net/ice: support flow director counter
authorYahui Cao <yahui.cao@intel.com>
Fri, 18 Oct 2019 11:15:59 +0000 (19:15 +0800)
committerFerruh Yigit <ferruh.yigit@intel.com>
Wed, 23 Oct 2019 14:43:10 +0000 (16:43 +0200)
This patch add FDIR statistical counter support
and it includes RTE_FLOW count actions support
and query support.

RTE_FLOW count actions support id and shared.
RTE_FLOW query record packet hits by default.

Signed-off-by: Yahui Cao <yahui.cao@intel.com>
Acked-by: Qi Zhang <qi.z.zhang@intel.com>
Reviewed-by: Xiaolong Ye <xiaolong.ye@intel.com>
drivers/net/ice/ice_ethdev.h
drivers/net/ice/ice_fdir_filter.c

index 5947a78..5129b3b 100644 (file)
@@ -249,6 +249,10 @@ TAILQ_HEAD(ice_parser_list, ice_flow_parser_node);
 
 struct ice_fdir_filter_conf {
        struct ice_fdir_fltr input;
+
+       struct ice_fdir_counter *counter; /* flow specific counter context */
+       struct rte_flow_action_count act_count;
+
        uint64_t input_set;
 };
 
@@ -257,8 +261,11 @@ struct ice_fdir_filter_conf {
 #define ICE_FDIR_COUNTERS_PER_BLOCK            256
 #define ICE_FDIR_COUNTER_INDEX(base_idx) \
                                ((base_idx) * ICE_FDIR_COUNTERS_PER_BLOCK)
+struct ice_fdir_counter_pool;
+
 struct ice_fdir_counter {
        TAILQ_ENTRY(ice_fdir_counter) next;
+       struct ice_fdir_counter_pool *pool;
        uint8_t shared;
        uint32_t ref_cnt;
        uint32_t id;
index 8dbd6fc..b327ead 100644 (file)
@@ -183,6 +183,95 @@ ice_fdir_counter_release(struct ice_pf *pf)
        return 0;
 }
 
+static struct ice_fdir_counter *
+ice_fdir_counter_shared_search(struct ice_fdir_counter_pool_container
+                                       *container,
+                              uint32_t id)
+{
+       struct ice_fdir_counter_pool *pool;
+       struct ice_fdir_counter *counter;
+       int i;
+
+       TAILQ_FOREACH(pool, &container->pool_list, next) {
+               for (i = 0; i < ICE_FDIR_COUNTERS_PER_BLOCK; i++) {
+                       counter = &pool->counters[i];
+
+                       if (counter->shared &&
+                           counter->ref_cnt &&
+                           counter->id == id)
+                               return counter;
+               }
+       }
+
+       return NULL;
+}
+
+static struct ice_fdir_counter *
+ice_fdir_counter_alloc(struct ice_pf *pf, uint32_t shared, uint32_t id)
+{
+       struct ice_hw *hw = ICE_PF_TO_HW(pf);
+       struct ice_fdir_info *fdir_info = &pf->fdir;
+       struct ice_fdir_counter_pool_container *container =
+                               &fdir_info->counter;
+       struct ice_fdir_counter_pool *pool = NULL;
+       struct ice_fdir_counter *counter_free = NULL;
+
+       if (shared) {
+               counter_free = ice_fdir_counter_shared_search(container, id);
+               if (counter_free) {
+                       if (counter_free->ref_cnt + 1 == 0) {
+                               rte_errno = E2BIG;
+                               return NULL;
+                       }
+                       counter_free->ref_cnt++;
+                       return counter_free;
+               }
+       }
+
+       TAILQ_FOREACH(pool, &container->pool_list, next) {
+               counter_free = TAILQ_FIRST(&pool->counter_list);
+               if (counter_free)
+                       break;
+               counter_free = NULL;
+       }
+
+       if (!counter_free) {
+               PMD_DRV_LOG(ERR, "No free counter found\n");
+               return NULL;
+       }
+
+       counter_free->shared = shared;
+       counter_free->id = id;
+       counter_free->ref_cnt = 1;
+       counter_free->pool = pool;
+
+       /* reset statistic counter value */
+       ICE_WRITE_REG(hw, GLSTAT_FD_CNT0H(counter_free->hw_index), 0);
+       ICE_WRITE_REG(hw, GLSTAT_FD_CNT0L(counter_free->hw_index), 0);
+
+       TAILQ_REMOVE(&pool->counter_list, counter_free, next);
+       if (TAILQ_EMPTY(&pool->counter_list)) {
+               TAILQ_REMOVE(&container->pool_list, pool, next);
+               TAILQ_INSERT_TAIL(&container->pool_list, pool, next);
+       }
+
+       return counter_free;
+}
+
+static void
+ice_fdir_counter_free(__rte_unused struct ice_pf *pf,
+                     struct ice_fdir_counter *counter)
+{
+       if (!counter)
+               return;
+
+       if (--counter->ref_cnt == 0) {
+               struct ice_fdir_counter_pool *pool = counter->pool;
+
+               TAILQ_INSERT_TAIL(&pool->counter_list, counter, next);
+       }
+}
+
 /*
  * ice_fdir_setup - reserve and initialize the Flow Director resources
  * @pf: board private structure
@@ -691,12 +780,28 @@ ice_fdir_create_filter(struct ice_adapter *ad,
                goto free_entry;
        }
 
+       /* alloc counter for FDIR */
+       if (filter->input.cnt_ena) {
+               struct rte_flow_action_count *act_count = &filter->act_count;
+
+               filter->counter = ice_fdir_counter_alloc(pf,
+                                                        act_count->shared,
+                                                        act_count->id);
+               if (!filter->counter) {
+                       rte_flow_error_set(error, EINVAL,
+                                       RTE_FLOW_ERROR_TYPE_ACTION, NULL,
+                                       "Failed to alloc FDIR counter.");
+                       goto free_entry;
+               }
+               filter->input.cnt_index = filter->counter->hw_index;
+       }
+
        ret = ice_fdir_add_del_filter(pf, filter, true);
        if (ret) {
                rte_flow_error_set(error, -ret,
                                   RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
                                   "Add filter rule failed.");
-               goto free_entry;
+               goto free_counter;
        }
 
        rte_memcpy(rule, filter, sizeof(*rule));
@@ -704,6 +809,12 @@ ice_fdir_create_filter(struct ice_adapter *ad,
        ice_fdir_cnt_update(pf, filter->input.flow_type, false, true);
        return 0;
 
+free_counter:
+       if (filter->counter) {
+               ice_fdir_counter_free(pf, filter->counter);
+               filter->counter = NULL;
+       }
+
 free_entry:
        rte_free(rule);
        return -rte_errno;
@@ -720,6 +831,11 @@ ice_fdir_destroy_filter(struct ice_adapter *ad,
 
        filter = (struct ice_fdir_filter_conf *)flow->rule;
 
+       if (filter->counter) {
+               ice_fdir_counter_free(pf, filter->counter);
+               filter->counter = NULL;
+       }
+
        ret = ice_fdir_add_del_filter(pf, filter, false);
        if (ret) {
                rte_flow_error_set(error, -ret,
@@ -736,11 +852,54 @@ ice_fdir_destroy_filter(struct ice_adapter *ad,
        return 0;
 }
 
+static int
+ice_fdir_query_count(struct ice_adapter *ad,
+                     struct rte_flow *flow,
+                     struct rte_flow_query_count *flow_stats,
+                     struct rte_flow_error *error)
+{
+       struct ice_pf *pf = &ad->pf;
+       struct ice_hw *hw = ICE_PF_TO_HW(pf);
+       struct ice_fdir_filter_conf *filter = flow->rule;
+       struct ice_fdir_counter *counter = filter->counter;
+       uint64_t hits_lo, hits_hi;
+
+       if (!counter) {
+               rte_flow_error_set(error, EINVAL,
+                                 RTE_FLOW_ERROR_TYPE_ACTION,
+                                 NULL,
+                                 "FDIR counters not available");
+               return -rte_errno;
+       }
+
+       /*
+        * Reading the low 32-bits latches the high 32-bits into a shadow
+        * register. Reading the high 32-bit returns the value in the
+        * shadow register.
+        */
+       hits_lo = ICE_READ_REG(hw, GLSTAT_FD_CNT0L(counter->hw_index));
+       hits_hi = ICE_READ_REG(hw, GLSTAT_FD_CNT0H(counter->hw_index));
+
+       flow_stats->hits_set = 1;
+       flow_stats->hits = hits_lo | (hits_hi << 32);
+       flow_stats->bytes_set = 0;
+       flow_stats->bytes = 0;
+
+       if (flow_stats->reset) {
+               /* reset statistic counter value */
+               ICE_WRITE_REG(hw, GLSTAT_FD_CNT0H(counter->hw_index), 0);
+               ICE_WRITE_REG(hw, GLSTAT_FD_CNT0L(counter->hw_index), 0);
+       }
+
+       return 0;
+}
+
 static struct ice_flow_engine ice_fdir_engine = {
        .init = ice_fdir_init,
        .uninit = ice_fdir_uninit,
        .create = ice_fdir_create_filter,
        .destroy = ice_fdir_destroy_filter,
+       .query_count = ice_fdir_query_count,
        .type = ICE_FLOW_ENGINE_FDIR,
 };
 
@@ -810,8 +969,10 @@ ice_fdir_parse_action(struct ice_adapter *ad,
        struct ice_pf *pf = &ad->pf;
        const struct rte_flow_action_queue *act_q;
        const struct rte_flow_action_mark *mark_spec = NULL;
+       const struct rte_flow_action_count *act_count;
        uint32_t dest_num = 0;
        uint32_t mark_num = 0;
+       uint32_t counter_num = 0;
        int ret;
 
        for (; actions->type != RTE_FLOW_ACTION_TYPE_END; actions++) {
@@ -860,6 +1021,15 @@ ice_fdir_parse_action(struct ice_adapter *ad,
 
                        mark_spec = actions->conf;
                        filter->input.fltr_id = mark_spec->id;
+                       break;
+               case RTE_FLOW_ACTION_TYPE_COUNT:
+                       counter_num++;
+
+                       act_count = actions->conf;
+                       filter->input.cnt_ena = ICE_FXD_FLTR_QW0_STAT_ENA_PKTS;
+                       rte_memcpy(&filter->act_count, act_count,
+                                               sizeof(filter->act_count));
+
                        break;
                default:
                        rte_flow_error_set(error, EINVAL,
@@ -883,6 +1053,13 @@ ice_fdir_parse_action(struct ice_adapter *ad,
                return -rte_errno;
        }
 
+       if (counter_num >= 2) {
+               rte_flow_error_set(error, EINVAL,
+                          RTE_FLOW_ERROR_TYPE_ACTION, actions,
+                          "Too many count actions");
+               return -rte_errno;
+       }
+
        return 0;
 }