sfc: handle XDP_TX outcomes of XDP eBPF programs

Provide an ndo_xdp_xmit function that uses the XDP tx queue for this
CPU to send the packet.
Signed-off-by: Charles McLachlan <cmclachlan@solarflare.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/sfc/efx.c

@@ -228,6 +228,8 @@ static void efx_start_all(struct efx_nic *efx);
 static void efx_stop_all(struct efx_nic *efx);
 static int efx_xdp_setup_prog(struct efx_nic *efx, struct bpf_prog *prog);
 static int efx_xdp(struct net_device *dev, struct netdev_bpf *xdp);
+static int efx_xdp_xmit(struct net_device *dev, int n, struct xdp_frame **xdpfs,
+			u32 flags);
 
 #define EFX_ASSERT_RESET_SERIALISED(efx)		\
 	do {						\
@@ -2633,6 +2635,7 @@ static const struct net_device_ops efx_netdev_ops = {
 #endif
 	.ndo_udp_tunnel_add	= efx_udp_tunnel_add,
 	.ndo_udp_tunnel_del	= efx_udp_tunnel_del,
+	.ndo_xdp_xmit		= efx_xdp_xmit,
 	.ndo_bpf		= efx_xdp
 };
 
@@ -2680,6 +2683,17 @@ static int efx_xdp(struct net_device *dev, struct netdev_bpf *xdp)
 	}
 }
 
+static int efx_xdp_xmit(struct net_device *dev, int n, struct xdp_frame **xdpfs,
+			u32 flags)
+{
+	struct efx_nic *efx = netdev_priv(dev);
+
+	if (!netif_running(dev))
+		return -EINVAL;
+
+	return efx_xdp_tx_buffers(efx, n, xdpfs, flags & XDP_XMIT_FLUSH);
+}
+
 static void efx_update_name(struct efx_nic *efx)
 {
 	strcpy(efx->name, efx->net_dev->name);

drivers/net/ethernet/sfc/efx.h

@@ -322,4 +322,7 @@ static inline bool efx_rwsem_assert_write_locked(struct rw_semaphore *sem)
 	return true;
 }
 
+int efx_xdp_tx_buffers(struct efx_nic *efx, int n, struct xdp_frame **xdpfs,
+		       bool flush);
+
 #endif /* EFX_EFX_H */

drivers/net/ethernet/sfc/rx.c

@@ -653,6 +653,7 @@ static bool efx_do_xdp(struct efx_nic *efx, struct efx_channel *channel,
 	u8 rx_prefix[EFX_MAX_RX_PREFIX_SIZE];
 	struct efx_rx_queue *rx_queue;
 	struct bpf_prog *xdp_prog;
+	struct xdp_frame *xdpf;
 	struct xdp_buff xdp;
 	u32 xdp_act;
 	s16 offset;
@@ -713,7 +714,16 @@ static bool efx_do_xdp(struct efx_nic *efx, struct efx_channel *channel,
 		break;
 
 	case XDP_TX:
-		return -EOPNOTSUPP;
+		/* Buffer ownership passes to tx on success. */
+		xdpf = convert_to_xdp_frame(&xdp);
+		err = efx_xdp_tx_buffers(efx, 1, &xdpf, true);
+		if (unlikely(err != 1)) {
+			efx_free_rx_buffers(rx_queue, rx_buf, 1);
+			if (net_ratelimit())
+				netif_err(efx, rx_err, efx->net_dev,
+					  "XDP TX failed (%d)\n", err);
+		}
+		break;
 
 	case XDP_REDIRECT:
 		err = xdp_do_redirect(efx->net_dev, &xdp, xdp_prog);

drivers/net/ethernet/sfc/tx.c

@@ -599,6 +599,94 @@ netdev_tx_t efx_enqueue_skb(struct efx_tx_queue *tx_queue, struct sk_buff *skb)
 	return NETDEV_TX_OK;
 }
 
+static void efx_xdp_return_frames(int n,  struct xdp_frame **xdpfs)
+{
+	int i;
+
+	for (i = 0; i < n; i++)
+		xdp_return_frame_rx_napi(xdpfs[i]);
+}
+
+/* Transmit a packet from an XDP buffer
+ *
+ * Returns number of packets sent on success, error code otherwise.
+ * Runs in NAPI context, either in our poll (for XDP TX) or a different NIC
+ * (for XDP redirect).
+ */
+int efx_xdp_tx_buffers(struct efx_nic *efx, int n, struct xdp_frame **xdpfs,
+		       bool flush)
+{
+	struct efx_tx_buffer *tx_buffer;
+	struct efx_tx_queue *tx_queue;
+	struct xdp_frame *xdpf;
+	dma_addr_t dma_addr;
+	unsigned int len;
+	int space;
+	int cpu;
+	int i;
+
+	cpu = raw_smp_processor_id();
+
+	if (!efx->xdp_tx_queue_count ||
+	    unlikely(cpu >= efx->xdp_tx_queue_count))
+		return -EINVAL;
+
+	tx_queue = efx->xdp_tx_queues[cpu];
+	if (unlikely(!tx_queue))
+		return -EINVAL;
+
+	if (unlikely(n && !xdpfs))
+		return -EINVAL;
+
+	if (!n)
+		return 0;
+
+	/* Check for available space. We should never need multiple
+	 * descriptors per frame.
+	 */
+	space = efx->txq_entries +
+		tx_queue->read_count - tx_queue->insert_count;
+
+	for (i = 0; i < n; i++) {
+		xdpf = xdpfs[i];
+
+		if (i >= space)
+			break;
+
+		/* We'll want a descriptor for this tx. */
+		prefetchw(__efx_tx_queue_get_insert_buffer(tx_queue));
+
+		len = xdpf->len;
+
+		/* Map for DMA. */
+		dma_addr = dma_map_single(&efx->pci_dev->dev,
+					  xdpf->data, len,
+					  DMA_TO_DEVICE);
+		if (dma_mapping_error(&efx->pci_dev->dev, dma_addr))
+			break;
+
+		/*  Create descriptor and set up for unmapping DMA. */
+		tx_buffer = efx_tx_map_chunk(tx_queue, dma_addr, len);
+		tx_buffer->xdpf = xdpf;
+		tx_buffer->flags = EFX_TX_BUF_XDP |
+				   EFX_TX_BUF_MAP_SINGLE;
+		tx_buffer->dma_offset = 0;
+		tx_buffer->unmap_len = len;
+		tx_queue->tx_packets++;
+	}
+
+	/* Pass mapped frames to hardware. */
+	if (flush && i > 0)
+		efx_nic_push_buffers(tx_queue);
+
+	if (i == 0)
+		return -EIO;
+
+	efx_xdp_return_frames(n - i, xdpfs + i);
+
+	return i;
+}
+
 /* Remove packets from the TX queue
  *
  * This removes packets from the TX queue, up to and including the