -rte_mtr_error_set(error, ENOTSUP,
                                          RTE_MTR_ERROR_TYPE_MTR_PARAMS, NULL,
-                                         "Metering profile parameter cbs is"
-                                         " invalid.");
-       /* ebs = ebs_mantissa * 2^ebs_exponent */
-       mlx5_flow_meter_xbs_man_exp_calc(fmp->profile.srtcm_rfc2697.ebs,
-                                   &man, &exp);
-       srtcm->ebs_mantissa = man;
-       srtcm->ebs_exponent = exp;
+                                         "meter profile parameter cbs is"
+                                         " not supported.");
+       cbs_man = man;
+       cbs_exp = exp;
+       srtcm->cbs_cir = rte_cpu_to_be_32(cbs_exp << ASO_DSEG_CBS_EXP_OFFSET |
+                               cbs_man << ASO_DSEG_CBS_MAN_OFFSET |
+                               cir_exp << ASO_DSEG_CIR_EXP_OFFSET |
+                               cir_man);
+       mlx5_flow_meter_xbs_man_exp_calc(ebs, &man, &exp);
        /* Check if ebs mantissa is too large. */
-       if (srtcm->ebs_exponent != exp)
-               return -rte_mtr_error_set(error, EINVAL,
-                                         RTE_MTR_ERROR_TYPE_MTR_PARAMS, NULL,
-                                         "Metering profile parameter ebs is"
-                                         " invalid.");
-       /* cir = 8G * cir_mantissa * 1/(2^cir_exponent)) Bytes/Sec */
-       mlx5_flow_meter_cir_man_exp_calc(fmp->profile.srtcm_rfc2697.cir,
-                                   &man, &exp);
-       srtcm->cir_mantissa = man;
-       srtcm->cir_exponent = exp;
-       /* Check if cir mantissa is too large. */
-       if (srtcm->cir_exponent != exp)
-               return -rte_mtr_error_set(error, EINVAL,
+       if (exp > ASO_DSEG_EXP_MASK)
+               return -rte_mtr_error_set(error, ENOTSUP,
                                          RTE_MTR_ERROR_TYPE_MTR_PARAMS, NULL,
-                                         "Metering profile parameter cir is"
-                                         " invalid.");
+                                         "meter profile parameter ebs is"
+                                         " not supported.");
+       ebs_man = man;
+       ebs_exp = exp;
+       srtcm->ebs_eir = rte_cpu_to_be_32(ebs_exp << ASO_DSEG_EBS_EXP_OFFSET |
+                                       ebs_man << ASO_DSEG_EBS_MAN_OFFSET);
        return 0;
 }
 
@@ -179,18 +360,25 @@ mlx5_flow_mtr_cap_get(struct rte_eth_dev *dev,
        if (!priv->mtr_en)
                return -rte_mtr_error_set(error, ENOTSUP,
                                          RTE_MTR_ERROR_TYPE_UNSPECIFIED, NULL,
-                                         "Meter is not support");
+                                         "Meter is not supported");
        memset(cap, 0, sizeof(*cap));
-       cap->n_max = 1 << qattr->log_max_flow_meter;
+       if (priv->sh->meter_aso_en) {
+               /* 2 meters per one ASO cache line. */
+               cap->n_max = 1 << (qattr->log_max_num_meter_aso + 1);
+               cap->srtcm_rfc2697_packet_mode_supported = 1;
+       } else {
+               cap->n_max = 1 << qattr->log_max_flow_meter;
+               cap->srtcm_rfc2697_packet_mode_supported = 0;
+       }
+       cap->srtcm_rfc2697_byte_mode_supported = 1;
        cap->n_shared_max = cap->n_max;
        cap->identical = 1;
        cap->shared_identical = 1;
        cap->shared_n_flows_per_mtr_max = 4 << 20;
        /* 2M flows can share the same meter. */
        cap->chaining_n_mtrs_per_flow_max = 1; /* Chaining is not supported. */
-       cap->meter_srtcm_rfc2697_n_max = qattr->srtcm_sup ? cap->n_max : 0;
+       cap->meter_srtcm_rfc2697_n_max = qattr->flow_meter_old ? cap->n_max : 0;
        cap->meter_rate_max = 1ULL << 40; /* 1 Tera tokens per sec. */
-       cap->policer_action_drop_supported = 1;
        cap->stats_mask = RTE_MTR_STATS_N_BYTES_DROPPED |
                          RTE_MTR_STATS_N_PKTS_DROPPED;
        return 0;
@@ -218,34 +406,44 @@ mlx5_flow_meter_profile_add(struct rte_eth_dev *dev,
                       struct rte_mtr_error *error)
 {
        struct mlx5_priv *priv = dev->data->dev_private;
-       struct mlx5_mtr_profiles *fmps = &priv->flow_meter_profiles;
        struct mlx5_flow_meter_profile *fmp;
+       union mlx5_l3t_data data;
        int ret;
 
        if (!priv->mtr_en)
                return -rte_mtr_error_set(error, ENOTSUP,
                                          RTE_MTR_ERROR_TYPE_UNSPECIFIED, NULL,
-                                         "Meter is not support");
+                                         "Meter is not supported");
+       /* Check input params. */
+       ret = mlx5_flow_meter_profile_validate(dev, meter_profile_id,
+                                              profile, error);
+       if (ret)
+               return ret;
        /* Meter profile memory allocation. */
-       fmp = rte_calloc(__func__, 1, sizeof(struct mlx5_flow_meter_profile),
-                        RTE_CACHE_LINE_SIZE);
+       fmp = mlx5_malloc(MLX5_MEM_ZERO, sizeof(struct mlx5_flow_meter_profile),
+                        RTE_CACHE_LINE_SIZE, SOCKET_ID_ANY);
        if (fmp == NULL)
                return -rte_mtr_error_set(error, ENOMEM,
                                          RTE_MTR_ERROR_TYPE_UNSPECIFIED,
                                          NULL, "Meter profile memory "
                                          "alloc failed.");
        /* Fill profile info. */
-       fmp->meter_profile_id = meter_profile_id;
+       fmp->id = meter_profile_id;
        fmp->profile = *profile;
        /* Fill the flow meter parameters for the PRM. */
-       ret = mlx5_flow_meter_param_fill(fmp, error);
+       ret = mlx5_flow_meter_param_fill(fmp, priv, error);
        if (ret)
                goto error;
-       /* Add to list. */
-       TAILQ_INSERT_TAIL(fmps, fmp, next);
+       data.ptr = fmp;
+       ret = mlx5_l3t_set_entry(priv->mtr_profile_tbl,
+                                meter_profile_id, &data);
+       if (ret)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                                         RTE_MTR_ERROR_TYPE_UNSPECIFIED,
+                                         NULL, "Meter profile insert fail.");
        return 0;
 error:
-       rte_free(fmp);
+       mlx5_free(fmp);
        return ret;
 }
 
@@ -273,54 +471,1650 @@ mlx5_flow_meter_profile_delete(struct rte_eth_dev *dev,
        if (!priv->mtr_en)
                return -rte_mtr_error_set(error, ENOTSUP,
                                          RTE_MTR_ERROR_TYPE_UNSPECIFIED, NULL,
-                                         "Meter is not support");
+                                         "Meter is not supported");
        /* Meter profile must exist. */
        fmp = mlx5_flow_meter_profile_find(priv, meter_profile_id);
        if (fmp == NULL)
                return -rte_mtr_error_set(error, ENOENT,
                                          RTE_MTR_ERROR_TYPE_METER_PROFILE_ID,
                                          &meter_profile_id,
-                                         "Meter profile id invalid.");
+                                         "Meter profile id is invalid.");
        /* Check profile is unused. */
        if (fmp->ref_cnt)
                return -rte_mtr_error_set(error, EBUSY,
                                          RTE_MTR_ERROR_TYPE_METER_PROFILE_ID,
-                                         NULL, "Meter profile in use.");
-       /* Remove from list. */
-       TAILQ_REMOVE(&priv->flow_meter_profiles, fmp, next);
-       rte_free(fmp);
+                                         NULL, "Meter profile is in use.");
+       if (mlx5_l3t_clear_entry(priv->mtr_profile_tbl, meter_profile_id))
+               return -rte_mtr_error_set(error, EBUSY,
+                                         RTE_MTR_ERROR_TYPE_METER_PROFILE_ID,
+                                         NULL, "Meter profile remove fail.");
+       mlx5_free(fmp);
        return 0;
 }
 
-static const struct rte_mtr_ops mlx5_flow_mtr_ops = {
-       .capabilities_get = mlx5_flow_mtr_cap_get,
-       .meter_profile_add = mlx5_flow_meter_profile_add,
-       .meter_profile_delete = mlx5_flow_meter_profile_delete,
-       .create = NULL,
-       .destroy = NULL,
-       .meter_enable = NULL,
-       .meter_disable = NULL,
-       .meter_profile_update = NULL,
-       .meter_dscp_table_update = NULL,
-       .policer_actions_update = NULL,
-       .stats_update = NULL,
-       .stats_read = NULL,
-};
+/**
+ * Find policy by id.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param policy_id
+ *   Policy id.
+ *
+ * @return
+ *   Pointer to the policy found on success, NULL otherwise.
+ */
+struct mlx5_flow_meter_policy *
+mlx5_flow_meter_policy_find(struct rte_eth_dev *dev,
+                           uint32_t policy_id,
+                           uint32_t *policy_idx)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_flow_meter_sub_policy *sub_policy = NULL;
+       union mlx5_l3t_data data;
+
+       if (policy_id > MLX5_MAX_SUB_POLICY_TBL_NUM || !priv->policy_idx_tbl)
+               return NULL;
+       if (mlx5_l3t_get_entry(priv->policy_idx_tbl, policy_id, &data) ||
+                               !data.dword)
+               return NULL;
+       if (policy_idx)
+               *policy_idx = data.dword;
+       sub_policy = mlx5_ipool_get(priv->sh->ipool[MLX5_IPOOL_MTR_POLICY],
+                                       data.dword);
+       /* Remove reference taken by the mlx5_l3t_get_entry. */
+       mlx5_l3t_clear_entry(priv->policy_idx_tbl, policy_id);
+       if (sub_policy)
+               if (sub_policy->main_policy_id)
+                       return sub_policy->main_policy;
+       return NULL;
+}
 
 /**
- * Get meter operations.
+ * Get the last meter's policy from one meter's policy in hierarchy.
  *
- * @param dev
- *   Pointer to Ethernet device structure.
- * @param arg
- *   Pointer to set the mtr operations.
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[in] policy
+ *   Pointer to flow meter policy.
  *
  * @return
- *   Always 0.
+ *   Pointer to the final meter's policy, or NULL when fail.
  */
-int
-mlx5_flow_meter_ops_get(struct rte_eth_dev *dev __rte_unused, void *arg)
+struct mlx5_flow_meter_policy *
+mlx5_flow_meter_hierarchy_get_final_policy(struct rte_eth_dev *dev,
+                                       struct mlx5_flow_meter_policy *policy)
 {
-       *(const struct rte_mtr_ops **)arg = &mlx5_flow_mtr_ops;
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_flow_meter_info *next_fm;
+       struct mlx5_flow_meter_policy *next_policy = policy;
+
+       while (next_policy->is_hierarchy) {
+               next_fm = mlx5_flow_meter_find(priv,
+                      next_policy->act_cnt[RTE_COLOR_GREEN].next_mtr_id, NULL);
+               if (!next_fm || next_fm->def_policy)
+                       return NULL;
+               next_policy = mlx5_flow_meter_policy_find(dev,
+                                               next_fm->policy_id, NULL);
+               MLX5_ASSERT(next_policy);
+       }
+       return next_policy;
+}
+
+/**
+ * Callback to check MTR policy action validate
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[in] actions
+ *   Pointer to meter policy action detail.
+ * @param[out] error
+ *   Pointer to the error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_policy_validate(struct rte_eth_dev *dev,
+       struct rte_mtr_meter_policy_params *policy,
+       struct rte_mtr_error *error)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct rte_flow_attr attr = { .transfer =
+                       priv->config.dv_esw_en ? 1 : 0};
+       bool is_rss = false;
+       bool is_def_policy = false;
+       uint8_t domain_bitmap;
+       int ret;
+
+       if (!priv->mtr_en || !priv->sh->meter_aso_en)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                               RTE_MTR_ERROR_TYPE_METER_POLICY,
+                               NULL, "meter policy unsupported.");
+       ret = mlx5_flow_validate_mtr_acts(dev, policy->actions, &attr,
+                       &is_rss, &domain_bitmap, &is_def_policy, error);
+       if (ret)
+               return ret;
+       return 0;
+}
+
+static int
+__mlx5_flow_meter_policy_delete(struct rte_eth_dev *dev,
+                       uint32_t policy_id,
+                       struct mlx5_flow_meter_policy *mtr_policy,
+                       struct rte_mtr_error *error,
+                       bool clear_l3t)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_flow_meter_sub_policy *sub_policy;
+       uint32_t i, j;
+       uint16_t sub_policy_num;
+
+       rte_spinlock_lock(&mtr_policy->sl);
+       if (mtr_policy->ref_cnt) {
+               rte_spinlock_unlock(&mtr_policy->sl);
+               return -rte_mtr_error_set(error, EBUSY,
+                               RTE_MTR_ERROR_TYPE_METER_POLICY_ID,
+                                NULL,
+                               "Meter policy object is being used.");
+       }
+       mlx5_flow_destroy_policy_rules(dev, mtr_policy);
+       mlx5_flow_destroy_mtr_acts(dev, mtr_policy);
+       for (i = 0; i < MLX5_MTR_DOMAIN_MAX; i++) {
+               sub_policy_num = (mtr_policy->sub_policy_num >>
+                       (MLX5_MTR_SUB_POLICY_NUM_SHIFT * i)) &
+                       MLX5_MTR_SUB_POLICY_NUM_MASK;
+               if (sub_policy_num) {
+                       for (j = 0; j < sub_policy_num; j++) {
+                               sub_policy = mtr_policy->sub_policys[i][j];
+                               if (sub_policy)
+                                       mlx5_ipool_free
+                                       (priv->sh->ipool[MLX5_IPOOL_MTR_POLICY],
+                                       sub_policy->idx);
+                       }
+               }
+       }
+       if (priv->policy_idx_tbl && clear_l3t) {
+               if (mlx5_l3t_clear_entry(priv->policy_idx_tbl, policy_id)) {
+                       rte_spinlock_unlock(&mtr_policy->sl);
+                       return -rte_mtr_error_set(error, ENOTSUP,
+                               RTE_MTR_ERROR_TYPE_METER_POLICY_ID, NULL,
+                               "Fail to delete policy in index table.");
+               }
+       }
+       rte_spinlock_unlock(&mtr_policy->sl);
+       return 0;
+}
+
+/**
+ * Callback to add MTR policy.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[out] policy_id
+ *   Pointer to policy id
+ * @param[in] actions
+ *   Pointer to meter policy action detail.
+ * @param[out] error
+ *   Pointer to the error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_policy_add(struct rte_eth_dev *dev,
+                       uint32_t policy_id,
+                       struct rte_mtr_meter_policy_params *policy,
+                       struct rte_mtr_error *error)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct rte_flow_attr attr = { .transfer =
+                       priv->config.dv_esw_en ? 1 : 0};
+       uint32_t sub_policy_idx = 0;
+       uint32_t policy_idx = 0;
+       struct mlx5_flow_meter_policy *mtr_policy = NULL;
+       struct mlx5_flow_meter_sub_policy *sub_policy;
+       bool is_rss = false;
+       bool is_def_policy = false;
+       uint32_t i;
+       int ret;
+       uint32_t policy_size = sizeof(struct mlx5_flow_meter_policy);
+       uint16_t sub_policy_num;
+       uint8_t domain_bitmap = 0;
+       union mlx5_l3t_data data;
+       bool skip_rule = false;
+
+       if (!priv->mtr_en)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                                         RTE_MTR_ERROR_TYPE_METER_POLICY,
+                                         NULL, "meter policy unsupported.");
+       if (policy_id == MLX5_INVALID_POLICY_ID)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                       RTE_MTR_ERROR_TYPE_METER_POLICY_ID, NULL,
+                       "policy ID is invalid. ");
+       if (policy_id == priv->sh->mtrmng->def_policy_id)
+               return -rte_mtr_error_set(error, EEXIST,
+                       RTE_MTR_ERROR_TYPE_METER_POLICY_ID, NULL,
+                       "policy ID exists. ");
+       mtr_policy = mlx5_flow_meter_policy_find(dev, policy_id,
+                               &policy_idx);
+       if (mtr_policy)
+               return -rte_mtr_error_set(error, EEXIST,
+                       RTE_MTR_ERROR_TYPE_METER_POLICY_ID, NULL,
+                       "policy ID exists. ");
+       ret = mlx5_flow_validate_mtr_acts(dev, policy->actions, &attr,
+                       &is_rss, &domain_bitmap, &is_def_policy, error);
+       if (ret)
+               return ret;
+       if (!domain_bitmap)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                               RTE_MTR_ERROR_TYPE_METER_POLICY,
+                               NULL, "fail to find policy domain.");
+       if (is_def_policy) {
+               if (priv->sh->mtrmng->def_policy_id != MLX5_INVALID_POLICY_ID)
+                       return -rte_mtr_error_set(error, EEXIST,
+                               RTE_MTR_ERROR_TYPE_METER_POLICY_ID,
+                               NULL, "a policy with similar actions "
+                               "is already configured");
+               if (mlx5_flow_create_def_policy(dev))
+                       return -rte_mtr_error_set(error, ENOTSUP,
+                               RTE_MTR_ERROR_TYPE_METER_POLICY,
+                               NULL,
+                               "fail to create non-terminated policy.");
+               priv->sh->mtrmng->def_policy_id = policy_id;
+               return 0;
+       }
+       if (!priv->sh->meter_aso_en)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                       RTE_MTR_ERROR_TYPE_METER_POLICY, NULL,
+                       "no ASO capability to support the policy ");
+       for (i = 0; i < MLX5_MTR_DOMAIN_MAX; i++) {
+               if (!(domain_bitmap & (1 << i)))
+                       continue;
+               if (is_rss) {
+                       policy_size +=
+                       sizeof(struct mlx5_flow_meter_sub_policy *) *
+                       MLX5_MTR_RSS_MAX_SUB_POLICY;
+                       break;
+               }
+               policy_size += sizeof(struct mlx5_flow_meter_sub_policy *);
+       }
+       mtr_policy = mlx5_malloc(MLX5_MEM_ZERO, policy_size,
+                        RTE_CACHE_LINE_SIZE, SOCKET_ID_ANY);
+       if (!mtr_policy)
+               return -rte_mtr_error_set(error, ENOMEM,
+                               RTE_MTR_ERROR_TYPE_METER_POLICY, NULL,
+                               "Memory alloc failed for meter policy.");
+       policy_size = sizeof(struct mlx5_flow_meter_policy);
+       for (i = 0; i < MLX5_MTR_DOMAIN_MAX; i++) {
+               if (!(domain_bitmap & (1 << i)))
+                       continue;
+               if (i == MLX5_MTR_DOMAIN_INGRESS)
+                       mtr_policy->ingress = 1;
+               if (i == MLX5_MTR_DOMAIN_EGRESS)
+                       mtr_policy->egress = 1;
+               if (i == MLX5_MTR_DOMAIN_TRANSFER)
+                       mtr_policy->transfer = 1;
+               sub_policy = mlx5_ipool_zmalloc
+                               (priv->sh->ipool[MLX5_IPOOL_MTR_POLICY],
+                               &sub_policy_idx);
+               if (!sub_policy)
+                       goto policy_add_err;
+               if (sub_policy_idx > MLX5_MAX_SUB_POLICY_TBL_NUM)
+                       goto policy_add_err;
+               sub_policy->idx = sub_policy_idx;
+               sub_policy->main_policy = mtr_policy;
+               if (!policy_idx) {
+                       policy_idx = sub_policy_idx;
+                       sub_policy->main_policy_id = 1;
+               }
+               mtr_policy->sub_policys[i] =
+               (struct mlx5_flow_meter_sub_policy **)
+                       ((uint8_t *)mtr_policy + policy_size);
+               mtr_policy->sub_policys[i][0] = sub_policy;
+               sub_policy_num = (mtr_policy->sub_policy_num >>
+                       (MLX5_MTR_SUB_POLICY_NUM_SHIFT * i)) &
+                       MLX5_MTR_SUB_POLICY_NUM_MASK;
+               sub_policy_num++;
+               mtr_policy->sub_policy_num &= ~(MLX5_MTR_SUB_POLICY_NUM_MASK <<
+                       (MLX5_MTR_SUB_POLICY_NUM_SHIFT * i));
+               mtr_policy->sub_policy_num |=
+                       (sub_policy_num & MLX5_MTR_SUB_POLICY_NUM_MASK) <<
+                       (MLX5_MTR_SUB_POLICY_NUM_SHIFT * i);
+               if (is_rss) {
+                       mtr_policy->is_rss = 1;
+                       break;
+               }
+               policy_size += sizeof(struct mlx5_flow_meter_sub_policy  *);
+       }
+       rte_spinlock_init(&mtr_policy->sl);
+       ret = mlx5_flow_create_mtr_acts(dev, mtr_policy,
+                                       policy->actions, error);
+       if (ret)
+               goto policy_add_err;
+       if (mtr_policy->is_hierarchy) {
+               struct mlx5_flow_meter_policy *final_policy;
+
+               final_policy =
+               mlx5_flow_meter_hierarchy_get_final_policy(dev, mtr_policy);
+               if (!final_policy)
+                       goto policy_add_err;
+               skip_rule = (final_policy->is_rss || final_policy->is_queue);
+       }
+       if (!is_rss && !mtr_policy->is_queue && !skip_rule) {
+               /* Create policy rules in HW. */
+               ret = mlx5_flow_create_policy_rules(dev, mtr_policy);
+               if (ret)
+                       goto policy_add_err;
+       }
+       data.dword = policy_idx;
+       if (!priv->policy_idx_tbl) {
+               priv->policy_idx_tbl = mlx5_l3t_create(MLX5_L3T_TYPE_DWORD);
+               if (!priv->policy_idx_tbl)
+                       goto policy_add_err;
+       }
+       if (mlx5_l3t_set_entry(priv->policy_idx_tbl, policy_id, &data))
+               goto policy_add_err;
+       return 0;
+policy_add_err:
+       if (mtr_policy) {
+               ret = __mlx5_flow_meter_policy_delete(dev, policy_id,
+                       mtr_policy, error, false);
+               mlx5_free(mtr_policy);
+               if (ret)
+                       return ret;
+       }
+       return -rte_mtr_error_set(error, ENOTSUP,
+                                 RTE_MTR_ERROR_TYPE_UNSPECIFIED,
+                                 NULL, "Failed to create devx policy.");
+}
+
+/**
+ * Callback to delete MTR policy.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[in] policy_id
+ *   Meter policy id.
+ * @param[out] error
+ *   Pointer to the error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_policy_delete(struct rte_eth_dev *dev,
+                         uint32_t policy_id,
+                         struct rte_mtr_error *error)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_flow_meter_policy *mtr_policy;
+       uint32_t policy_idx;
+       int ret;
+
+       if (policy_id == priv->sh->mtrmng->def_policy_id) {
+               if (priv->sh->mtrmng->def_policy_ref_cnt > 0)
+                       return -rte_mtr_error_set(error, ENOTSUP,
+                               RTE_MTR_ERROR_TYPE_METER_POLICY_ID, NULL,
+                               "Meter policy object is being used.");
+               priv->sh->mtrmng->def_policy_id = MLX5_INVALID_POLICY_ID;
+               return 0;
+       }
+       mtr_policy = mlx5_flow_meter_policy_find(dev, policy_id, &policy_idx);
+       if (!mtr_policy)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                       RTE_MTR_ERROR_TYPE_METER_POLICY_ID, NULL,
+                       "Meter policy id is invalid. ");
+       ret = __mlx5_flow_meter_policy_delete(dev, policy_id, mtr_policy,
+                                               error, true);
+       if (ret)
+               return ret;
+       mlx5_free(mtr_policy);
+       return 0;
+}
+
+/**
+ * Check meter validation.
+ *
+ * @param[in] priv
+ *   Pointer to mlx5 private data structure.
+ * @param[in] meter_id
+ *   Meter id.
+ * @param[in] params
+ *   Pointer to rte meter parameters.
+ * @param[out] error
+ *   Pointer to rte meter error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_validate(struct mlx5_priv *priv, uint32_t meter_id,
+                        struct rte_mtr_params *params,
+                        struct rte_mtr_error *error)
+{
+       /* Meter must use global drop action. */
+       if (!priv->sh->dr_drop_action)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                                         RTE_MTR_ERROR_TYPE_MTR_PARAMS,
+                                         NULL,
+                                         "No drop action ready for meter.");
+       /* Meter params must not be NULL. */
+       if (params == NULL)
+               return -rte_mtr_error_set(error, EINVAL,
+                                         RTE_MTR_ERROR_TYPE_MTR_PARAMS,
+                                         NULL, "Meter object params null.");
+       /* Previous meter color is not supported. */
+       if (params->use_prev_mtr_color)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                                         RTE_MTR_ERROR_TYPE_MTR_PARAMS,
+                                         NULL,
+                                         "Previous meter color "
+                                         "not supported.");
+       if (params->meter_policy_id == MLX5_INVALID_POLICY_ID)
+               return -rte_mtr_error_set(error, ENOENT,
+                               RTE_MTR_ERROR_TYPE_METER_POLICY_ID,
+                               NULL, "Meter policy id not valid.");
+       /* Validate meter id. */
+       if (mlx5_flow_meter_find(priv, meter_id, NULL))
+               return -rte_mtr_error_set(error, EEXIST,
+                       RTE_MTR_ERROR_TYPE_MTR_ID, NULL,
+                       "Meter object already exists.");
+       return 0;
+}
+
+/**
+ * Modify the flow meter action.
+ *
+ * @param[in] priv
+ *   Pointer to mlx5 private data structure.
+ * @param[in] fm
+ *   Pointer to flow meter to be modified.
+ * @param[in] srtcm
+ *   Pointer to meter srtcm description parameter.
+ * @param[in] modify_bits
+ *   The bit in srtcm to be updated.
+ * @param[in] active_state
+ *   The state to be updated.
+ * @return
+ *   0 on success, o negative value otherwise.
+ */
+static int
+mlx5_flow_meter_action_modify(struct mlx5_priv *priv,
+               struct mlx5_flow_meter_info *fm,
+               const struct mlx5_flow_meter_srtcm_rfc2697_prm *srtcm,
+               uint64_t modify_bits, uint32_t active_state, uint32_t is_enable)
+{
+#ifdef HAVE_MLX5_DR_CREATE_ACTION_FLOW_METER
+       uint32_t in[MLX5_ST_SZ_DW(flow_meter_parameters)] = { 0 };
+       uint32_t *attr;
+       struct mlx5dv_dr_flow_meter_attr mod_attr = { 0 };
+       int ret;
+       struct mlx5_aso_mtr *aso_mtr = NULL;
+       uint32_t cbs_cir, ebs_eir, val;
+
+       if (priv->sh->meter_aso_en) {
+               fm->is_enable = !!is_enable;
+               aso_mtr = container_of(fm, struct mlx5_aso_mtr, fm);
+               ret = mlx5_aso_meter_update_by_wqe(priv->sh, aso_mtr);
+               if (ret)
+                       return ret;
+               ret = mlx5_aso_mtr_wait(priv->sh, aso_mtr);
+               if (ret)
+                       return ret;
+       } else {
+               /* Fill command parameters. */
+               mod_attr.reg_c_index = priv->mtr_color_reg - REG_C_0;
+               mod_attr.flow_meter_parameter = in;
+               mod_attr.flow_meter_parameter_sz =
+                               MLX5_ST_SZ_BYTES(flow_meter_parameters);
+               if (modify_bits & MLX5_FLOW_METER_OBJ_MODIFY_FIELD_ACTIVE)
+                       mod_attr.active = !!active_state;
+               else
+                       mod_attr.active = 0;
+               attr = in;
+               cbs_cir = rte_be_to_cpu_32(srtcm->cbs_cir);
+               ebs_eir = rte_be_to_cpu_32(srtcm->ebs_eir);
+               if (modify_bits & MLX5_FLOW_METER_OBJ_MODIFY_FIELD_CBS) {
+                       val = (cbs_cir >> ASO_DSEG_CBS_EXP_OFFSET) &
+                               ASO_DSEG_EXP_MASK;
+                       MLX5_SET(flow_meter_parameters, attr,
+                               cbs_exponent, val);
+                       val = (cbs_cir >> ASO_DSEG_CBS_MAN_OFFSET) &
+                               ASO_DSEG_MAN_MASK;
+                       MLX5_SET(flow_meter_parameters, attr,
+                               cbs_mantissa, val);
+               }
+               if (modify_bits & MLX5_FLOW_METER_OBJ_MODIFY_FIELD_CIR) {
+                       val = (cbs_cir >> ASO_DSEG_CIR_EXP_OFFSET) &
+                               ASO_DSEG_EXP_MASK;
+                       MLX5_SET(flow_meter_parameters, attr,
+                               cir_exponent, val);
+                       val = cbs_cir & ASO_DSEG_MAN_MASK;
+                       MLX5_SET(flow_meter_parameters, attr,
+                               cir_mantissa, val);
+               }
+               if (modify_bits & MLX5_FLOW_METER_OBJ_MODIFY_FIELD_EBS) {
+                       val = (ebs_eir >> ASO_DSEG_EBS_EXP_OFFSET) &
+                               ASO_DSEG_EXP_MASK;
+                       MLX5_SET(flow_meter_parameters, attr,
+                               ebs_exponent, val);
+                       val = (ebs_eir >> ASO_DSEG_EBS_MAN_OFFSET) &
+                               ASO_DSEG_MAN_MASK;
+                       MLX5_SET(flow_meter_parameters, attr,
+                               ebs_mantissa, val);
+               }
+               /* Apply modifications to meter only if it was created. */
+               if (fm->meter_action) {
+                       ret = mlx5_glue->dv_modify_flow_action_meter
+                                       (fm->meter_action, &mod_attr,
+                                       rte_cpu_to_be_64(modify_bits));
+                       if (ret)
+                               return ret;
+               }
+               /* Update succeedded modify meter parameters. */
+               if (modify_bits & MLX5_FLOW_METER_OBJ_MODIFY_FIELD_ACTIVE)
+                       fm->active_state = !!active_state;
+       }
+       return 0;
+#else
+       (void)priv;
+       (void)fm;
+       (void)srtcm;
+       (void)modify_bits;
+       (void)active_state;
+       (void)is_enable;
+       return -ENOTSUP;
+#endif
+}
+
+static int
+mlx5_flow_meter_stats_enable_update(struct rte_eth_dev *dev,
+                               struct mlx5_flow_meter_info *fm,
+                               uint64_t stats_mask)
+{
+       fm->bytes_dropped =
+               (stats_mask & RTE_MTR_STATS_N_BYTES_DROPPED) ? 1 : 0;
+       fm->pkts_dropped = (stats_mask & RTE_MTR_STATS_N_PKTS_DROPPED) ? 1 : 0;
+       if (fm->bytes_dropped || fm->pkts_dropped) {
+               if (!fm->drop_cnt) {
+                       /* Alloc policer counters. */
+                       fm->drop_cnt = mlx5_counter_alloc(dev);
+                       if (!fm->drop_cnt)
+                               return -1;
+               }
+       } else {
+               if (fm->drop_cnt) {
+                       mlx5_counter_free(dev, fm->drop_cnt);
+                       fm->drop_cnt = 0;
+               }
+       }
+       return 0;
+}
+
+/**
+ * Create meter rules.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[in] meter_id
+ *   Meter id.
+ * @param[in] params
+ *   Pointer to rte meter parameters.
+ * @param[in] shared
+ *   Meter shared with other flow or not.
+ * @param[out] error
+ *   Pointer to rte meter error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_create(struct rte_eth_dev *dev, uint32_t meter_id,
+                      struct rte_mtr_params *params, int shared,
+                      struct rte_mtr_error *error)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_legacy_flow_meters *fms = &priv->flow_meters;
+       struct mlx5_flow_meter_profile *fmp;
+       struct mlx5_flow_meter_info *fm;
+       struct mlx5_legacy_flow_meter *legacy_fm;
+       struct mlx5_flow_meter_policy *mtr_policy = NULL;
+       struct mlx5_indexed_pool_config flow_ipool_cfg = {
+               .size = 0,
+               .trunk_size = 64,
+               .need_lock = 1,
+               .type = "mlx5_flow_mtr_flow_id_pool",
+       };
+       struct mlx5_aso_mtr *aso_mtr;
+       uint32_t mtr_idx, policy_idx;
+       union mlx5_l3t_data data;
+       int ret;
+       uint8_t domain_bitmap;
+       uint8_t mtr_id_bits;
+       uint8_t mtr_reg_bits = priv->mtr_reg_share ?
+                               MLX5_MTR_IDLE_BITS_IN_COLOR_REG : MLX5_REG_BITS;
+
+       if (!priv->mtr_en)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                                       RTE_MTR_ERROR_TYPE_UNSPECIFIED, NULL,
+                                       "Meter is not supported");
+       /* Validate the parameters. */
+       ret = mlx5_flow_meter_validate(priv, meter_id, params, error);
+       if (ret)
+               return ret;
+       /* Meter profile must exist. */
+       fmp = mlx5_flow_meter_profile_find(priv, params->meter_profile_id);
+       if (fmp == NULL)
+               return -rte_mtr_error_set(error, ENOENT,
+                       RTE_MTR_ERROR_TYPE_METER_PROFILE_ID,
+                       NULL, "Meter profile id not valid.");
+       /* Meter policy must exist. */
+       if (params->meter_policy_id == priv->sh->mtrmng->def_policy_id) {
+               __atomic_add_fetch
+                       (&priv->sh->mtrmng->def_policy_ref_cnt,
+                       1, __ATOMIC_RELAXED);
+               domain_bitmap = MLX5_MTR_ALL_DOMAIN_BIT;
+               if (!priv->config.dv_esw_en)
+                       domain_bitmap &= ~MLX5_MTR_DOMAIN_TRANSFER_BIT;
+       } else {
+               mtr_policy = mlx5_flow_meter_policy_find(dev,
+                               params->meter_policy_id, &policy_idx);
+               if (!priv->sh->meter_aso_en)
+                       return -rte_mtr_error_set(error, ENOTSUP,
+                               RTE_MTR_ERROR_TYPE_UNSPECIFIED, NULL,
+                               "Part of the policies cannot be "
+                               "supported without ASO ");
+               if (!mtr_policy)
+                       return -rte_mtr_error_set(error, ENOENT,
+                               RTE_MTR_ERROR_TYPE_METER_POLICY_ID,
+                               NULL, "Meter policy id not valid.");
+               domain_bitmap = (mtr_policy->ingress ?
+                                       MLX5_MTR_DOMAIN_INGRESS_BIT : 0) |
+                               (mtr_policy->egress ?
+                                       MLX5_MTR_DOMAIN_EGRESS_BIT : 0) |
+                               (mtr_policy->transfer ?
+                                       MLX5_MTR_DOMAIN_TRANSFER_BIT : 0);
+       }
+       /* Allocate the flow meter memory. */
+       if (priv->sh->meter_aso_en) {
+               mtr_idx = mlx5_flow_mtr_alloc(dev);
+               if (!mtr_idx)
+                       return -rte_mtr_error_set(error, ENOMEM,
+                               RTE_MTR_ERROR_TYPE_UNSPECIFIED, NULL,
+                               "Memory alloc failed for meter.");
+               aso_mtr = mlx5_aso_meter_by_idx(priv, mtr_idx);
+               fm = &aso_mtr->fm;
+       } else {
+               legacy_fm = mlx5_ipool_zmalloc
+                               (priv->sh->ipool[MLX5_IPOOL_MTR], &mtr_idx);
+               if (legacy_fm == NULL)
+                       return -rte_mtr_error_set(error, ENOMEM,
+                               RTE_MTR_ERROR_TYPE_UNSPECIFIED, NULL,
+                               "Memory alloc failed for meter.");
+               legacy_fm->idx = mtr_idx;
+               fm = &legacy_fm->fm;
+       }
+       mtr_id_bits = MLX5_REG_BITS - __builtin_clz(mtr_idx);
+       if ((mtr_id_bits + priv->sh->mtrmng->max_mtr_flow_bits) >
+           mtr_reg_bits) {
+               DRV_LOG(ERR, "Meter number exceeds max limit.");
+               goto error;
+       }
+       if (mtr_id_bits > priv->sh->mtrmng->max_mtr_bits)
+               priv->sh->mtrmng->max_mtr_bits = mtr_id_bits;
+       /* Fill the flow meter parameters. */
+       fm->meter_id = meter_id;
+       fm->policy_id = params->meter_policy_id;
+       fm->profile = fmp;
+       if (mlx5_flow_meter_stats_enable_update(dev, fm, params->stats_mask))
+               goto error;
+       if (mlx5_flow_create_mtr_tbls(dev, fm, mtr_idx, domain_bitmap))
+               goto error;
+       /* Add to the flow meter list. */
+       if (!priv->sh->meter_aso_en)
+               TAILQ_INSERT_TAIL(fms, legacy_fm, next);
+       /* Add to the flow meter list. */
+       fm->active_state = 1; /* Config meter starts as active. */
+       fm->is_enable = 1;
+       fm->shared = !!shared;
+       __atomic_add_fetch(&fm->profile->ref_cnt, 1, __ATOMIC_RELAXED);
+       if (params->meter_policy_id == priv->sh->mtrmng->def_policy_id) {
+               fm->def_policy = 1;
+               fm->flow_ipool = mlx5_ipool_create(&flow_ipool_cfg);
+               if (!fm->flow_ipool)
+                       goto error;
+       }
+       rte_spinlock_init(&fm->sl);
+       /* If ASO meter supported, update ASO flow meter by wqe. */
+       if (priv->sh->meter_aso_en) {
+               aso_mtr = container_of(fm, struct mlx5_aso_mtr, fm);
+               ret = mlx5_aso_meter_update_by_wqe(priv->sh, aso_mtr);
+               if (ret)
+                       goto error;
+               if (!priv->mtr_idx_tbl) {
+                       priv->mtr_idx_tbl =
+                               mlx5_l3t_create(MLX5_L3T_TYPE_DWORD);
+                       if (!priv->mtr_idx_tbl)
+                               goto error;
+               }
+               data.dword = mtr_idx;
+               if (mlx5_l3t_set_entry(priv->mtr_idx_tbl, meter_id, &data))
+                       goto error;
+       }
+       if (mtr_policy)
+               __atomic_add_fetch(&mtr_policy->ref_cnt, 1, __ATOMIC_RELAXED);
+       return 0;
+error:
+       mlx5_flow_destroy_mtr_tbls(dev, fm);
+       /* Free policer counters. */
+       if (fm->drop_cnt)
+               mlx5_counter_free(dev, fm->drop_cnt);
+       if (priv->sh->meter_aso_en)
+               mlx5_flow_mtr_free(dev, mtr_idx);
+       else
+               mlx5_ipool_free(priv->sh->ipool[MLX5_IPOOL_MTR], mtr_idx);
+       return -rte_mtr_error_set(error, ENOTSUP,
+               RTE_MTR_ERROR_TYPE_UNSPECIFIED,
+               NULL, "Failed to create devx meter.");
+}
+
+static int
+mlx5_flow_meter_params_flush(struct rte_eth_dev *dev,
+                       struct mlx5_flow_meter_info *fm,
+                       uint32_t mtr_idx)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_legacy_flow_meters *fms = &priv->flow_meters;
+       struct mlx5_flow_meter_profile *fmp;
+       struct mlx5_legacy_flow_meter *legacy_fm = NULL;
+       struct mlx5_flow_meter_policy *mtr_policy;
+
+       /* Meter object must not have any owner. */
+       MLX5_ASSERT(!fm->ref_cnt);
+       /* Get meter profile. */
+       fmp = fm->profile;
+       if (fmp == NULL)
+               return -1;
+       /* Update dependencies. */
+       __atomic_sub_fetch(&fmp->ref_cnt, 1, __ATOMIC_RELAXED);
+       fm->profile = NULL;
+       /* Remove from list. */
+       if (!priv->sh->meter_aso_en) {
+               legacy_fm = container_of(fm,
+                       struct mlx5_legacy_flow_meter, fm);
+               TAILQ_REMOVE(fms, legacy_fm, next);
+       }
+       /* Free drop counters. */
+       if (fm->drop_cnt)
+               mlx5_counter_free(dev, fm->drop_cnt);
+       /* Free meter flow table. */
+       if (fm->flow_ipool) {
+               mlx5_ipool_destroy(fm->flow_ipool);
+               fm->flow_ipool = 0;
+       }
+       mlx5_flow_destroy_mtr_tbls(dev, fm);
+       if (fm->def_policy)
+               __atomic_sub_fetch(&priv->sh->mtrmng->def_policy_ref_cnt,
+                               1, __ATOMIC_RELAXED);
+       if (priv->sh->meter_aso_en) {
+               if (!fm->def_policy) {
+                       mtr_policy = mlx5_flow_meter_policy_find(dev,
+                                               fm->policy_id, NULL);
+                       if (mtr_policy)
+                               __atomic_sub_fetch(&mtr_policy->ref_cnt,
+                                               1, __ATOMIC_RELAXED);
+                       fm->policy_id = 0;
+               }
+               fm->def_policy = 0;
+               if (mlx5_l3t_clear_entry(priv->mtr_idx_tbl, fm->meter_id))
+                       return -1;
+               mlx5_flow_mtr_free(dev, mtr_idx);
+       } else {
+               mlx5_ipool_free(priv->sh->ipool[MLX5_IPOOL_MTR],
+                                       legacy_fm->idx);
+       }
+       return 0;
+}
+
+/**
+ * Destroy meter rules.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[in] meter_id
+ *   Meter id.
+ * @param[out] error
+ *   Pointer to rte meter error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_destroy(struct rte_eth_dev *dev, uint32_t meter_id,
+                       struct rte_mtr_error *error)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_flow_meter_info *fm;
+       uint32_t mtr_idx = 0;
+
+       if (!priv->mtr_en)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                                         RTE_MTR_ERROR_TYPE_UNSPECIFIED,
+                                         NULL,
+                                         "Meter is not supported");
+       /* Meter object must exist. */
+       fm = mlx5_flow_meter_find(priv, meter_id, &mtr_idx);
+       if (fm == NULL)
+               return -rte_mtr_error_set(error, ENOENT,
+                                         RTE_MTR_ERROR_TYPE_MTR_ID,
+                                         NULL,
+                                         "Meter object id not valid.");
+       /* Meter object must not have any owner. */
+       if (fm->ref_cnt > 0)
+               return -rte_mtr_error_set(error, EBUSY,
+                                         RTE_MTR_ERROR_TYPE_UNSPECIFIED,
+                                         NULL,
+                                         "Meter object is being used.");
+       /* Destroy the meter profile. */
+       if (mlx5_flow_meter_params_flush(dev, fm, mtr_idx))
+               return -rte_mtr_error_set(error, EINVAL,
+                                       RTE_MTR_ERROR_TYPE_METER_PROFILE_ID,
+                                       NULL,
+                                       "MTR object meter profile invalid.");
+       return 0;
+}
+
+/**
+ * Modify meter state.
+ *
+ * @param[in] priv
+ *   Pointer to mlx5 private data structure.
+ * @param[in] fm
+ *   Pointer to flow meter.
+ * @param[in] new_state
+ *   New state to update.
+ * @param[out] error
+ *   Pointer to rte meter error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_modify_state(struct mlx5_priv *priv,
+                            struct mlx5_flow_meter_info *fm,
+                            uint32_t new_state,
+                            struct rte_mtr_error *error)
+{
+       static const struct mlx5_flow_meter_srtcm_rfc2697_prm srtcm = {
+               .cbs_cir = RTE_BE32(MLX5_IFC_FLOW_METER_DISABLE_CBS_CIR_VAL),
+               .ebs_eir = 0,
+       };
+       uint64_t modify_bits = MLX5_FLOW_METER_OBJ_MODIFY_FIELD_CBS |
+                              MLX5_FLOW_METER_OBJ_MODIFY_FIELD_CIR;
+       int ret;
+
+       if (new_state == MLX5_FLOW_METER_DISABLE)
+               ret = mlx5_flow_meter_action_modify(priv, fm,
+                               &srtcm, modify_bits, 0, 0);
+       else
+               ret = mlx5_flow_meter_action_modify(priv, fm,
+                                                  &fm->profile->srtcm_prm,
+                                                   modify_bits, 0, 1);
+       if (ret)
+               return -rte_mtr_error_set(error, -ret,
+                                         RTE_MTR_ERROR_TYPE_MTR_PARAMS,
+                                         NULL,
+                                         new_state ?
+                                         "Failed to enable meter." :
+                                         "Failed to disable meter.");
+       return 0;
+}
+
+/**
+ * Callback to enable flow meter.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[in] meter_id
+ *   Meter id.
+ * @param[out] error
+ *   Pointer to rte meter error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_enable(struct rte_eth_dev *dev,
+                      uint32_t meter_id,
+                      struct rte_mtr_error *error)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_flow_meter_info *fm;
+       int ret;
+
+       if (!priv->mtr_en)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                                         RTE_MTR_ERROR_TYPE_UNSPECIFIED, NULL,
+                                         "Meter is not supported");
+       /* Meter object must exist. */
+       fm = mlx5_flow_meter_find(priv, meter_id, NULL);
+       if (fm == NULL)
+               return -rte_mtr_error_set(error, ENOENT,
+                                         RTE_MTR_ERROR_TYPE_MTR_ID,
+                                         NULL, "Meter not found.");
+       if (fm->active_state == MLX5_FLOW_METER_ENABLE)
+               return 0;
+       ret = mlx5_flow_meter_modify_state(priv, fm, MLX5_FLOW_METER_ENABLE,
+                                          error);
+       if (!ret)
+               fm->active_state = MLX5_FLOW_METER_ENABLE;
+       return ret;
+}
+
+/**
+ * Callback to disable flow meter.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[in] meter_id
+ *   Meter id.
+ * @param[out] error
+ *   Pointer to rte meter error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_disable(struct rte_eth_dev *dev,
+                       uint32_t meter_id,
+                       struct rte_mtr_error *error)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_flow_meter_info *fm;
+       int ret;
+
+       if (!priv->mtr_en)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                                         RTE_MTR_ERROR_TYPE_UNSPECIFIED, NULL,
+                                         "Meter is not supported");
+       /* Meter object must exist. */
+       fm = mlx5_flow_meter_find(priv, meter_id, NULL);
+       if (fm == NULL)
+               return -rte_mtr_error_set(error, ENOENT,
+                                         RTE_MTR_ERROR_TYPE_MTR_ID,
+                                         NULL, "Meter not found.");
+       if (fm->active_state == MLX5_FLOW_METER_DISABLE)
+               return 0;
+       ret = mlx5_flow_meter_modify_state(priv, fm, MLX5_FLOW_METER_DISABLE,
+                                          error);
+       if (!ret)
+               fm->active_state = MLX5_FLOW_METER_DISABLE;
+       return ret;
+}
+
+/**
+ * Callback to update meter profile.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[in] meter_id
+ *   Meter id.
+ * @param[in] meter_profile_id
+ *   To be updated meter profile id.
+ * @param[out] error
+ *   Pointer to rte meter error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_profile_update(struct rte_eth_dev *dev,
+                              uint32_t meter_id,
+                              uint32_t meter_profile_id,
+                              struct rte_mtr_error *error)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_flow_meter_profile *fmp;
+       struct mlx5_flow_meter_profile *old_fmp;
+       struct mlx5_flow_meter_info *fm;
+       uint64_t modify_bits = MLX5_FLOW_METER_OBJ_MODIFY_FIELD_CBS |
+                              MLX5_FLOW_METER_OBJ_MODIFY_FIELD_CIR;
+       int ret;
+
+       if (!priv->mtr_en)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                                         RTE_MTR_ERROR_TYPE_UNSPECIFIED, NULL,
+                                         "Meter is not supported");
+       /* Meter profile must exist. */
+       fmp = mlx5_flow_meter_profile_find(priv, meter_profile_id);
+       if (fmp == NULL)
+               return -rte_mtr_error_set(error, ENOENT,
+                                         RTE_MTR_ERROR_TYPE_METER_PROFILE_ID,
+                                         NULL, "Meter profile not found.");
+       /* Meter object must exist. */
+       fm = mlx5_flow_meter_find(priv, meter_id, NULL);
+       if (fm == NULL)
+               return -rte_mtr_error_set(error, ENOENT,
+                                         RTE_MTR_ERROR_TYPE_MTR_ID,
+                                         NULL, "Meter not found.");
+       /* MTR object already set to meter profile id. */
+       old_fmp = fm->profile;
+       if (fmp == old_fmp)
+               return 0;
+       /* Update the profile. */
+       fm->profile = fmp;
+       /* Update meter params in HW (if not disabled). */
+       if (fm->active_state == MLX5_FLOW_METER_DISABLE)
+               return 0;
+       ret = mlx5_flow_meter_action_modify(priv, fm, &fm->profile->srtcm_prm,
+                                             modify_bits, fm->active_state, 1);
+       if (ret) {
+               fm->profile = old_fmp;
+               return -rte_mtr_error_set(error, -ret,
+                                         RTE_MTR_ERROR_TYPE_MTR_PARAMS,
+                                         NULL, "Failed to update meter"
+                                         " parmeters in hardware.");
+       }
+       old_fmp->ref_cnt--;
+       fmp->ref_cnt++;
+       return 0;
+}
+
+/**
+ * Callback to update meter stats mask.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[in] meter_id
+ *   Meter id.
+ * @param[in] stats_mask
+ *   To be updated stats_mask.
+ * @param[out] error
+ *   Pointer to rte meter error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_stats_update(struct rte_eth_dev *dev,
+                            uint32_t meter_id,
+                            uint64_t stats_mask,
+                            struct rte_mtr_error *error)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_flow_meter_info *fm;
+
+       if (!priv->mtr_en)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                                         RTE_MTR_ERROR_TYPE_UNSPECIFIED, NULL,
+                                         "Meter is not supported");
+       /* Meter object must exist. */
+       fm = mlx5_flow_meter_find(priv, meter_id, NULL);
+       if (fm == NULL)
+               return -rte_mtr_error_set(error, ENOENT,
+                                         RTE_MTR_ERROR_TYPE_MTR_ID,
+                                         NULL, "Meter object id not valid.");
+       if (mlx5_flow_meter_stats_enable_update(dev, fm, stats_mask))
+               return -rte_mtr_error_set(error, ENOENT,
+                                         RTE_MTR_ERROR_TYPE_MTR_ID,
+                                         NULL, "Fail to allocate "
+                                         "counter for meter.");
+       return 0;
+}
+
+/**
+ * Callback to read meter statistics.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[in] meter_id
+ *   Meter id.
+ * @param[out] stats
+ *   Pointer to store the statistics.
+ * @param[out] stats_mask
+ *   Pointer to store the stats_mask.
+ * @param[in] clear
+ *   Statistic to be cleared after read or not.
+ * @param[out] error
+ *   Pointer to rte meter error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_stats_read(struct rte_eth_dev *dev,
+                          uint32_t meter_id,
+                          struct rte_mtr_stats *stats,
+                          uint64_t *stats_mask,
+                          int clear,
+                          struct rte_mtr_error *error)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_flow_meter_info *fm;
+       uint64_t pkts;
+       uint64_t bytes;
+       int ret = 0;
+
+       if (!priv->mtr_en)
+               return -rte_mtr_error_set(error, ENOTSUP,
+                                         RTE_MTR_ERROR_TYPE_UNSPECIFIED, NULL,
+                                         "Meter is not supported");
+       /* Meter object must exist. */
+       fm = mlx5_flow_meter_find(priv, meter_id, NULL);
+       if (fm == NULL)
+               return -rte_mtr_error_set(error, ENOENT,
+                                         RTE_MTR_ERROR_TYPE_MTR_ID,
+                                         NULL, "Meter object id not valid.");
+       *stats_mask = 0;
+       if (fm->bytes_dropped)
+               *stats_mask |= RTE_MTR_STATS_N_BYTES_DROPPED;
+       if (fm->pkts_dropped)
+               *stats_mask |= RTE_MTR_STATS_N_PKTS_DROPPED;
+       memset(stats, 0, sizeof(*stats));
+       if (fm->drop_cnt) {
+               ret = mlx5_counter_query(dev, fm->drop_cnt, clear, &pkts,
+                                                &bytes);
+               if (ret)
+                       goto error;
+               /* If need to read the packets, set it. */
+               if (fm->pkts_dropped)
+                       stats->n_pkts_dropped = pkts;
+               /* If need to read the bytes, set it. */
+               if (fm->bytes_dropped)
+                       stats->n_bytes_dropped = bytes;
+       }
+       return 0;
+error:
+       return -rte_mtr_error_set(error, ret, RTE_MTR_ERROR_TYPE_STATS, NULL,
+                                "Failed to read meter drop counters.");
+}
+
+static const struct rte_mtr_ops mlx5_flow_mtr_ops = {
+       .capabilities_get = mlx5_flow_mtr_cap_get,
+       .meter_profile_add = mlx5_flow_meter_profile_add,
+       .meter_profile_delete = mlx5_flow_meter_profile_delete,
+       .meter_policy_validate = mlx5_flow_meter_policy_validate,
+       .meter_policy_add = mlx5_flow_meter_policy_add,
+       .meter_policy_delete = mlx5_flow_meter_policy_delete,
+       .create = mlx5_flow_meter_create,
+       .destroy = mlx5_flow_meter_destroy,
+       .meter_enable = mlx5_flow_meter_enable,
+       .meter_disable = mlx5_flow_meter_disable,
+       .meter_profile_update = mlx5_flow_meter_profile_update,
+       .meter_dscp_table_update = NULL,
+       .stats_update = mlx5_flow_meter_stats_update,
+       .stats_read = mlx5_flow_meter_stats_read,
+};
+
+/**
+ * Get meter operations.
+ *
+ * @param dev
+ *   Pointer to Ethernet device structure.
+ * @param arg
+ *   Pointer to set the mtr operations.
+ *
+ * @return
+ *   Always 0.
+ */
+int
+mlx5_flow_meter_ops_get(struct rte_eth_dev *dev __rte_unused, void *arg)
+{
+       *(const struct rte_mtr_ops **)arg = &mlx5_flow_mtr_ops;
+       return 0;
+}
+
+/**
+ * Find meter by id.
+ *
+ * @param priv
+ *   Pointer to mlx5_priv.
+ * @param meter_id
+ *   Meter id.
+ * @param mtr_idx
+ *   Pointer to Meter index.
+ *
+ * @return
+ *   Pointer to the meter info found on success, NULL otherwise.
+ */
+struct mlx5_flow_meter_info *
+mlx5_flow_meter_find(struct mlx5_priv *priv, uint32_t meter_id,
+               uint32_t *mtr_idx)
+{
+       struct mlx5_legacy_flow_meter *legacy_fm;
+       struct mlx5_legacy_flow_meters *fms = &priv->flow_meters;
+       struct mlx5_aso_mtr *aso_mtr;
+       struct mlx5_aso_mtr_pools_mng *pools_mng =
+                               &priv->sh->mtrmng->pools_mng;
+       union mlx5_l3t_data data;
+
+       if (priv->sh->meter_aso_en) {
+               rte_spinlock_lock(&pools_mng->mtrsl);
+               if (!pools_mng->n_valid || !priv->mtr_idx_tbl) {
+                       rte_spinlock_unlock(&pools_mng->mtrsl);
+                       return NULL;
+               }
+               if (mlx5_l3t_get_entry(priv->mtr_idx_tbl, meter_id, &data) ||
+                       !data.dword) {
+                       rte_spinlock_unlock(&pools_mng->mtrsl);
+                       return NULL;
+               }
+               if (mtr_idx)
+                       *mtr_idx = data.dword;
+               aso_mtr = mlx5_aso_meter_by_idx(priv, data.dword);
+               /* Remove reference taken by the mlx5_l3t_get_entry. */
+               mlx5_l3t_clear_entry(priv->mtr_idx_tbl, meter_id);
+               rte_spinlock_unlock(&pools_mng->mtrsl);
+               if (!aso_mtr || aso_mtr->state == ASO_METER_FREE)
+                       return NULL;
+               return &aso_mtr->fm;
+       }
+       TAILQ_FOREACH(legacy_fm, fms, next)
+               if (meter_id == legacy_fm->fm.meter_id) {
+                       if (mtr_idx)
+                               *mtr_idx = legacy_fm->idx;
+                       return &legacy_fm->fm;
+               }
+       return NULL;
+}
+
+/**
+ * Find meter by index.
+ *
+ * @param priv
+ *   Pointer to mlx5_priv.
+ * @param idx
+ *   Meter index.
+ *
+ * @return
+ *   Pointer to the meter info found on success, NULL otherwise.
+ */
+struct mlx5_flow_meter_info *
+flow_dv_meter_find_by_idx(struct mlx5_priv *priv, uint32_t idx)
+{
+       struct mlx5_aso_mtr *aso_mtr;
+
+       if (priv->sh->meter_aso_en) {
+               aso_mtr = mlx5_aso_meter_by_idx(priv, idx);
+               if (!aso_mtr)
+                       return NULL;
+               return &aso_mtr->fm;
+       } else {
+               return mlx5_ipool_get(priv->sh->ipool[MLX5_IPOOL_MTR], idx);
+       }
+}
+
+/**
+ * Attach meter to flow.
+ * Unidirectional Meter creation can only be done
+ * when flow direction is known, i.e. when calling meter_attach.
+ *
+ * @param [in] priv
+ *  Pointer to mlx5 private data.
+ * @param[in] fm
+ *   Pointer to flow meter.
+ * @param [in] attr
+ *  Pointer to flow attributes.
+ * @param [out] error
+ *  Pointer to error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+int
+mlx5_flow_meter_attach(struct mlx5_priv *priv,
+                      struct mlx5_flow_meter_info *fm,
+                      const struct rte_flow_attr *attr,
+                      struct rte_flow_error *error)
+{
+       int ret = 0;
+
+       if (priv->sh->meter_aso_en) {
+               struct mlx5_aso_mtr *aso_mtr;
+
+               aso_mtr = container_of(fm, struct mlx5_aso_mtr, fm);
+               if (mlx5_aso_mtr_wait(priv->sh, aso_mtr)) {
+                       return rte_flow_error_set(error, ENOENT,
+                                       RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
+                                       NULL,
+                                       "Timeout in meter configuration");
+               }
+               rte_spinlock_lock(&fm->sl);
+               if (fm->shared || !fm->ref_cnt) {
+                       fm->ref_cnt++;
+               } else {
+                       rte_flow_error_set(error, EINVAL,
+                                  RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+                                  "Meter cannot be shared");
+                       ret = -1;
+               }
+               rte_spinlock_unlock(&fm->sl);
+       } else {
+               rte_spinlock_lock(&fm->sl);
+               if (fm->meter_action) {
+                       if (fm->shared &&
+                           attr->transfer == fm->transfer &&
+                           attr->ingress == fm->ingress &&
+                           attr->egress == fm->egress) {
+                               fm->ref_cnt++;
+                       } else {
+                               rte_flow_error_set(error, EINVAL,
+                                       RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+                                       fm->shared ?
+                                       "Meter attr not match." :
+                                       "Meter cannot be shared.");
+                               ret = -1;
+                       }
+               } else {
+                       fm->ingress = attr->ingress;
+                       fm->egress = attr->egress;
+                       fm->transfer = attr->transfer;
+                       fm->ref_cnt = 1;
+                       /* This also creates the meter object. */
+                       fm->meter_action = mlx5_flow_meter_action_create(priv,
+                                                                        fm);
+                       if (!fm->meter_action) {
+                               fm->ref_cnt = 0;
+                               fm->ingress = 0;
+                               fm->egress = 0;
+                               fm->transfer = 0;
+                               rte_flow_error_set(error, EINVAL,
+                                       RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+                                       "Meter action create failed.");
+                               ret = -1;
+                       }
+               }
+               rte_spinlock_unlock(&fm->sl);
+       }
+       return ret ? -rte_errno : 0;
+}
+
+/**
+ * Detach meter from flow.
+ *
+ * @param [in] priv
+ *  Pointer to mlx5 private data.
+ * @param [in] fm
+ *  Pointer to flow meter.
+ */
+void
+mlx5_flow_meter_detach(struct mlx5_priv *priv,
+                      struct mlx5_flow_meter_info *fm)
+{
+#ifdef HAVE_MLX5_DR_CREATE_ACTION_FLOW_METER
+       rte_spinlock_lock(&fm->sl);
+       MLX5_ASSERT(fm->ref_cnt);
+       if (--fm->ref_cnt == 0 && !priv->sh->meter_aso_en) {
+               mlx5_glue->destroy_flow_action(fm->meter_action);
+               fm->meter_action = NULL;
+               fm->ingress = 0;
+               fm->egress = 0;
+               fm->transfer = 0;
+       }
+       rte_spinlock_unlock(&fm->sl);
+#else
+       (void)priv;
+       (void)fm;
+#endif
+}
+
+/**
+ * Flush meter with Rx queue configuration.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ */
+void
+mlx5_flow_meter_rxq_flush(struct rte_eth_dev *dev)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_flow_meter_sub_policy *sub_policy;
+       struct mlx5_flow_meter_policy *mtr_policy;
+       void *entry;
+       uint32_t i, policy_idx;
+
+       if (!priv->mtr_en)
+               return;
+       if (priv->policy_idx_tbl) {
+               MLX5_L3T_FOREACH(priv->policy_idx_tbl, i, entry) {
+                       policy_idx = *(uint32_t *)entry;
+                       sub_policy = mlx5_ipool_get
+                               (priv->sh->ipool[MLX5_IPOOL_MTR_POLICY],
+                               policy_idx);
+                       if (!sub_policy || !sub_policy->main_policy)
+                               continue;
+                       mtr_policy = sub_policy->main_policy;
+                       if (mtr_policy->is_queue || mtr_policy->is_rss)
+                               mlx5_flow_destroy_sub_policy_with_rxq(dev,
+                                       mtr_policy);
+               }
+       }
+}
+
+/**
+ * Iterate a meter hierarchy and flush all meters and policies if possible.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[in] fm
+ *   Pointer to flow meter.
+ * @param[in] mtr_idx
+ *   .Meter's index
+ * @param[out] error
+ *   Pointer to rte meter error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_flush_hierarchy(struct rte_eth_dev *dev,
+                               struct mlx5_flow_meter_info *fm,
+                               uint32_t mtr_idx,
+                               struct rte_mtr_error *error)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_flow_meter_policy *policy;
+       uint32_t policy_id;
+       struct mlx5_flow_meter_info *next_fm;
+       uint32_t next_mtr_idx;
+       struct mlx5_flow_meter_policy *next_policy = NULL;
+
+       policy = mlx5_flow_meter_policy_find(dev, fm->policy_id, NULL);
+       MLX5_ASSERT(policy);
+       while (!fm->ref_cnt && policy->is_hierarchy) {
+               policy_id = fm->policy_id;
+               next_fm = mlx5_flow_meter_find(priv,
+                               policy->act_cnt[RTE_COLOR_GREEN].next_mtr_id,
+                               &next_mtr_idx);
+               if (next_fm) {
+                       next_policy = mlx5_flow_meter_policy_find(dev,
+                                                       next_fm->policy_id,
+                                                       NULL);
+                       MLX5_ASSERT(next_policy);
+               }
+               if (mlx5_flow_meter_params_flush(dev, fm, mtr_idx))
+                       return -rte_mtr_error_set(error, ENOTSUP,
+                                               RTE_MTR_ERROR_TYPE_MTR_ID,
+                                               NULL,
+                                               "Failed to flush meter.");
+               if (policy->ref_cnt)
+                       break;
+               if (__mlx5_flow_meter_policy_delete(dev, policy_id,
+                                               policy, error, true))
+                       return -rte_errno;
+               mlx5_free(policy);
+               if (!next_fm || !next_policy)
+                       break;
+               fm = next_fm;
+               mtr_idx = next_mtr_idx;
+               policy = next_policy;
+       }
+       return 0;
+}
+
+/**
+ * Flush all the hierarchy meters and their policies.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[out] error
+ *   Pointer to rte meter error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+static int
+mlx5_flow_meter_flush_all_hierarchies(struct rte_eth_dev *dev,
+                                     struct rte_mtr_error *error)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_flow_meter_info *fm;
+       struct mlx5_flow_meter_policy *policy;
+       struct mlx5_flow_meter_sub_policy *sub_policy;
+       struct mlx5_flow_meter_info *next_fm;
+       struct mlx5_aso_mtr *aso_mtr;
+       uint32_t mtr_idx = 0;
+       uint32_t i, policy_idx;
+       void *entry;
+
+       if (!priv->mtr_idx_tbl || !priv->policy_idx_tbl)
+               return 0;
+       MLX5_L3T_FOREACH(priv->mtr_idx_tbl, i, entry) {
+               mtr_idx = *(uint32_t *)entry;
+               if (!mtr_idx)
+                       continue;
+               aso_mtr = mlx5_aso_meter_by_idx(priv, mtr_idx);
+               fm = &aso_mtr->fm;
+               if (fm->ref_cnt || fm->def_policy)
+                       continue;
+               if (mlx5_flow_meter_flush_hierarchy(dev, fm, mtr_idx, error))
+                       return -rte_errno;
+       }
+       MLX5_L3T_FOREACH(priv->policy_idx_tbl, i, entry) {
+               policy_idx = *(uint32_t *)entry;
+               sub_policy = mlx5_ipool_get
+                               (priv->sh->ipool[MLX5_IPOOL_MTR_POLICY],
+                               policy_idx);
+               if (!sub_policy)
+                       return -rte_mtr_error_set(error,
+                                       EINVAL,
+                                       RTE_MTR_ERROR_TYPE_METER_POLICY_ID,
+                                       NULL, "Meter policy invalid.");
+               policy = sub_policy->main_policy;
+               if (!policy || !policy->is_hierarchy || policy->ref_cnt)
+                       continue;
+               next_fm = mlx5_flow_meter_find(priv,
+                               policy->act_cnt[RTE_COLOR_GREEN].next_mtr_id,
+                               &mtr_idx);
+               if (__mlx5_flow_meter_policy_delete(dev, i, policy,
+                                                   error, true))
+                       return -rte_mtr_error_set(error,
+                                       EINVAL,
+                                       RTE_MTR_ERROR_TYPE_METER_POLICY_ID,
+                                       NULL, "Meter policy invalid.");
+               mlx5_free(policy);
+               if (!next_fm || next_fm->ref_cnt || next_fm->def_policy)
+                       continue;
+               if (mlx5_flow_meter_flush_hierarchy(dev, next_fm,
+                                                   mtr_idx, error))
+                       return -rte_errno;
+       }
+       return 0;
+}
+/**
+ * Flush meter configuration.
+ *
+ * @param[in] dev
+ *   Pointer to Ethernet device.
+ * @param[out] error
+ *   Pointer to rte meter error structure.
+ *
+ * @return
+ *   0 on success, a negative errno value otherwise and rte_errno is set.
+ */
+int
+mlx5_flow_meter_flush(struct rte_eth_dev *dev, struct rte_mtr_error *error)
+{
+       struct mlx5_priv *priv = dev->data->dev_private;
+       struct mlx5_legacy_flow_meters *fms = &priv->flow_meters;
+       struct mlx5_flow_meter_profile *fmp;
+       struct mlx5_legacy_flow_meter *legacy_fm;
+       struct mlx5_flow_meter_info *fm;
+       struct mlx5_flow_meter_sub_policy *sub_policy;
+       void *tmp;
+       uint32_t i, mtr_idx, policy_idx;
+       void *entry;
+       struct mlx5_aso_mtr *aso_mtr;
+
+       if (!priv->mtr_en)
+               return 0;
+       if (priv->sh->meter_aso_en) {
+               if (mlx5_flow_meter_flush_all_hierarchies(dev, error))
+                       return -rte_errno;
+               if (priv->mtr_idx_tbl) {
+                       MLX5_L3T_FOREACH(priv->mtr_idx_tbl, i, entry) {
+                               mtr_idx = *(uint32_t *)entry;
+                               if (mtr_idx) {
+                                       aso_mtr =
+                                       mlx5_aso_meter_by_idx(priv, mtr_idx);
+                                       fm = &aso_mtr->fm;
+                                       (void)mlx5_flow_meter_params_flush(dev,
+                                               fm, mtr_idx);
+                               }
+                       }
+                       mlx5_l3t_destroy(priv->mtr_idx_tbl);
+                       priv->mtr_idx_tbl = NULL;
+               }
+       } else {
+               TAILQ_FOREACH_SAFE(legacy_fm, fms, next, tmp) {
+                       fm = &legacy_fm->fm;
+                       if (mlx5_flow_meter_params_flush(dev, fm, 0))
+                               return -rte_mtr_error_set(error, EINVAL,
+                               RTE_MTR_ERROR_TYPE_METER_PROFILE_ID,
+                               NULL, "MTR object meter profile invalid.");
+               }
+       }
+       if (priv->policy_idx_tbl) {
+               MLX5_L3T_FOREACH(priv->policy_idx_tbl, i, entry) {
+                       policy_idx = *(uint32_t *)entry;
+                       sub_policy = mlx5_ipool_get
+                               (priv->sh->ipool[MLX5_IPOOL_MTR_POLICY],
+                               policy_idx);
+                       if (!sub_policy)
+                               return -rte_mtr_error_set(error,
+                                               EINVAL,
+                                       RTE_MTR_ERROR_TYPE_METER_POLICY_ID,
+                                               NULL, "MTR object "
+                                               "meter policy invalid.");
+                       if (__mlx5_flow_meter_policy_delete(dev, i,
+                                               sub_policy->main_policy,
+                                               error, true))
+                               return -rte_mtr_error_set(error,
+                                               EINVAL,
+                                       RTE_MTR_ERROR_TYPE_METER_POLICY_ID,
+                                               NULL, "MTR object "
+                                               "meter policy invalid.");
+                       mlx5_free(sub_policy->main_policy);
+               }
+               mlx5_l3t_destroy(priv->policy_idx_tbl);
+               priv->policy_idx_tbl = NULL;
+       }
+       if (priv->mtr_profile_tbl) {
+               MLX5_L3T_FOREACH(priv->mtr_profile_tbl, i, entry) {
+                       fmp = entry;
+                       if (mlx5_flow_meter_profile_delete(dev, fmp->id,
+                                                          error))
+                               return -rte_mtr_error_set(error, EINVAL,
+                                       RTE_MTR_ERROR_TYPE_METER_POLICY_ID,
+                                               NULL, "Fail to destroy "
+                                               "meter profile.");
+               }
+               mlx5_l3t_destroy(priv->mtr_profile_tbl);
+               priv->mtr_profile_tbl = NULL;
+       }
+       /* Delete default policy table. */
+       mlx5_flow_destroy_def_policy(dev);
+       if (priv->sh->refcnt == 1)
+               mlx5_flow_destroy_mtr_drop_tbls(dev);
        return 0;
 }