nfp: tls: add basic statistics

Count TX TLS packets: successes, out of order, and dropped due to
missing record info.  Make sure the RX and TX completion statistics
don't share cache lines with TX ones as much as possible.  With TLS
stats they are no longer reasonably aligned.
Signed-off-by: Dirk van der Merwe <dirk.vandermerwe@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/netronome/nfp/crypto/tls.c

@@ -324,9 +324,13 @@ nfp_net_tls_add(struct net_device *netdev, struct sock *sk,
 	reply = (void *)skb->data;
 	err = -be32_to_cpu(reply->error);
 	if (err) {
-		if (err != -ENOSPC)
+		if (err == -ENOSPC) {
+			if (!atomic_fetch_inc(&nn->ktls_no_space))
+				nn_info(nn, "HW TLS table full\n");
+		} else {
 			nn_dp_warn(&nn->dp,
 				   "failed to add TLS, FW replied: %d\n", err);
+		}
 		goto err_free_skb;
 	}
 

drivers/net/ethernet/netronome/nfp/nfp_net.h

@@ -12,6 +12,7 @@
 #ifndef _NFP_NET_H_
 #define _NFP_NET_H_
 
+#include <linux/atomic.h>
 #include <linux/interrupt.h>
 #include <linux/list.h>
 #include <linux/netdevice.h>
@@ -373,6 +374,11 @@ struct nfp_net_rx_ring {
  * @hw_csum_tx_inner:	 Counter of inner TX checksum offload requests
  * @tx_gather:	    Counter of packets with Gather DMA
  * @tx_lso:	    Counter of LSO packets sent
+ * @hw_tls_tx:	    Counter of TLS packets sent with crypto offloaded to HW
+ * @tls_tx_fallback:	Counter of TLS packets sent which had to be encrypted
+ *			by the fallback path because packets came out of order
+ * @tls_tx_no_fallback:	Counter of TLS packets not sent because the fallback
+ *			path could not encrypt them
  * @tx_errors:	    How many TX errors were encountered
  * @tx_busy:        How often was TX busy (no space)?
  * @rx_replace_buf_alloc_fail:	Counter of RX buffer allocation failures
@@ -410,21 +416,28 @@ struct nfp_net_r_vector {
 	u64 hw_csum_rx_inner_ok;
 	u64 hw_csum_rx_complete;
 
+	u64 hw_csum_rx_error;
+	u64 rx_replace_buf_alloc_fail;
+
 	struct nfp_net_tx_ring *xdp_ring;
 
 	struct u64_stats_sync tx_sync;
 	u64 tx_pkts;
 	u64 tx_bytes;
-	u64 hw_csum_tx;
+
+	u64 ____cacheline_aligned_in_smp hw_csum_tx;
 	u64 hw_csum_tx_inner;
 	u64 tx_gather;
 	u64 tx_lso;
+	u64 hw_tls_tx;
 
-	u64 hw_csum_rx_error;
-	u64 rx_replace_buf_alloc_fail;
+	u64 tls_tx_fallback;
+	u64 tls_tx_no_fallback;
 	u64 tx_errors;
 	u64 tx_busy;
 
+	/* Cold data follows */
+
 	u32 irq_vector;
 	irq_handler_t handler;
 	char name[IFNAMSIZ + 8];
@@ -566,6 +579,8 @@ struct nfp_net_dp {
  * @rx_bar:             Pointer to mapped FL/RX queues
  * @tlv_caps:		Parsed TLV capabilities
  * @ktls_tx_conn_cnt:	Number of offloaded kTLS TX connections
+ * @ktls_no_space:	Counter of firmware rejecting kTLS connection due to
+ *			lack of space
  * @mbox_cmsg:		Common Control Message via vNIC mailbox state
  * @mbox_cmsg.queue:	CCM mbox queue of pending messages
  * @mbox_cmsg.wq:	CCM mbox wait queue of waiting processes
@@ -647,6 +662,8 @@ struct nfp_net {
 
 	unsigned int ktls_tx_conn_cnt;
 
+	atomic_t ktls_no_space;
+
 	struct {
 		struct sk_buff_head queue;
 		wait_queue_head_t wq;

drivers/net/ethernet/netronome/nfp/nfp_net_common.c

@@ -804,8 +804,8 @@ static void nfp_net_tx_csum(struct nfp_net_dp *dp,
 
 #ifdef CONFIG_TLS_DEVICE
 static struct sk_buff *
-nfp_net_tls_tx(struct nfp_net_dp *dp, struct sk_buff *skb, u64 *tls_handle,
-	       int *nr_frags)
+nfp_net_tls_tx(struct nfp_net_dp *dp, struct nfp_net_r_vector *r_vec,
+	       struct sk_buff *skb, u64 *tls_handle, int *nr_frags)
 {
 	struct nfp_net_tls_offload_ctx *ntls;
 	struct sk_buff *nskb;
@@ -824,15 +824,26 @@ nfp_net_tls_tx(struct nfp_net_dp *dp, struct sk_buff *skb, u64 *tls_handle,
 		if (!datalen)
 			return skb;
 
+		u64_stats_update_begin(&r_vec->tx_sync);
+		r_vec->tls_tx_fallback++;
+		u64_stats_update_end(&r_vec->tx_sync);
+
 		nskb = tls_encrypt_skb(skb);
-		if (!nskb)
+		if (!nskb) {
+			u64_stats_update_begin(&r_vec->tx_sync);
+			r_vec->tls_tx_no_fallback++;
+			u64_stats_update_end(&r_vec->tx_sync);
 			return NULL;
+		}
 		/* encryption wasn't necessary */
 		if (nskb == skb)
 			return skb;
 		/* we don't re-check ring space */
 		if (unlikely(skb_is_nonlinear(nskb))) {
 			nn_dp_warn(dp, "tls_encrypt_skb() produced fragmented frame\n");
+			u64_stats_update_begin(&r_vec->tx_sync);
+			r_vec->tx_errors++;
+			u64_stats_update_end(&r_vec->tx_sync);
 			dev_kfree_skb_any(nskb);
 			return NULL;
 		}
@@ -845,6 +856,12 @@ nfp_net_tls_tx(struct nfp_net_dp *dp, struct sk_buff *skb, u64 *tls_handle,
 		return nskb;
 	}
 
+	if (datalen) {
+		u64_stats_update_begin(&r_vec->tx_sync);
+		r_vec->hw_tls_tx++;
+		u64_stats_update_end(&r_vec->tx_sync);
+	}
+
 	memcpy(tls_handle, ntls->fw_handle, sizeof(ntls->fw_handle));
 	ntls->next_seq += datalen;
 	return skb;
@@ -944,9 +961,11 @@ static int nfp_net_tx(struct sk_buff *skb, struct net_device *netdev)
 	}
 
 #ifdef CONFIG_TLS_DEVICE
-	skb = nfp_net_tls_tx(dp, skb, &tls_handle, &nr_frags);
-	if (unlikely(!skb))
-		goto err_flush;
+	skb = nfp_net_tls_tx(dp, r_vec, skb, &tls_handle, &nr_frags);
+	if (unlikely(!skb)) {
+		nfp_net_tx_xmit_more_flush(tx_ring);
+		return NETDEV_TX_OK;
+	}
 #endif
 
 	md_bytes = nfp_net_prep_tx_meta(skb, tls_handle);

drivers/net/ethernet/netronome/nfp/nfp_net_ethtool.c

@@ -150,8 +150,9 @@ static const struct nfp_et_stat nfp_mac_et_stats[] = {
 
 #define NN_ET_GLOBAL_STATS_LEN ARRAY_SIZE(nfp_net_et_stats)
 #define NN_ET_SWITCH_STATS_LEN 9
-#define NN_RVEC_GATHER_STATS	9
+#define NN_RVEC_GATHER_STATS	12
 #define NN_RVEC_PER_Q_STATS	3
+#define NN_CTRL_PATH_STATS	1
 
 #define SFP_SFF_REV_COMPLIANCE	1
 
@@ -423,7 +424,8 @@ static unsigned int nfp_vnic_get_sw_stats_count(struct net_device *netdev)
 {
 	struct nfp_net *nn = netdev_priv(netdev);
 
-	return NN_RVEC_GATHER_STATS + nn->max_r_vecs * NN_RVEC_PER_Q_STATS;
+	return NN_RVEC_GATHER_STATS + nn->max_r_vecs * NN_RVEC_PER_Q_STATS +
+		NN_CTRL_PATH_STATS;
 }
 
 static u8 *nfp_vnic_get_sw_stats_strings(struct net_device *netdev, u8 *data)
@@ -446,6 +448,11 @@ static u8 *nfp_vnic_get_sw_stats_strings(struct net_device *netdev, u8 *data)
 	data = nfp_pr_et(data, "hw_tx_inner_csum");
 	data = nfp_pr_et(data, "tx_gather");
 	data = nfp_pr_et(data, "tx_lso");
+	data = nfp_pr_et(data, "tx_tls_encrypted");
+	data = nfp_pr_et(data, "tx_tls_ooo");
+	data = nfp_pr_et(data, "tx_tls_drop_no_sync_data");
+
+	data = nfp_pr_et(data, "hw_tls_no_space");
 
 	return data;
 }
@@ -478,6 +485,9 @@ static u64 *nfp_vnic_get_sw_stats(struct net_device *netdev, u64 *data)
 			tmp[6] = nn->r_vecs[i].hw_csum_tx_inner;
 			tmp[7] = nn->r_vecs[i].tx_gather;
 			tmp[8] = nn->r_vecs[i].tx_lso;
+			tmp[9] = nn->r_vecs[i].hw_tls_tx;
+			tmp[10] = nn->r_vecs[i].tls_tx_fallback;
+			tmp[11] = nn->r_vecs[i].tls_tx_no_fallback;
 		} while (u64_stats_fetch_retry(&nn->r_vecs[i].tx_sync, start));
 
 		data += NN_RVEC_PER_Q_STATS;
@@ -489,6 +499,8 @@ static u64 *nfp_vnic_get_sw_stats(struct net_device *netdev, u64 *data)
 	for (j = 0; j < NN_RVEC_GATHER_STATS; j++)
 		*data++ = gathered_stats[j];
 
+	*data++ = atomic_read(&nn->ktls_no_space);
+
 	return data;
 }