net/octeontx2: support flow destroy ops
authorKiran Kumar K <kirankumark@marvell.com>
Wed, 26 Jun 2019 10:42:14 +0000 (16:12 +0530)
committerFerruh Yigit <ferruh.yigit@intel.com>
Thu, 4 Jul 2019 23:52:02 +0000 (01:52 +0200)
Adding few more flow operations like flow_destroy, flow_isolate
and flow_flush.

Signed-off-by: Kiran Kumar K <kirankumark@marvell.com>
Signed-off-by: Vivek Sharma <viveksharma@marvell.com>
drivers/net/octeontx2/otx2_flow.c
drivers/net/octeontx2/otx2_rx.h

index 896aef0..24bde62 100644 (file)
@@ -5,6 +5,48 @@
 #include "otx2_ethdev.h"
 #include "otx2_flow.h"
 
+int
+otx2_flow_free_all_resources(struct otx2_eth_dev *hw)
+{
+       struct otx2_npc_flow_info *npc = &hw->npc_flow;
+       struct otx2_mbox *mbox = hw->mbox;
+       struct otx2_mcam_ents_info *info;
+       struct rte_bitmap *bmap;
+       struct rte_flow *flow;
+       int entry_count = 0;
+       int rc, idx;
+
+       for (idx = 0; idx < npc->flow_max_priority; idx++) {
+               info = &npc->flow_entry_info[idx];
+               entry_count += info->live_ent;
+       }
+
+       if (entry_count == 0)
+               return 0;
+
+       /* Free all MCAM entries allocated */
+       rc = otx2_flow_mcam_free_all_entries(mbox);
+
+       /* Free any MCAM counters and delete flow list */
+       for (idx = 0; idx < npc->flow_max_priority; idx++) {
+               while ((flow = TAILQ_FIRST(&npc->flow_list[idx])) != NULL) {
+                       if (flow->ctr_id != NPC_COUNTER_NONE)
+                               rc |= otx2_flow_mcam_free_counter(mbox,
+                                                            flow->ctr_id);
+
+                       TAILQ_REMOVE(&npc->flow_list[idx], flow, next);
+                       rte_free(flow);
+                       bmap = npc->live_entries[flow->priority];
+                       rte_bitmap_clear(bmap, flow->mcam_id);
+               }
+               info = &npc->flow_entry_info[idx];
+               info->free_ent = 0;
+               info->live_ent = 0;
+       }
+       return rc;
+}
+
+
 static int
 flow_program_npc(struct otx2_parse_state *pst, struct otx2_mbox *mbox,
                 struct otx2_npc_flow_info *flow_info)
@@ -237,6 +279,27 @@ flow_program_rss_action(struct rte_eth_dev *eth_dev,
        return 0;
 }
 
+static int
+flow_free_rss_action(struct rte_eth_dev *eth_dev,
+                    struct rte_flow *flow)
+{
+       struct otx2_eth_dev *dev = eth_dev->data->dev_private;
+       struct otx2_npc_flow_info *npc = &dev->npc_flow;
+       uint32_t rss_grp;
+
+       if (flow->npc_action & NIX_RX_ACTIONOP_RSS) {
+               rss_grp = (flow->npc_action >> NIX_RSS_ACT_GRP_OFFSET) &
+                       NIX_RSS_ACT_GRP_MASK;
+               if (rss_grp == 0 || rss_grp >= npc->rss_grps)
+                       return -EINVAL;
+
+               rte_bitmap_clear(npc->rss_grp_entries, rss_grp);
+       }
+
+       return 0;
+}
+
+
 static int
 flow_parse_meta_items(__rte_unused struct otx2_parse_state *pst)
 {
@@ -445,7 +508,150 @@ err_exit:
        return NULL;
 }
 
+static int
+otx2_flow_destroy(struct rte_eth_dev *dev,
+                 struct rte_flow *flow,
+                 struct rte_flow_error *error)
+{
+       struct otx2_eth_dev *hw = dev->data->dev_private;
+       struct otx2_npc_flow_info *npc = &hw->npc_flow;
+       struct otx2_mbox *mbox = hw->mbox;
+       struct rte_bitmap *bmap;
+       uint16_t match_id;
+       int rc;
+
+       match_id = (flow->npc_action >> NIX_RX_ACT_MATCH_OFFSET) &
+               NIX_RX_ACT_MATCH_MASK;
+
+       if (match_id && match_id < OTX2_FLOW_ACTION_FLAG_DEFAULT) {
+               if (rte_atomic32_read(&npc->mark_actions) == 0)
+                       return -EINVAL;
+
+               /* Clear mark offload flag if there are no more mark actions */
+               if (rte_atomic32_sub_return(&npc->mark_actions, 1) == 0)
+                       hw->rx_offload_flags &= ~NIX_RX_OFFLOAD_MARK_UPDATE_F;
+       }
+
+       rc = flow_free_rss_action(dev, flow);
+       if (rc != 0) {
+               rte_flow_error_set(error, EIO,
+                                  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
+                                  NULL,
+                                  "Failed to free rss action");
+       }
+
+       rc = otx2_flow_mcam_free_entry(mbox, flow->mcam_id);
+       if (rc != 0) {
+               rte_flow_error_set(error, EIO,
+                                  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
+                                  NULL,
+                                  "Failed to destroy filter");
+       }
+
+       TAILQ_REMOVE(&npc->flow_list[flow->priority], flow, next);
+
+       bmap = npc->live_entries[flow->priority];
+       rte_bitmap_clear(bmap, flow->mcam_id);
+
+       rte_free(flow);
+       return 0;
+}
+
+static int
+otx2_flow_flush(struct rte_eth_dev *dev,
+               struct rte_flow_error *error)
+{
+       struct otx2_eth_dev *hw = dev->data->dev_private;
+       int rc;
+
+       rc = otx2_flow_free_all_resources(hw);
+       if (rc) {
+               otx2_err("Error when deleting NPC MCAM entries "
+                               ", counters");
+               rte_flow_error_set(error, EIO,
+                                  RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
+                                  NULL,
+                                  "Failed to flush filter");
+               return -rte_errno;
+       }
+
+       return 0;
+}
+
+static int
+otx2_flow_isolate(struct rte_eth_dev *dev __rte_unused,
+                 int enable __rte_unused,
+                 struct rte_flow_error *error)
+{
+       /*
+        * If we support, we need to un-install the default mcam
+        * entry for this port.
+        */
+
+       rte_flow_error_set(error, ENOTSUP,
+                          RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
+                          NULL,
+                          "Flow isolation not supported");
+
+       return -rte_errno;
+}
+
+static int
+otx2_flow_query(struct rte_eth_dev *dev,
+               struct rte_flow *flow,
+               const struct rte_flow_action *action,
+               void *data,
+               struct rte_flow_error *error)
+{
+       struct otx2_eth_dev *hw = dev->data->dev_private;
+       struct rte_flow_query_count *query = data;
+       struct otx2_mbox *mbox = hw->mbox;
+       const char *errmsg = NULL;
+       int errcode = ENOTSUP;
+       int rc;
+
+       if (action->type != RTE_FLOW_ACTION_TYPE_COUNT) {
+               errmsg = "Only COUNT is supported in query";
+               goto err_exit;
+       }
+
+       if (flow->ctr_id == NPC_COUNTER_NONE) {
+               errmsg = "Counter is not available";
+               goto err_exit;
+       }
+
+       rc = otx2_flow_mcam_read_counter(mbox, flow->ctr_id, &query->hits);
+       if (rc != 0) {
+               errcode = EIO;
+               errmsg = "Error reading flow counter";
+               goto err_exit;
+       }
+       query->hits_set = 1;
+       query->bytes_set = 0;
+
+       if (query->reset)
+               rc = otx2_flow_mcam_clear_counter(mbox, flow->ctr_id);
+       if (rc != 0) {
+               errcode = EIO;
+               errmsg = "Error clearing flow counter";
+               goto err_exit;
+       }
+
+       return 0;
+
+err_exit:
+       rte_flow_error_set(error, errcode,
+                          RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
+                          NULL,
+                          errmsg);
+       return -rte_errno;
+}
+
 const struct rte_flow_ops otx2_flow_ops = {
        .validate = otx2_flow_validate,
        .create = otx2_flow_create,
+       .destroy = otx2_flow_destroy,
+       .flush = otx2_flow_flush,
+       .query = otx2_flow_query,
+       .isolate = otx2_flow_isolate,
 };
index db79451..e18e046 100644 (file)
@@ -5,6 +5,9 @@
 #ifndef __OTX2_RX_H__
 #define __OTX2_RX_H__
 
+/* Default mark value used when none is provided. */
+#define OTX2_FLOW_ACTION_FLAG_DEFAULT  0xffff
+
 #define PTYPE_WIDTH 12
 #define PTYPE_NON_TUNNEL_ARRAY_SZ      BIT(PTYPE_WIDTH)
 #define PTYPE_TUNNEL_ARRAY_SZ          BIT(PTYPE_WIDTH)