octeontx2-pf: add tc flower stats handler for hw offloads

Add support to get the stats for tc flower flows that are
offloaded to hardware. To support this feature, added a
new AF mbox handler which returns the MCAM entry stats
for a flow that has hardware stat counter enabled.

Signed-off-by: Naveen Mamindlapalli <naveenm@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/marvell/octeontx2/af/mbox.h

@@ -216,6 +216,9 @@ M(NPC_MCAM_READ_ENTRY,	  0x600f, npc_mcam_read_entry,			\
 				  npc_mcam_read_entry_rsp)		\
 M(NPC_MCAM_READ_BASE_RULE, 0x6011, npc_read_base_steer_rule,            \
 				   msg_req, npc_mcam_read_base_rule_rsp)  \
+M(NPC_MCAM_GET_STATS, 0x6012, npc_mcam_entry_stats,                     \
+				   npc_mcam_get_stats_req,              \
+				   npc_mcam_get_stats_rsp)              \
 /* NIX mbox IDs (range 0x8000 - 0xFFFF) */				\
 M(NIX_LF_ALLOC,		0x8000, nix_lf_alloc,				\
 				 nix_lf_alloc_req, nix_lf_alloc_rsp)	\
@@ -1195,6 +1198,17 @@ struct npc_mcam_read_base_rule_rsp {
 	struct mcam_entry entry;
 };
 
+struct npc_mcam_get_stats_req {
+	struct mbox_msghdr hdr;
+	u16 entry; /* mcam entry */
+};
+
+struct npc_mcam_get_stats_rsp {
+	struct mbox_msghdr hdr;
+	u64 stat;  /* counter stats */
+	u8 stat_ena; /* enabled */
+};
+
 enum ptp_op {
 	PTP_OP_ADJFINE = 0,
 	PTP_OP_GET_CLOCK = 1,

drivers/net/ethernet/marvell/octeontx2/af/rvu_npc.c

@@ -2806,3 +2806,42 @@ int rvu_mbox_handler_npc_read_base_steer_rule(struct rvu *rvu,
 out:
 	return rc;
 }
+
+int rvu_mbox_handler_npc_mcam_entry_stats(struct rvu *rvu,
+					  struct npc_mcam_get_stats_req *req,
+					  struct npc_mcam_get_stats_rsp *rsp)
+{
+	struct npc_mcam *mcam = &rvu->hw->mcam;
+	u16 index, cntr;
+	int blkaddr;
+	u64 regval;
+	u32 bank;
+
+	blkaddr = rvu_get_blkaddr(rvu, BLKTYPE_NPC, 0);
+	if (blkaddr < 0)
+		return NPC_MCAM_INVALID_REQ;
+
+	mutex_lock(&mcam->lock);
+
+	index = req->entry & (mcam->banksize - 1);
+	bank = npc_get_bank(mcam, req->entry);
+
+	/* read MCAM entry STAT_ACT register */
+	regval = rvu_read64(rvu, blkaddr, NPC_AF_MCAMEX_BANKX_STAT_ACT(index, bank));
+
+	if (!(regval & BIT_ULL(9))) {
+		rsp->stat_ena = 0;
+		mutex_unlock(&mcam->lock);
+		return 0;
+	}
+
+	cntr = regval & 0x1FF;
+
+	rsp->stat_ena = 1;
+	rsp->stat = rvu_read64(rvu, blkaddr, NPC_AF_MATCH_STATX(cntr));
+	rsp->stat &= BIT_ULL(48) - 1;
+
+	mutex_unlock(&mcam->lock);
+
+	return 0;
+}

drivers/net/ethernet/marvell/octeontx2/nic/otx2_tc.c

@@ -16,12 +16,20 @@
 
 #include "otx2_common.h"
 
+struct otx2_tc_flow_stats {
+	u64 bytes;
+	u64 pkts;
+	u64 used;
+};
+
 struct otx2_tc_flow {
 	struct rhash_head		node;
 	unsigned long			cookie;
 	u16				entry;
 	unsigned int			bitpos;
 	struct rcu_head			rcu;
+	struct otx2_tc_flow_stats	stats;
+	spinlock_t			lock; /* lock for stats */
 };
 
 static int otx2_tc_parse_actions(struct otx2_nic *nic,
@@ -403,6 +411,66 @@ static int otx2_tc_add_flow(struct otx2_nic *nic,
 	return rc;
 }
 
+static int otx2_tc_get_flow_stats(struct otx2_nic *nic,
+				  struct flow_cls_offload *tc_flow_cmd)
+{
+	struct otx2_tc_info *tc_info = &nic->tc_info;
+	struct npc_mcam_get_stats_req *req;
+	struct npc_mcam_get_stats_rsp *rsp;
+	struct otx2_tc_flow_stats *stats;
+	struct otx2_tc_flow *flow_node;
+	int err;
+
+	flow_node = rhashtable_lookup_fast(&tc_info->flow_table,
+					   &tc_flow_cmd->cookie,
+					   tc_info->flow_ht_params);
+	if (!flow_node) {
+		netdev_info(nic->netdev, "tc flow not found for cookie %lx",
+			    tc_flow_cmd->cookie);
+		return -EINVAL;
+	}
+
+	mutex_lock(&nic->mbox.lock);
+
+	req = otx2_mbox_alloc_msg_npc_mcam_entry_stats(&nic->mbox);
+	if (!req) {
+		mutex_unlock(&nic->mbox.lock);
+		return -ENOMEM;
+	}
+
+	req->entry = flow_node->entry;
+
+	err = otx2_sync_mbox_msg(&nic->mbox);
+	if (err) {
+		netdev_err(nic->netdev, "Failed to get stats for MCAM flow entry %d\n",
+			   req->entry);
+		mutex_unlock(&nic->mbox.lock);
+		return -EFAULT;
+	}
+
+	rsp = (struct npc_mcam_get_stats_rsp *)otx2_mbox_get_rsp
+		(&nic->mbox.mbox, 0, &req->hdr);
+	if (IS_ERR(rsp)) {
+		mutex_unlock(&nic->mbox.lock);
+		return PTR_ERR(rsp);
+	}
+
+	mutex_unlock(&nic->mbox.lock);
+
+	if (!rsp->stat_ena)
+		return -EINVAL;
+
+	stats = &flow_node->stats;
+
+	spin_lock(&flow_node->lock);
+	flow_stats_update(&tc_flow_cmd->stats, 0x0, rsp->stat - stats->pkts, 0x0, 0x0,
+			  FLOW_ACTION_HW_STATS_IMMEDIATE);
+	stats->pkts = rsp->stat;
+	spin_unlock(&flow_node->lock);
+
+	return 0;
+}
+
 static int otx2_setup_tc_cls_flower(struct otx2_nic *nic,
 				    struct flow_cls_offload *cls_flower)
 {
@@ -412,7 +480,7 @@ static int otx2_setup_tc_cls_flower(struct otx2_nic *nic,
 	case FLOW_CLS_DESTROY:
 		return otx2_tc_del_flow(nic, cls_flower);
 	case FLOW_CLS_STATS:
-		return -EOPNOTSUPP;
+		return otx2_tc_get_flow_stats(nic, cls_flower);
 	default:
 		return -EOPNOTSUPP;
 	}