net: mana: Add support of XDP_REDIRECT action

Add a handler of the XDP_REDIRECT return code from a XDP program. The
packets will be flushed at the end of each RX/CQ NAPI poll cycle.
ndo_xdp_xmit() is implemented by sharing the code in mana_xdp_tx().
Ethtool per queue counters are added for XDP redirect and xmit operations.

Signed-off-by: Haiyang Zhang <[e-mail]>
Signed-off-by: Paolo Abeni <[e-mail]>

drivers/net/ethernet/microsoft/mana/mana.h

···
 	u64 bytes;
 	u64 xdp_drop;
 	u64 xdp_tx;
+	u64 xdp_redirect;
 	struct u64_stats_sync syncp;
 };
 
 struct mana_stats_tx {
 	u64 packets;
 	u64 bytes;
+	u64 xdp_xmit;
 	struct u64_stats_sync syncp;
 };
 
···
 	struct bpf_prog __rcu *bpf_prog;
 	struct xdp_rxq_info xdp_rxq;
 	struct page *xdp_save_page;
+	bool xdp_flush;
+	int xdp_rc; /* XDP redirect return code */
 
 	/* MUST BE THE LAST MEMBER:
 	 * Each receive buffer has an associated mana_recv_buf_oob.
···
 void mana_remove(struct gdma_dev *gd, bool suspending);
 
 void mana_xdp_tx(struct sk_buff *skb, struct net_device *ndev);
+int mana_xdp_xmit(struct net_device *ndev, int n, struct xdp_frame **frames,
+		  u32 flags);
 u32 mana_run_xdp(struct net_device *ndev, struct mana_rxq *rxq,
 		 struct xdp_buff *xdp, void *buf_va, uint pkt_len);
 struct bpf_prog *mana_xdp_get(struct mana_port_context *apc);

drivers/net/ethernet/microsoft/mana/mana_bpf.c

···
 	ndev->stats.tx_dropped++;
 }
 
+static int mana_xdp_xmit_fm(struct net_device *ndev, struct xdp_frame *frame,
+			    u16 q_idx)
+{
+	struct sk_buff *skb;
+
+	skb = xdp_build_skb_from_frame(frame, ndev);
+	if (unlikely(!skb))
+		return -ENOMEM;
+
+	skb_set_queue_mapping(skb, q_idx);
+
+	mana_xdp_tx(skb, ndev);
+
+	return 0;
+}
+
+int mana_xdp_xmit(struct net_device *ndev, int n, struct xdp_frame **frames,
+		  u32 flags)
+{
+	struct mana_port_context *apc = netdev_priv(ndev);
+	struct mana_stats_tx *tx_stats;
+	int i, count = 0;
+	u16 q_idx;
+
+	if (unlikely(!apc->port_is_up))
+		return 0;
+
+	q_idx = smp_processor_id() % ndev->real_num_tx_queues;
+
+	for (i = 0; i < n; i++) {
+		if (mana_xdp_xmit_fm(ndev, frames[i], q_idx))
+			break;
+
+		count++;
+	}
+
+	tx_stats = &apc->tx_qp[q_idx].txq.stats;
+
+	u64_stats_update_begin(&tx_stats->syncp);
+	tx_stats->xdp_xmit += count;
+	u64_stats_update_end(&tx_stats->syncp);
+
+	return count;
+}
+
 u32 mana_run_xdp(struct net_device *ndev, struct mana_rxq *rxq,
 		 struct xdp_buff *xdp, void *buf_va, uint pkt_len)
 {
+	struct mana_stats_rx *rx_stats;
 	struct bpf_prog *prog;
 	u32 act = XDP_PASS;
 
···
 
 	act = bpf_prog_run_xdp(prog, xdp);
 
+	rx_stats = &rxq->stats;
+
 	switch (act) {
 	case XDP_PASS:
 	case XDP_TX:
 	case XDP_DROP:
 		break;
+
+	case XDP_REDIRECT:
+		rxq->xdp_rc = xdp_do_redirect(ndev, xdp, prog);
+		if (!rxq->xdp_rc) {
+			rxq->xdp_flush = true;
+
+			u64_stats_update_begin(&rx_stats->syncp);
+			rx_stats->packets++;
+			rx_stats->bytes += pkt_len;
+			rx_stats->xdp_redirect++;
+			u64_stats_update_end(&rx_stats->syncp);
+
+			break;
+		}
+
+		fallthrough;
 
 	case XDP_ABORTED:
 		trace_xdp_exception(ndev, prog, act);

drivers/net/ethernet/microsoft/mana/mana_en.c

···
 #include <linux/inetdevice.h>
 #include <linux/etherdevice.h>
 #include <linux/ethtool.h>
+#include <linux/filter.h>
 #include <linux/mm.h>
 
 #include <net/checksum.h>
···
 	.ndo_validate_addr	= eth_validate_addr,
 	.ndo_get_stats64	= mana_get_stats64,
 	.ndo_bpf		= mana_bpf,
+	.ndo_xdp_xmit		= mana_xdp_xmit,
 };
 
 static void mana_cleanup_port_context(struct mana_port_context *apc)
···
 
 	act = mana_run_xdp(ndev, rxq, &xdp, buf_va, pkt_len);
 
+	if (act == XDP_REDIRECT && !rxq->xdp_rc)
+		return;
+
 	if (act != XDP_PASS && act != XDP_TX)
 		goto drop_xdp;
 
···
 static void mana_poll_rx_cq(struct mana_cq *cq)
 {
 	struct gdma_comp *comp = cq->gdma_comp_buf;
+	struct mana_rxq *rxq = cq->rxq;
 	int comp_read, i;
 
 	comp_read = mana_gd_poll_cq(cq->gdma_cq, comp, CQE_POLLING_BUFFER);
 	WARN_ON_ONCE(comp_read > CQE_POLLING_BUFFER);
+
+	rxq->xdp_flush = false;
 
 	for (i = 0; i < comp_read; i++) {
 		if (WARN_ON_ONCE(comp[i].is_sq))
···
 		if (WARN_ON_ONCE(comp[i].wq_num != cq->rxq->gdma_id))
 			return;
 
-		mana_process_rx_cqe(cq->rxq, cq, &comp[i]);
+		mana_process_rx_cqe(rxq, cq, &comp[i]);
 	}
+
+	if (rxq->xdp_flush)
+		xdp_do_flush();
 }
 
 static void mana_cq_handler(void *context, struct gdma_queue *gdma_queue)

drivers/net/ethernet/microsoft/mana/mana_ethtool.c

···
 	if (stringset != ETH_SS_STATS)
 		return -EINVAL;
 
-	return ARRAY_SIZE(mana_eth_stats) + num_queues * 6;
+	return ARRAY_SIZE(mana_eth_stats) + num_queues * 8;
 }
 
 static void mana_get_strings(struct net_device *ndev, u32 stringset, u8 *data)
···
 		p += ETH_GSTRING_LEN;
 		sprintf(p, "rx_%d_xdp_tx", i);
 		p += ETH_GSTRING_LEN;
+		sprintf(p, "rx_%d_xdp_redirect", i);
+		p += ETH_GSTRING_LEN;
 	}
 
 	for (i = 0; i < num_queues; i++) {
 		sprintf(p, "tx_%d_packets", i);
 		p += ETH_GSTRING_LEN;
 		sprintf(p, "tx_%d_bytes", i);
+		p += ETH_GSTRING_LEN;
+		sprintf(p, "tx_%d_xdp_xmit", i);
 		p += ETH_GSTRING_LEN;
 	}
 }
···
 	struct mana_stats_tx *tx_stats;
 	unsigned int start;
 	u64 packets, bytes;
+	u64 xdp_redirect;
+	u64 xdp_xmit;
 	u64 xdp_drop;
 	u64 xdp_tx;
 	int q, i = 0;
···
 			bytes = rx_stats->bytes;
 			xdp_drop = rx_stats->xdp_drop;
 			xdp_tx = rx_stats->xdp_tx;
+			xdp_redirect = rx_stats->xdp_redirect;
 		} while (u64_stats_fetch_retry_irq(&rx_stats->syncp, start));
 
 		data[i++] = packets;
 		data[i++] = bytes;
 		data[i++] = xdp_drop;
 		data[i++] = xdp_tx;
+		data[i++] = xdp_redirect;
 	}
 
 	for (q = 0; q < num_queues; q++) {
···
 			start = u64_stats_fetch_begin_irq(&tx_stats->syncp);
 			packets = tx_stats->packets;
 			bytes = tx_stats->bytes;
+			xdp_xmit = tx_stats->xdp_xmit;
 		} while (u64_stats_fetch_retry_irq(&tx_stats->syncp, start));
 
 		data[i++] = packets;
 		data[i++] = bytes;
+		data[i++] = xdp_xmit;
 	}
 }