eth: fbnic: Collect packet statistics for XDP

Add support for XDP statistics collection and reporting via rtnl_link
and netdev_queue API.

For XDP programs without frags support, fbnic requires MTU to be less
than the HDS threshold. If an over-sized frame is received, the frame
is dropped and recorded as rx_length_errors reported via ip stats to
highlight that this is an error.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Mohsin Bashir <mohsin.bashr@gmail.com>
Link: https://patch.msgid.link/20250813221319.3367670-9-mohsin.bashr@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>

Documentation/networking/device_drivers/ethernet/meta/fbnic.rst

@@ -160,3 +160,14 @@ behavior and potential performance bottlenecks.
 	  credit exhaustion
         - ``pcie_ob_rd_no_np_cred``: Read requests dropped due to non-posted
 	  credit exhaustion
+
+XDP Length Error:
+~~~~~~~~~~~~~~~~~
+
+For XDP programs without frags support, fbnic tries to make sure that MTU fits
+into a single buffer. If an oversized frame is received and gets fragmented,
+it is dropped and the following netlink counters are updated
+
+   - ``rx-length``: number of frames dropped due to lack of fragmentation
+     support in the attached XDP program
+   - ``rx-errors``: total number of packets with errors received on the interface

drivers/net/ethernet/meta/fbnic/fbnic_netdev.c

@@ -407,11 +407,12 @@ static void fbnic_get_stats64(struct net_device *dev,
 			      struct rtnl_link_stats64 *stats64)
 {
 	u64 rx_bytes, rx_packets, rx_dropped = 0, rx_errors = 0;
+	u64 rx_over = 0, rx_missed = 0, rx_length = 0;
 	u64 tx_bytes, tx_packets, tx_dropped = 0;
 	struct fbnic_net *fbn = netdev_priv(dev);
 	struct fbnic_dev *fbd = fbn->fbd;
 	struct fbnic_queue_stats *stats;
-	u64 rx_over = 0, rx_missed = 0;
+
 	unsigned int start, i;
 
 	fbnic_get_hw_stats(fbd);
@@ -489,6 +490,7 @@ static void fbnic_get_stats64(struct net_device *dev,
 	stats64->rx_missed_errors = rx_missed;
 
 	for (i = 0; i < fbn->num_rx_queues; i++) {
+		struct fbnic_ring *xdpr = fbn->tx[FBNIC_MAX_TXQS + i];
 		struct fbnic_ring *rxr = fbn->rx[i];
 
 		if (!rxr)
@@ -500,11 +502,29 @@ static void fbnic_get_stats64(struct net_device *dev,
 			rx_bytes = stats->bytes;
 			rx_packets = stats->packets;
 			rx_dropped = stats->dropped;
+			rx_length = stats->rx.length_errors;
 		} while (u64_stats_fetch_retry(&stats->syncp, start));
 
 		stats64->rx_bytes += rx_bytes;
 		stats64->rx_packets += rx_packets;
 		stats64->rx_dropped += rx_dropped;
+		stats64->rx_errors += rx_length;
+		stats64->rx_length_errors += rx_length;
+
+		if (!xdpr)
+			continue;
+
+		stats = &xdpr->stats;
+		do {
+			start = u64_stats_fetch_begin(&stats->syncp);
+			tx_bytes = stats->bytes;
+			tx_packets = stats->packets;
+			tx_dropped = stats->dropped;
+		} while (u64_stats_fetch_retry(&stats->syncp, start));
+
+		stats64->tx_bytes += tx_bytes;
+		stats64->tx_packets += tx_packets;
+		stats64->tx_dropped += tx_dropped;
 	}
 }
 
@@ -603,6 +623,7 @@ static void fbnic_get_queue_stats_tx(struct net_device *dev, int idx,
 	struct fbnic_ring *txr = fbn->tx[idx];
 	struct fbnic_queue_stats *stats;
 	u64 stop, wake, csum, lso;
+	struct fbnic_ring *xdpr;
 	unsigned int start;
 	u64 bytes, packets;
 
@@ -626,6 +647,19 @@ static void fbnic_get_queue_stats_tx(struct net_device *dev, int idx,
 	tx->hw_gso_wire_packets = lso;
 	tx->stop = stop;
 	tx->wake = wake;
+
+	xdpr = fbn->tx[FBNIC_MAX_TXQS + idx];
+	if (xdpr) {
+		stats = &xdpr->stats;
+		do {
+			start = u64_stats_fetch_begin(&stats->syncp);
+			bytes = stats->bytes;
+			packets = stats->packets;
+		} while (u64_stats_fetch_retry(&stats->syncp, start));
+
+		tx->bytes += bytes;
+		tx->packets += packets;
+	}
 }
 
 static void fbnic_get_base_stats(struct net_device *dev,

drivers/net/ethernet/meta/fbnic/fbnic_txrx.c

@@ -620,8 +620,8 @@ static void fbnic_clean_twq1(struct fbnic_napi_vector *nv, bool pp_allow_direct,
 			     struct fbnic_ring *ring, bool discard,
 			     unsigned int hw_head)
 {
+	u64 total_bytes = 0, total_packets = 0;
 	unsigned int head = ring->head;
-	u64 total_bytes = 0;
 
 	while (hw_head != head) {
 		struct page *page;
@@ -633,6 +633,11 @@ static void fbnic_clean_twq1(struct fbnic_napi_vector *nv, bool pp_allow_direct,
 		twd = le64_to_cpu(ring->desc[head]);
 		page = ring->tx_buf[head];
 
+		/* TYPE_AL is 2, TYPE_LAST_AL is 3. So this trick gives
+		 * us one increment per packet, with no branches.
+		 */
+		total_packets += FIELD_GET(FBNIC_TWD_TYPE_MASK, twd) -
+				 FBNIC_TWD_TYPE_AL;
 		total_bytes += FIELD_GET(FBNIC_TWD_LEN_MASK, twd);
 
 		page_pool_put_page(nv->page_pool, page, -1, pp_allow_direct);
@@ -645,6 +650,18 @@ next_desc:
 		return;
 
 	ring->head = head;
+
+	if (discard) {
+		u64_stats_update_begin(&ring->stats.syncp);
+		ring->stats.dropped += total_packets;
+		u64_stats_update_end(&ring->stats.syncp);
+		return;
+	}
+
+	u64_stats_update_begin(&ring->stats.syncp);
+	ring->stats.bytes += total_bytes;
+	ring->stats.packets += total_packets;
+	u64_stats_update_end(&ring->stats.syncp);
 }
 
 static void fbnic_clean_tsq(struct fbnic_napi_vector *nv,
@@ -1040,8 +1057,12 @@ static long fbnic_pkt_tx(struct fbnic_napi_vector *nv,
 		frag = &shinfo->frags[0];
 	}
 
-	if (fbnic_desc_unused(ring) < nsegs)
+	if (fbnic_desc_unused(ring) < nsegs) {
+		u64_stats_update_begin(&ring->stats.syncp);
+		ring->stats.dropped++;
+		u64_stats_update_end(&ring->stats.syncp);
 		return -FBNIC_XDP_CONSUME;
+	}
 
 	page = virt_to_page(pkt->buff.data_hard_start);
 	offset = offset_in_page(pkt->buff.data);
@@ -1181,8 +1202,8 @@ static int fbnic_clean_rcq(struct fbnic_napi_vector *nv,
 			   struct fbnic_q_triad *qt, int budget)
 {
 	unsigned int packets = 0, bytes = 0, dropped = 0, alloc_failed = 0;
+	u64 csum_complete = 0, csum_none = 0, length_errors = 0;
 	s32 head0 = -1, head1 = -1, pkt_tail = -1;
-	u64 csum_complete = 0, csum_none = 0;
 	struct fbnic_ring *rcq = &qt->cmpl;
 	struct fbnic_pkt_buff *pkt;
 	__le64 *raw_rcd, done;
@@ -1247,6 +1268,8 @@ static int fbnic_clean_rcq(struct fbnic_napi_vector *nv,
 				if (!skb) {
 					alloc_failed++;
 					dropped++;
+				} else if (skb == ERR_PTR(-FBNIC_XDP_LEN_ERR)) {
+					length_errors++;
 				} else {
 					dropped++;
 				}
@@ -1276,6 +1299,7 @@ static int fbnic_clean_rcq(struct fbnic_napi_vector *nv,
 	rcq->stats.rx.alloc_failed += alloc_failed;
 	rcq->stats.rx.csum_complete += csum_complete;
 	rcq->stats.rx.csum_none += csum_none;
+	rcq->stats.rx.length_errors += length_errors;
 	u64_stats_update_end(&rcq->stats.syncp);
 
 	if (pkt_tail >= 0)
@@ -1359,8 +1383,9 @@ void fbnic_aggregate_ring_rx_counters(struct fbnic_net *fbn,
 	fbn->rx_stats.rx.alloc_failed += stats->rx.alloc_failed;
 	fbn->rx_stats.rx.csum_complete += stats->rx.csum_complete;
 	fbn->rx_stats.rx.csum_none += stats->rx.csum_none;
+	fbn->rx_stats.rx.length_errors += stats->rx.length_errors;
 	/* Remember to add new stats here */
-	BUILD_BUG_ON(sizeof(fbn->rx_stats.rx) / 8 != 3);
+	BUILD_BUG_ON(sizeof(fbn->rx_stats.rx) / 8 != 4);
 }
 
 void fbnic_aggregate_ring_tx_counters(struct fbnic_net *fbn,
@@ -1382,6 +1407,22 @@ void fbnic_aggregate_ring_tx_counters(struct fbnic_net *fbn,
 	BUILD_BUG_ON(sizeof(fbn->tx_stats.twq) / 8 != 6);
 }
 
+static void fbnic_aggregate_ring_xdp_counters(struct fbnic_net *fbn,
+					      struct fbnic_ring *xdpr)
+{
+	struct fbnic_queue_stats *stats = &xdpr->stats;
+
+	if (!(xdpr->flags & FBNIC_RING_F_STATS))
+		return;
+
+	/* Capture stats from queues before dissasociating them */
+	fbn->rx_stats.bytes += stats->bytes;
+	fbn->rx_stats.packets += stats->packets;
+	fbn->rx_stats.dropped += stats->dropped;
+	fbn->tx_stats.bytes += stats->bytes;
+	fbn->tx_stats.packets += stats->packets;
+}
+
 static void fbnic_remove_tx_ring(struct fbnic_net *fbn,
 				 struct fbnic_ring *txr)
 {
@@ -1401,6 +1442,8 @@ static void fbnic_remove_xdp_ring(struct fbnic_net *fbn,
 	if (!(xdpr->flags & FBNIC_RING_F_STATS))
 		return;
 
+	fbnic_aggregate_ring_xdp_counters(fbn, xdpr);
+
 	/* Remove pointer to the Tx ring */
 	WARN_ON(fbn->tx[xdpr->q_idx] && fbn->tx[xdpr->q_idx] != xdpr);
 	fbn->tx[xdpr->q_idx] = NULL;

drivers/net/ethernet/meta/fbnic/fbnic_txrx.h

@@ -90,6 +90,7 @@ struct fbnic_queue_stats {
 			u64 alloc_failed;
 			u64 csum_complete;
 			u64 csum_none;
+			u64 length_errors;
 		} rx;
 	};
 	u64 dropped;