[PATCH 0/2] ixgbe updates

[PATCH 0/2] ixgbe updates

[John Fastabend]

Hi Jeff, Here are two patch updates for your tree. I merged in Alex's
build fix and Jakub's comments. Lightly tested xdp1 and xdp2.

---

John Fastabend (2):
      ixgbe: add XDP support for pass and drop actions
      ixgbe: add support for XDP_TX action


 0 files changed

[PATCH 1/2] ixgbe: add XDP support for pass and drop actions

[John Fastabend]

Basic XDP drop support for ixgbe. Uses READ_ONCE/xchg semantics on XDP
programs instead of rcu primitives as suggested by Daniel Borkmann and
Alex Duyck.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Acked-by: Alexander Duyck <alexander.h.duyck@intel.com>
---
 0 files changed

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe.h b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
index 656ca8f..cb14813 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe.h
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
@@ -318,6 +318,7 @@ struct ixgbe_ring {
 	struct ixgbe_ring *next;	/* pointer to next ring in q_vector */
 	struct ixgbe_q_vector *q_vector; /* backpointer to host q_vector */
 	struct net_device *netdev;	/* netdev ring belongs to */
+	struct bpf_prog *xdp_prog;
 	struct device *dev;		/* device for DMA mapping */
 	struct ixgbe_fwd_adapter *l2_accel_priv;
 	void *desc;			/* descriptor ring memory */
@@ -555,6 +556,7 @@ struct ixgbe_adapter {
 	unsigned long active_vlans[BITS_TO_LONGS(VLAN_N_VID)];
 	/* OS defined structs */
 	struct net_device *netdev;
+	struct bpf_prog *xdp_prog;
 	struct pci_dev *pdev;
 
 	unsigned long state;
@@ -835,7 +837,7 @@ enum ixgbe_boards {
 void ixgbe_reinit_locked(struct ixgbe_adapter *adapter);
 void ixgbe_reset(struct ixgbe_adapter *adapter);
 void ixgbe_set_ethtool_ops(struct net_device *netdev);
-int ixgbe_setup_rx_resources(struct ixgbe_ring *);
+int ixgbe_setup_rx_resources(struct ixgbe_adapter *, struct ixgbe_ring *);
 int ixgbe_setup_tx_resources(struct ixgbe_ring *);
 void ixgbe_free_rx_resources(struct ixgbe_ring *);
 void ixgbe_free_tx_resources(struct ixgbe_ring *);
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c
index 59730ed..79a126d 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c
@@ -1128,7 +1128,7 @@ static int ixgbe_set_ringparam(struct net_device *netdev,
 			       sizeof(struct ixgbe_ring));
 
 			temp_ring[i].count = new_rx_count;
-			err = ixgbe_setup_rx_resources(&temp_ring[i]);
+			err = ixgbe_setup_rx_resources(adapter, &temp_ring[i]);
 			if (err) {
 				while (i) {
 					i--;
@@ -1761,7 +1761,7 @@ static int ixgbe_setup_desc_rings(struct ixgbe_adapter *adapter)
 	rx_ring->netdev = adapter->netdev;
 	rx_ring->reg_idx = adapter->rx_ring[0]->reg_idx;
 
-	err = ixgbe_setup_rx_resources(rx_ring);
+	err = ixgbe_setup_rx_resources(adapter, rx_ring);
 	if (err) {
 		ret_val = 4;
 		goto err_nomem;
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index 730742a..ebcf9c0 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -49,6 +49,9 @@
 #include <linux/if_macvlan.h>
 #include <linux/if_bridge.h>
 #include <linux/prefetch.h>
+#include <linux/bpf.h>
+#include <linux/bpf_trace.h>
+#include <linux/atomic.h>
 #include <scsi/fc/fc_fcoe.h>
 #include <net/udp_tunnel.h>
 #include <net/pkt_cls.h>
@@ -1856,6 +1859,10 @@ static void ixgbe_dma_sync_frag(struct ixgbe_ring *rx_ring,
  * @rx_desc: pointer to the EOP Rx descriptor
  * @skb: pointer to current skb being fixed
  *
+ * Check if the skb is valid in the XDP case it will be an error pointer.
+ * Return true in this case to abort processing and advance to next
+ * descriptor.
+ *
  * Check for corrupted packet headers caused by senders on the local L2
  * embedded NIC switch not setting up their Tx Descriptors right.  These
  * should be very rare.
@@ -1874,6 +1881,10 @@ static bool ixgbe_cleanup_headers(struct ixgbe_ring *rx_ring,
 {
 	struct net_device *netdev = rx_ring->netdev;
 
+	/* XDP packets use error pointer so abort at this point */
+	if (IS_ERR(skb))
+		return true;
+
 	/* verify that the packet does not have any known errors */
 	if (unlikely(ixgbe_test_staterr(rx_desc,
 					IXGBE_RXDADV_ERR_FRAME_ERR_MASK) &&
@@ -2049,7 +2060,7 @@ static void ixgbe_put_rx_buffer(struct ixgbe_ring *rx_ring,
 		/* hand second half of page back to the ring */
 		ixgbe_reuse_rx_page(rx_ring, rx_buffer);
 	} else {
-		if (IXGBE_CB(skb)->dma == rx_buffer->dma) {
+		if (!IS_ERR(skb) && IXGBE_CB(skb)->dma == rx_buffer->dma) {
 			/* the page has been released from the ring */
 			IXGBE_CB(skb)->page_released = true;
 		} else {
@@ -2070,21 +2081,22 @@ static void ixgbe_put_rx_buffer(struct ixgbe_ring *rx_ring,
 
 static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
 					   struct ixgbe_rx_buffer *rx_buffer,
-					   union ixgbe_adv_rx_desc *rx_desc,
-					   unsigned int size)
+					   struct xdp_buff *xdp,
+					   union ixgbe_adv_rx_desc *rx_desc)
 {
-	void *va = page_address(rx_buffer->page) + rx_buffer->page_offset;
+	unsigned int size = xdp->data_end - xdp->data;
 #if (PAGE_SIZE < 8192)
 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
 #else
-	unsigned int truesize = SKB_DATA_ALIGN(size);
+	unsigned int truesize = SKB_DATA_ALIGN(xdp->data_end -
+					       xdp->data_hard_start);
 #endif
 	struct sk_buff *skb;
 
 	/* prefetch first cache line of first page */
-	prefetch(va);
+	prefetch(xdp->data);
 #if L1_CACHE_BYTES < 128
-	prefetch(va + L1_CACHE_BYTES);
+	prefetch(xdp->data + L1_CACHE_BYTES);
 #endif
 
 	/* allocate a skb to store the frags */
@@ -2097,7 +2109,7 @@ static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
 			IXGBE_CB(skb)->dma = rx_buffer->dma;
 
 		skb_add_rx_frag(skb, 0, rx_buffer->page,
-				rx_buffer->page_offset,
+				xdp->data - page_address(rx_buffer->page),
 				size, truesize);
 #if (PAGE_SIZE < 8192)
 		rx_buffer->page_offset ^= truesize;
@@ -2105,7 +2117,8 @@ static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
 		rx_buffer->page_offset += truesize;
 #endif
 	} else {
-		memcpy(__skb_put(skb, size), va, ALIGN(size, sizeof(long)));
+		memcpy(__skb_put(skb, size),
+		       xdp->data, ALIGN(size, sizeof(long)));
 		rx_buffer->pagecnt_bias++;
 	}
 
@@ -2114,32 +2127,32 @@ static struct sk_buff *ixgbe_construct_skb(struct ixgbe_ring *rx_ring,
 
 static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
 				       struct ixgbe_rx_buffer *rx_buffer,
-				       union ixgbe_adv_rx_desc *rx_desc,
-				       unsigned int size)
+				       struct xdp_buff *xdp,
+				       union ixgbe_adv_rx_desc *rx_desc)
 {
-	void *va = page_address(rx_buffer->page) + rx_buffer->page_offset;
 #if (PAGE_SIZE < 8192)
 	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
 #else
 	unsigned int truesize = SKB_DATA_ALIGN(sizeof(struct skb_shared_info)) +
-				SKB_DATA_ALIGN(IXGBE_SKB_PAD + size);
+				SKB_DATA_ALIGN(xdp->data_end -
+					       xdp->data_hard_start);
 #endif
 	struct sk_buff *skb;
 
 	/* prefetch first cache line of first page */
-	prefetch(va);
+	prefetch(xdp->data);
 #if L1_CACHE_BYTES < 128
-	prefetch(va + L1_CACHE_BYTES);
+	prefetch(xdp->data + L1_CACHE_BYTES);
 #endif
 
-	/* build an skb around the page buffer */
-	skb = build_skb(va - IXGBE_SKB_PAD, truesize);
+	/* build an skb to around the page buffer */
+	skb = build_skb(xdp->data_hard_start, truesize);
 	if (unlikely(!skb))
 		return NULL;
 
 	/* update pointers within the skb to store the data */
-	skb_reserve(skb, IXGBE_SKB_PAD);
-	__skb_put(skb, size);
+	skb_reserve(skb, xdp->data - xdp->data_hard_start);
+	__skb_put(skb, xdp->data_end - xdp->data);
 
 	/* record DMA address if this is the start of a chain of buffers */
 	if (!ixgbe_test_staterr(rx_desc, IXGBE_RXD_STAT_EOP))
@@ -2155,6 +2168,41 @@ static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
 	return skb;
 }
 
+#define IXGBE_XDP_PASS 0
+#define IXGBE_XDP_CONSUMED 1
+
+static struct sk_buff *ixgbe_run_xdp(struct ixgbe_ring  *rx_ring,
+				     struct xdp_buff *xdp)
+{
+	int result = IXGBE_XDP_PASS;
+	struct bpf_prog *xdp_prog;
+	u32 act;
+
+	rcu_read_lock();
+	xdp_prog = READ_ONCE(rx_ring->xdp_prog);
+
+	if (!xdp_prog)
+		goto xdp_out;
+
+	act = bpf_prog_run_xdp(xdp_prog, xdp);
+	switch (act) {
+	case XDP_PASS:
+		break;
+	default:
+		bpf_warn_invalid_xdp_action(act);
+	case XDP_TX:
+	case XDP_ABORTED:
+		trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
+		/* fallthrough -- handle aborts by dropping packet */
+	case XDP_DROP:
+		result = IXGBE_XDP_CONSUMED;
+		break;
+	}
+xdp_out:
+	rcu_read_unlock();
+	return ERR_PTR(-result);
+}
+
 /**
  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
  * @q_vector: structure containing interrupt and ring information
@@ -2184,6 +2232,7 @@ static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
 		union ixgbe_adv_rx_desc *rx_desc;
 		struct ixgbe_rx_buffer *rx_buffer;
 		struct sk_buff *skb;
+		struct xdp_buff xdp;
 		unsigned int size;
 
 		/* return some buffers to hardware, one at a time is too slow */
@@ -2206,14 +2255,29 @@ static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
 		rx_buffer = ixgbe_get_rx_buffer(rx_ring, rx_desc, &skb, size);
 
 		/* retrieve a buffer from the ring */
-		if (skb)
+		if (!skb) {
+			xdp.data = page_address(rx_buffer->page) +
+						rx_buffer->page_offset;
+			xdp.data_hard_start = xdp.data -
+						ixgbe_rx_offset(rx_ring);
+			xdp.data_end = xdp.data + size;
+
+			skb = ixgbe_run_xdp(rx_ring, &xdp);
+		}
+
+		if (IS_ERR(skb)) {
+			total_rx_packets++;
+			total_rx_bytes += size;
+			rx_buffer->pagecnt_bias++;
+		} else if (skb) {
 			ixgbe_add_rx_frag(rx_ring, rx_buffer, skb, size);
-		else if (ring_uses_build_skb(rx_ring))
+		} else if (ring_uses_build_skb(rx_ring)) {
 			skb = ixgbe_build_skb(rx_ring, rx_buffer,
-					      rx_desc, size);
-		else
+					      &xdp, rx_desc);
+		} else {
 			skb = ixgbe_construct_skb(rx_ring, rx_buffer,
-						  rx_desc, size);
+						  &xdp, rx_desc);
+		}
 
 		/* exit if we failed to retrieve a buffer */
 		if (!skb) {
@@ -6072,7 +6136,8 @@ static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
  *
  * Returns 0 on success, negative on failure
  **/
-int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
+int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
+			     struct ixgbe_ring *rx_ring)
 {
 	struct device *dev = rx_ring->dev;
 	int orig_node = dev_to_node(dev);
@@ -6109,6 +6174,8 @@ int ixgbe_setup_rx_resources(struct ixgbe_ring *rx_ring)
 	rx_ring->next_to_clean = 0;
 	rx_ring->next_to_use = 0;
 
+	rx_ring->xdp_prog = adapter->xdp_prog;
+
 	return 0;
 err:
 	vfree(rx_ring->rx_buffer_info);
@@ -6132,7 +6199,7 @@ static int ixgbe_setup_all_rx_resources(struct ixgbe_adapter *adapter)
 	int i, err = 0;
 
 	for (i = 0; i < adapter->num_rx_queues; i++) {
-		err = ixgbe_setup_rx_resources(adapter->rx_ring[i]);
+		err = ixgbe_setup_rx_resources(adapter, adapter->rx_ring[i]);
 		if (!err)
 			continue;
 
@@ -6200,6 +6267,7 @@ void ixgbe_free_rx_resources(struct ixgbe_ring *rx_ring)
 {
 	ixgbe_clean_rx_ring(rx_ring);
 
+	rx_ring->xdp_prog = NULL;
 	vfree(rx_ring->rx_buffer_info);
 	rx_ring->rx_buffer_info = NULL;
 
@@ -9468,6 +9536,54 @@ static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
 	return features;
 }
 
+static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
+{
+	int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
+	struct ixgbe_adapter *adapter = netdev_priv(dev);
+	struct bpf_prog *old_prog;
+
+	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
+		return -EINVAL;
+
+	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
+		return -EINVAL;
+
+	/* verify ixgbe ring attributes are sufficient for XDP */
+	for (i = 0; i < adapter->num_rx_queues; i++) {
+		struct ixgbe_ring *ring = adapter->rx_ring[i];
+
+		if (ring_is_rsc_enabled(ring))
+			return -EINVAL;
+
+		if (frame_size > ixgbe_rx_bufsz(ring))
+			return -EINVAL;
+	}
+
+	old_prog = xchg(&adapter->xdp_prog, prog);
+	for (i = 0; i < adapter->num_rx_queues; i++)
+		xchg(&adapter->rx_ring[i]->xdp_prog, adapter->xdp_prog);
+
+	if (old_prog)
+		bpf_prog_put(old_prog);
+
+	return 0;
+}
+
+static int ixgbe_xdp(struct net_device *dev, struct netdev_xdp *xdp)
+{
+	struct ixgbe_adapter *adapter = netdev_priv(dev);
+
+	switch (xdp->command) {
+	case XDP_SETUP_PROG:
+		return ixgbe_xdp_setup(dev, xdp->prog);
+	case XDP_QUERY_PROG:
+		xdp->prog_attached = !!(adapter->xdp_prog);
+		return 0;
+	default:
+		return -EINVAL;
+	}
+}
+
 static const struct net_device_ops ixgbe_netdev_ops = {
 	.ndo_open		= ixgbe_open,
 	.ndo_stop		= ixgbe_close,
@@ -9513,6 +9629,7 @@ static void ixgbe_fwd_del(struct net_device *pdev, void *priv)
 	.ndo_udp_tunnel_add	= ixgbe_add_udp_tunnel_port,
 	.ndo_udp_tunnel_del	= ixgbe_del_udp_tunnel_port,
 	.ndo_features_check	= ixgbe_features_check,
+	.ndo_xdp		= ixgbe_xdp,
 };
 
 /**

[PATCH 2/2] ixgbe: add support for XDP_TX action

[John Fastabend]

From: John Fastabend <john.fastabend@gmail.com>

Add support for XDP_TX action.

A couple design choices were made here. First I use a new ring
pointer structure xdp_ring[] in the adapter struct instead of
pushing the newly allocated xdp TX rings into the tx_ring[]
structure. This means we have to duplicate loops around rings
in places we want to initialize both TX rings and XDP rings.
But by making it explicit it is obvious when we are using XDP
rings and when we are using TX rings. Further we don't have
to do ring arithmatic which is error prone. As a proof point
for doing this my first patches used only a single ring structure
and introduced bugs in FCoE code and macvlan code paths.

Second I am aware this is not the most optimized version of
this code possible. I want to get baseline support in using
the most readable format possible and then once this series
is included I will optimize the TX path in another series
of patches.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
---
 0 files changed

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe.h b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
index cb14813..e8cd449 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe.h
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
@@ -235,7 +235,11 @@ struct vf_macvlans {
 struct ixgbe_tx_buffer {
 	union ixgbe_adv_tx_desc *next_to_watch;
 	unsigned long time_stamp;
-	struct sk_buff *skb;
+	union {
+		struct sk_buff *skb;
+		/* XDP uses address ptr on irq_clean */
+		void *data;
+	};
 	unsigned int bytecount;
 	unsigned short gso_segs;
 	__be16 protocol;
@@ -288,6 +292,7 @@ enum ixgbe_ring_state_t {
 	__IXGBE_TX_XPS_INIT_DONE,
 	__IXGBE_TX_DETECT_HANG,
 	__IXGBE_HANG_CHECK_ARMED,
+	__IXGBE_TX_XDP_RING,
 };
 
 #define ring_uses_build_skb(ring) \
@@ -314,6 +319,12 @@ struct ixgbe_fwd_adapter {
 	set_bit(__IXGBE_RX_RSC_ENABLED, &(ring)->state)
 #define clear_ring_rsc_enabled(ring) \
 	clear_bit(__IXGBE_RX_RSC_ENABLED, &(ring)->state)
+#define ring_is_xdp(ring) \
+	test_bit(__IXGBE_TX_XDP_RING, &(ring)->state)
+#define set_ring_xdp(ring) \
+	set_bit(__IXGBE_TX_XDP_RING, &(ring)->state)
+#define clear_ring_xdp(ring) \
+	clear_bit(__IXGBE_TX_XDP_RING, &(ring)->state)
 struct ixgbe_ring {
 	struct ixgbe_ring *next;	/* pointer to next ring in q_vector */
 	struct ixgbe_q_vector *q_vector; /* backpointer to host q_vector */
@@ -380,6 +391,7 @@ enum ixgbe_ring_f_enum {
 #define IXGBE_MAX_FCOE_INDICES		8
 #define MAX_RX_QUEUES			(IXGBE_MAX_FDIR_INDICES + 1)
 #define MAX_TX_QUEUES			(IXGBE_MAX_FDIR_INDICES + 1)
+#define MAX_XDP_QUEUES			(IXGBE_MAX_FDIR_INDICES + 1)
 #define IXGBE_MAX_L2A_QUEUES		4
 #define IXGBE_BAD_L2A_QUEUE		3
 #define IXGBE_MAX_MACVLANS		31
@@ -623,6 +635,10 @@ struct ixgbe_adapter {
 	__be16 vxlan_port;
 	__be16 geneve_port;
 
+	/* XDP */
+	int num_xdp_queues;
+	struct ixgbe_ring *xdp_ring[MAX_XDP_QUEUES];
+
 	/* TX */
 	struct ixgbe_ring *tx_ring[MAX_TX_QUEUES] ____cacheline_aligned_in_smp;
 
@@ -669,6 +685,7 @@ struct ixgbe_adapter {
 
 	u64 tx_busy;
 	unsigned int tx_ring_count;
+	unsigned int xdp_ring_count;
 	unsigned int rx_ring_count;
 
 	u32 link_speed;
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c
index 79a126d..b0fd2f5 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c
@@ -1071,15 +1071,19 @@ static int ixgbe_set_ringparam(struct net_device *netdev,
 	if (!netif_running(adapter->netdev)) {
 		for (i = 0; i < adapter->num_tx_queues; i++)
 			adapter->tx_ring[i]->count = new_tx_count;
+		for (i = 0; i < adapter->num_xdp_queues; i++)
+			adapter->xdp_ring[i]->count = new_tx_count;
 		for (i = 0; i < adapter->num_rx_queues; i++)
 			adapter->rx_ring[i]->count = new_rx_count;
 		adapter->tx_ring_count = new_tx_count;
+		adapter->xdp_ring_count = new_tx_count;
 		adapter->rx_ring_count = new_rx_count;
 		goto clear_reset;
 	}
 
 	/* allocate temporary buffer to store rings in */
 	i = max_t(int, adapter->num_tx_queues, adapter->num_rx_queues);
+	i = max_t(int, i, adapter->num_xdp_queues);
 	temp_ring = vmalloc(i * sizeof(struct ixgbe_ring));
 
 	if (!temp_ring) {
@@ -1111,12 +1115,33 @@ static int ixgbe_set_ringparam(struct net_device *netdev,
 			}
 		}
 
+		for (i = 0; i < adapter->num_xdp_queues; i++) {
+			memcpy(&temp_ring[i], adapter->xdp_ring[i],
+			       sizeof(struct ixgbe_ring));
+
+			temp_ring[i].count = new_tx_count;
+			err = ixgbe_setup_tx_resources(&temp_ring[i]);
+			if (err) {
+				while (i) {
+					i--;
+					ixgbe_free_tx_resources(&temp_ring[i]);
+				}
+				goto err_setup;
+			}
+		}
+
 		for (i = 0; i < adapter->num_tx_queues; i++) {
 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
 
 			memcpy(adapter->tx_ring[i], &temp_ring[i],
 			       sizeof(struct ixgbe_ring));
 		}
+		for (i = 0; i < adapter->num_xdp_queues; i++) {
+			ixgbe_free_tx_resources(adapter->xdp_ring[i]);
+
+			memcpy(adapter->xdp_ring[i], &temp_ring[i],
+			       sizeof(struct ixgbe_ring));
+		}
 
 		adapter->tx_ring_count = new_tx_count;
 	}
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_lib.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_lib.c
index 1b8be7d..b45fdc9 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_lib.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_lib.c
@@ -267,12 +267,14 @@ static bool ixgbe_cache_ring_sriov(struct ixgbe_adapter *adapter)
  **/
 static bool ixgbe_cache_ring_rss(struct ixgbe_adapter *adapter)
 {
-	int i;
+	int i, reg_idx;
 
 	for (i = 0; i < adapter->num_rx_queues; i++)
 		adapter->rx_ring[i]->reg_idx = i;
-	for (i = 0; i < adapter->num_tx_queues; i++)
-		adapter->tx_ring[i]->reg_idx = i;
+	for (i = 0, reg_idx = 0; i < adapter->num_tx_queues; i++, reg_idx++)
+		adapter->tx_ring[i]->reg_idx = reg_idx;
+	for (i = 0; i < adapter->num_xdp_queues; i++, reg_idx++)
+		adapter->xdp_ring[i]->reg_idx = reg_idx;
 
 	return true;
 }
@@ -308,6 +310,11 @@ static void ixgbe_cache_ring_register(struct ixgbe_adapter *adapter)
 	ixgbe_cache_ring_rss(adapter);
 }
 
+static int ixgbe_xdp_queues(struct ixgbe_adapter *adapter)
+{
+	return adapter->xdp_prog ? nr_cpu_ids : 0;
+}
+
 #define IXGBE_RSS_64Q_MASK	0x3F
 #define IXGBE_RSS_16Q_MASK	0xF
 #define IXGBE_RSS_8Q_MASK	0x7
@@ -382,6 +389,7 @@ static bool ixgbe_set_dcb_sriov_queues(struct ixgbe_adapter *adapter)
 	adapter->num_rx_queues_per_pool = tcs;
 
 	adapter->num_tx_queues = vmdq_i * tcs;
+	adapter->num_xdp_queues = 0;
 	adapter->num_rx_queues = vmdq_i * tcs;
 
 #ifdef IXGBE_FCOE
@@ -479,6 +487,7 @@ static bool ixgbe_set_dcb_queues(struct ixgbe_adapter *adapter)
 		netdev_set_tc_queue(dev, i, rss_i, rss_i * i);
 
 	adapter->num_tx_queues = rss_i * tcs;
+	adapter->num_xdp_queues = 0;
 	adapter->num_rx_queues = rss_i * tcs;
 
 	return true;
@@ -549,6 +558,7 @@ static bool ixgbe_set_sriov_queues(struct ixgbe_adapter *adapter)
 
 	adapter->num_rx_queues = vmdq_i * rss_i;
 	adapter->num_tx_queues = vmdq_i * rss_i;
+	adapter->num_xdp_queues = 0;
 
 	/* disable ATR as it is not supported when VMDq is enabled */
 	adapter->flags &= ~IXGBE_FLAG_FDIR_HASH_CAPABLE;
@@ -669,6 +679,7 @@ static bool ixgbe_set_rss_queues(struct ixgbe_adapter *adapter)
 #endif /* IXGBE_FCOE */
 	adapter->num_rx_queues = rss_i;
 	adapter->num_tx_queues = rss_i;
+	adapter->num_xdp_queues = ixgbe_xdp_queues(adapter);
 
 	return true;
 }
@@ -689,6 +700,7 @@ static void ixgbe_set_num_queues(struct ixgbe_adapter *adapter)
 	/* Start with base case */
 	adapter->num_rx_queues = 1;
 	adapter->num_tx_queues = 1;
+	adapter->num_xdp_queues = 0;
 	adapter->num_rx_pools = adapter->num_rx_queues;
 	adapter->num_rx_queues_per_pool = 1;
 
@@ -719,8 +731,11 @@ static int ixgbe_acquire_msix_vectors(struct ixgbe_adapter *adapter)
 	struct ixgbe_hw *hw = &adapter->hw;
 	int i, vectors, vector_threshold;
 
-	/* We start by asking for one vector per queue pair */
+	/* We start by asking for one vector per queue pair with XDP queues
+	 * being stacked with TX queues.
+	 */
 	vectors = max(adapter->num_rx_queues, adapter->num_tx_queues);
+	vectors = max(vectors, adapter->num_xdp_queues);
 
 	/* It is easy to be greedy for MSI-X vectors. However, it really
 	 * doesn't do much good if we have a lot more vectors than CPUs. We'll
@@ -800,6 +815,8 @@ static void ixgbe_add_ring(struct ixgbe_ring *ring,
  * @v_idx: index of vector in adapter struct
  * @txr_count: total number of Tx rings to allocate
  * @txr_idx: index of first Tx ring to allocate
+ * @xdp_count: total number of XDP rings to allocate
+ * @xdp_idx: index of first XDP ring to allocate
  * @rxr_count: total number of Rx rings to allocate
  * @rxr_idx: index of first Rx ring to allocate
  *
@@ -808,6 +825,7 @@ static void ixgbe_add_ring(struct ixgbe_ring *ring,
 static int ixgbe_alloc_q_vector(struct ixgbe_adapter *adapter,
 				int v_count, int v_idx,
 				int txr_count, int txr_idx,
+				int xdp_count, int xdp_idx,
 				int rxr_count, int rxr_idx)
 {
 	struct ixgbe_q_vector *q_vector;
@@ -817,7 +835,7 @@ static int ixgbe_alloc_q_vector(struct ixgbe_adapter *adapter,
 	int ring_count, size;
 	u8 tcs = netdev_get_num_tc(adapter->netdev);
 
-	ring_count = txr_count + rxr_count;
+	ring_count = txr_count + rxr_count + xdp_count;
 	size = sizeof(struct ixgbe_q_vector) +
 	       (sizeof(struct ixgbe_ring) * ring_count);
 
@@ -909,6 +927,33 @@ static int ixgbe_alloc_q_vector(struct ixgbe_adapter *adapter,
 		ring++;
 	}
 
+	while (xdp_count) {
+		/* assign generic ring traits */
+		ring->dev = &adapter->pdev->dev;
+		ring->netdev = adapter->netdev;
+
+		/* configure backlink on ring */
+		ring->q_vector = q_vector;
+
+		/* update q_vector Tx values */
+		ixgbe_add_ring(ring, &q_vector->tx);
+
+		/* apply Tx specific ring traits */
+		ring->count = adapter->tx_ring_count;
+		ring->queue_index = xdp_idx;
+		set_ring_xdp(ring);
+
+		/* assign ring to adapter */
+		adapter->xdp_ring[xdp_idx] = ring;
+
+		/* update count and index */
+		xdp_count--;
+		xdp_idx++;
+
+		/* push pointer to next ring */
+		ring++;
+	}
+
 	while (rxr_count) {
 		/* assign generic ring traits */
 		ring->dev = &adapter->pdev->dev;
@@ -1002,17 +1047,18 @@ static int ixgbe_alloc_q_vectors(struct ixgbe_adapter *adapter)
 	int q_vectors = adapter->num_q_vectors;
 	int rxr_remaining = adapter->num_rx_queues;
 	int txr_remaining = adapter->num_tx_queues;
-	int rxr_idx = 0, txr_idx = 0, v_idx = 0;
+	int xdp_remaining = adapter->num_xdp_queues;
+	int rxr_idx = 0, txr_idx = 0, xdp_idx = 0, v_idx = 0;
 	int err;
 
 	/* only one q_vector if MSI-X is disabled. */
 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED))
 		q_vectors = 1;
 
-	if (q_vectors >= (rxr_remaining + txr_remaining)) {
+	if (q_vectors >= (rxr_remaining + txr_remaining + xdp_remaining)) {
 		for (; rxr_remaining; v_idx++) {
 			err = ixgbe_alloc_q_vector(adapter, q_vectors, v_idx,
-						   0, 0, 1, rxr_idx);
+						   0, 0, 0, 0, 1, rxr_idx);
 
 			if (err)
 				goto err_out;
@@ -1026,8 +1072,11 @@ static int ixgbe_alloc_q_vectors(struct ixgbe_adapter *adapter)
 	for (; v_idx < q_vectors; v_idx++) {
 		int rqpv = DIV_ROUND_UP(rxr_remaining, q_vectors - v_idx);
 		int tqpv = DIV_ROUND_UP(txr_remaining, q_vectors - v_idx);
+		int xqpv = DIV_ROUND_UP(xdp_remaining, q_vectors - v_idx);
+
 		err = ixgbe_alloc_q_vector(adapter, q_vectors, v_idx,
 					   tqpv, txr_idx,
+					   xqpv, xdp_idx,
 					   rqpv, rxr_idx);
 
 		if (err)
@@ -1036,14 +1085,17 @@ static int ixgbe_alloc_q_vectors(struct ixgbe_adapter *adapter)
 		/* update counts and index */
 		rxr_remaining -= rqpv;
 		txr_remaining -= tqpv;
+		xdp_remaining -= xqpv;
 		rxr_idx++;
 		txr_idx++;
+		xdp_idx += xqpv;
 	}
 
 	return 0;
 
 err_out:
 	adapter->num_tx_queues = 0;
+	adapter->num_xdp_queues = 0;
 	adapter->num_rx_queues = 0;
 	adapter->num_q_vectors = 0;
 
@@ -1066,6 +1118,7 @@ static void ixgbe_free_q_vectors(struct ixgbe_adapter *adapter)
 	int v_idx = adapter->num_q_vectors;
 
 	adapter->num_tx_queues = 0;
+	adapter->num_xdp_queues = 0;
 	adapter->num_rx_queues = 0;
 	adapter->num_q_vectors = 0;
 
@@ -1172,9 +1225,10 @@ int ixgbe_init_interrupt_scheme(struct ixgbe_adapter *adapter)
 
 	ixgbe_cache_ring_register(adapter);
 
-	e_dev_info("Multiqueue %s: Rx Queue count = %u, Tx Queue count = %u\n",
+	e_dev_info("Multiqueue %s: Rx Queue count = %u, Tx Queue count = %u XDP Queue count = %u\n",
 		   (adapter->num_rx_queues > 1) ? "Enabled" : "Disabled",
-		   adapter->num_rx_queues, adapter->num_tx_queues);
+		   adapter->num_rx_queues, adapter->num_tx_queues,
+		   adapter->num_xdp_queues);
 
 	set_bit(__IXGBE_DOWN, &adapter->state);
 
@@ -1195,6 +1249,7 @@ int ixgbe_init_interrupt_scheme(struct ixgbe_adapter *adapter)
 void ixgbe_clear_interrupt_scheme(struct ixgbe_adapter *adapter)
 {
 	adapter->num_tx_queues = 0;
+	adapter->num_xdp_queues = 0;
 	adapter->num_rx_queues = 0;
 
 	ixgbe_free_q_vectors(adapter);
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index ebcf9c0..a13d2bf 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -594,6 +594,19 @@ static void ixgbe_regdump(struct ixgbe_hw *hw, struct ixgbe_reg_info *reginfo)
 
 }
 
+static void ixgbe_print_buffer(struct ixgbe_ring *ring, int n)
+{
+	struct ixgbe_tx_buffer *tx_buffer;
+
+	tx_buffer = &ring->tx_buffer_info[ring->next_to_clean];
+	pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
+		n, ring->next_to_use, ring->next_to_clean,
+		(u64)dma_unmap_addr(tx_buffer, dma),
+		dma_unmap_len(tx_buffer, len),
+		tx_buffer->next_to_watch,
+		(u64)tx_buffer->time_stamp);
+}
+
 /*
  * ixgbe_dump - Print registers, tx-rings and rx-rings
  */
@@ -603,7 +616,7 @@ static void ixgbe_dump(struct ixgbe_adapter *adapter)
 	struct ixgbe_hw *hw = &adapter->hw;
 	struct ixgbe_reg_info *reginfo;
 	int n = 0;
-	struct ixgbe_ring *tx_ring;
+	struct ixgbe_ring *ring;
 	struct ixgbe_tx_buffer *tx_buffer;
 	union ixgbe_adv_tx_desc *tx_desc;
 	struct my_u0 { u64 a; u64 b; } *u0;
@@ -643,14 +656,13 @@ static void ixgbe_dump(struct ixgbe_adapter *adapter)
 		"Queue [NTU] [NTC] [bi(ntc)->dma  ]",
 		"leng", "ntw", "timestamp");
 	for (n = 0; n < adapter->num_tx_queues; n++) {
-		tx_ring = adapter->tx_ring[n];
-		tx_buffer = &tx_ring->tx_buffer_info[tx_ring->next_to_clean];
-		pr_info(" %5d %5X %5X %016llX %08X %p %016llX\n",
-			   n, tx_ring->next_to_use, tx_ring->next_to_clean,
-			   (u64)dma_unmap_addr(tx_buffer, dma),
-			   dma_unmap_len(tx_buffer, len),
-			   tx_buffer->next_to_watch,
-			   (u64)tx_buffer->time_stamp);
+		ring = adapter->tx_ring[n];
+		ixgbe_print_buffer(ring, n);
+	}
+
+	for (n = 0; n < adapter->num_xdp_queues; n++) {
+		ring = adapter->xdp_ring[n];
+		ixgbe_print_buffer(ring, n);
 	}
 
 	/* Print TX Rings */
@@ -695,28 +707,28 @@ static void ixgbe_dump(struct ixgbe_adapter *adapter)
 	 */
 
 	for (n = 0; n < adapter->num_tx_queues; n++) {
-		tx_ring = adapter->tx_ring[n];
+		ring = adapter->tx_ring[n];
 		pr_info("------------------------------------\n");
-		pr_info("TX QUEUE INDEX = %d\n", tx_ring->queue_index);
+		pr_info("TX QUEUE INDEX = %d\n", ring->queue_index);
 		pr_info("------------------------------------\n");
 		pr_info("%s%s    %s              %s        %s          %s\n",
 			"T [desc]     [address 63:0  ] ",
 			"[PlPOIdStDDt Ln] [bi->dma       ] ",
 			"leng", "ntw", "timestamp", "bi->skb");
 
-		for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
-			tx_desc = IXGBE_TX_DESC(tx_ring, i);
-			tx_buffer = &tx_ring->tx_buffer_info[i];
+		for (i = 0; ring->desc && (i < ring->count); i++) {
+			tx_desc = IXGBE_TX_DESC(ring, i);
+			tx_buffer = &ring->tx_buffer_info[i];
 			u0 = (struct my_u0 *)tx_desc;
 			if (dma_unmap_len(tx_buffer, len) > 0) {
 				const char *ring_desc;
 
-				if (i == tx_ring->next_to_use &&
-				    i == tx_ring->next_to_clean)
+				if (i == ring->next_to_use &&
+				    i == ring->next_to_clean)
 					ring_desc = " NTC/U";
-				else if (i == tx_ring->next_to_use)
+				else if (i == ring->next_to_use)
 					ring_desc = " NTU";
-				else if (i == tx_ring->next_to_clean)
+				else if (i == ring->next_to_clean)
 					ring_desc = " NTC";
 				else
 					ring_desc = "";
@@ -985,6 +997,10 @@ static void ixgbe_update_xoff_rx_lfc(struct ixgbe_adapter *adapter)
 	for (i = 0; i < adapter->num_tx_queues; i++)
 		clear_bit(__IXGBE_HANG_CHECK_ARMED,
 			  &adapter->tx_ring[i]->state);
+
+	for (i = 0; i < adapter->num_xdp_queues; i++)
+		clear_bit(__IXGBE_HANG_CHECK_ARMED,
+			  &adapter->xdp_ring[i]->state);
 }
 
 static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
@@ -1029,6 +1045,14 @@ static void ixgbe_update_xoff_received(struct ixgbe_adapter *adapter)
 		if (xoff[tc])
 			clear_bit(__IXGBE_HANG_CHECK_ARMED, &tx_ring->state);
 	}
+
+	for (i = 0; i < adapter->num_xdp_queues; i++) {
+		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
+
+		tc = xdp_ring->dcb_tc;
+		if (xoff[tc])
+			clear_bit(__IXGBE_HANG_CHECK_ARMED, &xdp_ring->state);
+	}
 }
 
 static u64 ixgbe_get_tx_completed(struct ixgbe_ring *ring)
@@ -1180,7 +1204,10 @@ static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
 		total_packets += tx_buffer->gso_segs;
 
 		/* free the skb */
-		napi_consume_skb(tx_buffer->skb, napi_budget);
+		if (ring_is_xdp(tx_ring))
+			page_frag_free(tx_buffer->data);
+		else
+			napi_consume_skb(tx_buffer->skb, napi_budget);
 
 		/* unmap skb header data */
 		dma_unmap_single(tx_ring->dev,
@@ -1241,7 +1268,7 @@ static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
 	if (check_for_tx_hang(tx_ring) && ixgbe_check_tx_hang(tx_ring)) {
 		/* schedule immediate reset if we believe we hung */
 		struct ixgbe_hw *hw = &adapter->hw;
-		e_err(drv, "Detected Tx Unit Hang\n"
+		e_err(drv, "Detected Tx Unit Hang %s\n"
 			"  Tx Queue             <%d>\n"
 			"  TDH, TDT             <%x>, <%x>\n"
 			"  next_to_use          <%x>\n"
@@ -1249,13 +1276,16 @@ static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
 			"tx_buffer_info[next_to_clean]\n"
 			"  time_stamp           <%lx>\n"
 			"  jiffies              <%lx>\n",
+			ring_is_xdp(tx_ring) ? "(XDP)" : "",
 			tx_ring->queue_index,
 			IXGBE_READ_REG(hw, IXGBE_TDH(tx_ring->reg_idx)),
 			IXGBE_READ_REG(hw, IXGBE_TDT(tx_ring->reg_idx)),
 			tx_ring->next_to_use, i,
 			tx_ring->tx_buffer_info[i].time_stamp, jiffies);
 
-		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
+		if (!ring_is_xdp(tx_ring))
+			netif_stop_subqueue(tx_ring->netdev,
+					    tx_ring->queue_index);
 
 		e_info(probe,
 		       "tx hang %d detected on queue %d, resetting adapter\n",
@@ -1268,6 +1298,9 @@ static bool ixgbe_clean_tx_irq(struct ixgbe_q_vector *q_vector,
 		return true;
 	}
 
+	if (ring_is_xdp(tx_ring))
+		return !!budget;
+
 	netdev_tx_completed_queue(txring_txq(tx_ring),
 				  total_packets, total_bytes);
 
@@ -2170,8 +2203,13 @@ static struct sk_buff *ixgbe_build_skb(struct ixgbe_ring *rx_ring,
 
 #define IXGBE_XDP_PASS 0
 #define IXGBE_XDP_CONSUMED 1
+#define IXGBE_XDP_TX 2
 
-static struct sk_buff *ixgbe_run_xdp(struct ixgbe_ring  *rx_ring,
+static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
+			       struct xdp_buff *xdp);
+
+static struct sk_buff *ixgbe_run_xdp(struct ixgbe_adapter *adapter,
+				     struct ixgbe_ring *rx_ring,
 				     struct xdp_buff *xdp)
 {
 	int result = IXGBE_XDP_PASS;
@@ -2188,9 +2226,11 @@ static struct sk_buff *ixgbe_run_xdp(struct ixgbe_ring  *rx_ring,
 	switch (act) {
 	case XDP_PASS:
 		break;
+	case XDP_TX:
+		result = ixgbe_xmit_xdp_ring(adapter, xdp);
+		break;
 	default:
 		bpf_warn_invalid_xdp_action(act);
-	case XDP_TX:
 	case XDP_ABORTED:
 		trace_xdp_exception(rx_ring->netdev, xdp_prog, act);
 		/* fallthrough -- handle aborts by dropping packet */
@@ -2203,6 +2243,23 @@ static struct sk_buff *ixgbe_run_xdp(struct ixgbe_ring  *rx_ring,
 	return ERR_PTR(-result);
 }
 
+static void ixgbe_rx_buffer_flip(struct ixgbe_ring *rx_ring,
+				 struct ixgbe_rx_buffer *rx_buffer,
+				 unsigned int size)
+{
+#if (PAGE_SIZE < 8192)
+	unsigned int truesize = ixgbe_rx_pg_size(rx_ring) / 2;
+
+	rx_buffer->page_offset ^= truesize;
+#else
+	unsigned int truesize = ring_uses_build_skb(rx_ring) ?
+				SKB_DATA_ALIGN(IXGBE_SKB_PAD + size) :
+				SKB_DATA_ALIGN(size);
+
+	rx_buffer->page_offset += truesize;
+#endif
+}
+
 /**
  * ixgbe_clean_rx_irq - Clean completed descriptors from Rx ring - bounce buf
  * @q_vector: structure containing interrupt and ring information
@@ -2221,8 +2278,8 @@ static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
 			       const int budget)
 {
 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
-#ifdef IXGBE_FCOE
 	struct ixgbe_adapter *adapter = q_vector->adapter;
+#ifdef IXGBE_FCOE
 	int ddp_bytes;
 	unsigned int mss = 0;
 #endif /* IXGBE_FCOE */
@@ -2262,13 +2319,16 @@ static int ixgbe_clean_rx_irq(struct ixgbe_q_vector *q_vector,
 						ixgbe_rx_offset(rx_ring);
 			xdp.data_end = xdp.data + size;
 
-			skb = ixgbe_run_xdp(rx_ring, &xdp);
+			skb = ixgbe_run_xdp(adapter, rx_ring, &xdp);
 		}
 
 		if (IS_ERR(skb)) {
+			if (PTR_ERR(skb) == -IXGBE_XDP_TX)
+				ixgbe_rx_buffer_flip(rx_ring, rx_buffer, size);
+			else
+				rx_buffer->pagecnt_bias++;
 			total_rx_packets++;
 			total_rx_bytes += size;
-			rx_buffer->pagecnt_bias++;
 		} else if (skb) {
 			ixgbe_add_rx_frag(rx_ring, rx_buffer, skb, size);
 		} else if (ring_uses_build_skb(rx_ring)) {
@@ -3438,6 +3498,8 @@ static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
 	/* Setup the HW Tx Head and Tail descriptor pointers */
 	for (i = 0; i < adapter->num_tx_queues; i++)
 		ixgbe_configure_tx_ring(adapter, adapter->tx_ring[i]);
+	for (i = 0; i < adapter->num_xdp_queues; i++)
+		ixgbe_configure_tx_ring(adapter, adapter->xdp_ring[i]);
 }
 
 static void ixgbe_enable_rx_drop(struct ixgbe_adapter *adapter,
@@ -5579,7 +5641,10 @@ static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
 		union ixgbe_adv_tx_desc *eop_desc, *tx_desc;
 
 		/* Free all the Tx ring sk_buffs */
-		dev_kfree_skb_any(tx_buffer->skb);
+		if (ring_is_xdp(tx_ring))
+			page_frag_free(tx_buffer->data);
+		else
+			dev_kfree_skb_any(tx_buffer->skb);
 
 		/* unmap skb header data */
 		dma_unmap_single(tx_ring->dev,
@@ -5620,7 +5685,8 @@ static void ixgbe_clean_tx_ring(struct ixgbe_ring *tx_ring)
 	}
 
 	/* reset BQL for queue */
-	netdev_tx_reset_queue(txring_txq(tx_ring));
+	if (!ring_is_xdp(tx_ring))
+		netdev_tx_reset_queue(txring_txq(tx_ring));
 
 	/* reset next_to_use and next_to_clean */
 	tx_ring->next_to_use = 0;
@@ -5649,6 +5715,8 @@ static void ixgbe_clean_all_tx_rings(struct ixgbe_adapter *adapter)
 
 	for (i = 0; i < adapter->num_tx_queues; i++)
 		ixgbe_clean_tx_ring(adapter->tx_ring[i]);
+	for (i = 0; i < adapter->num_xdp_queues; i++)
+		ixgbe_clean_tx_ring(adapter->xdp_ring[i]);
 }
 
 static void ixgbe_fdir_filter_exit(struct ixgbe_adapter *adapter)
@@ -5743,6 +5811,11 @@ void ixgbe_down(struct ixgbe_adapter *adapter)
 		u8 reg_idx = adapter->tx_ring[i]->reg_idx;
 		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
 	}
+	for (i = 0; i < adapter->num_xdp_queues; i++) {
+		u8 reg_idx = adapter->xdp_ring[i]->reg_idx;
+
+		IXGBE_WRITE_REG(hw, IXGBE_TXDCTL(reg_idx), IXGBE_TXDCTL_SWFLSH);
+	}
 
 	/* Disable the Tx DMA engine on 82599 and later MAC */
 	switch (hw->mac.type) {
@@ -6111,7 +6184,7 @@ int ixgbe_setup_tx_resources(struct ixgbe_ring *tx_ring)
  **/
 static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
 {
-	int i, err = 0;
+	int i, j = 0, err = 0;
 
 	for (i = 0; i < adapter->num_tx_queues; i++) {
 		err = ixgbe_setup_tx_resources(adapter->tx_ring[i]);
@@ -6121,10 +6194,20 @@ static int ixgbe_setup_all_tx_resources(struct ixgbe_adapter *adapter)
 		e_err(probe, "Allocation for Tx Queue %u failed\n", i);
 		goto err_setup_tx;
 	}
+	for (j = 0; j < adapter->num_xdp_queues; j++) {
+		err = ixgbe_setup_tx_resources(adapter->xdp_ring[j]);
+		if (!err)
+			continue;
+
+		e_err(probe, "Allocation for Tx Queue %u failed\n", j);
+		goto err_setup_tx;
+	}
 
 	return 0;
 err_setup_tx:
 	/* rewind the index freeing the rings as we go */
+	while (j--)
+		ixgbe_free_tx_resources(adapter->xdp_ring[j]);
 	while (i--)
 		ixgbe_free_tx_resources(adapter->tx_ring[i]);
 	return err;
@@ -6255,6 +6338,9 @@ static void ixgbe_free_all_tx_resources(struct ixgbe_adapter *adapter)
 	for (i = 0; i < adapter->num_tx_queues; i++)
 		if (adapter->tx_ring[i]->desc)
 			ixgbe_free_tx_resources(adapter->tx_ring[i]);
+	for (i = 0; i < adapter->num_xdp_queues; i++)
+		if (adapter->xdp_ring[i]->desc)
+			ixgbe_free_tx_resources(adapter->xdp_ring[i]);
 }
 
 /**
@@ -6674,6 +6760,14 @@ void ixgbe_update_stats(struct ixgbe_adapter *adapter)
 		bytes += tx_ring->stats.bytes;
 		packets += tx_ring->stats.packets;
 	}
+	for (i = 0; i < adapter->num_xdp_queues; i++) {
+		struct ixgbe_ring *xdp_ring = adapter->xdp_ring[i];
+
+		restart_queue += xdp_ring->tx_stats.restart_queue;
+		tx_busy += xdp_ring->tx_stats.tx_busy;
+		bytes += xdp_ring->stats.bytes;
+		packets += xdp_ring->stats.packets;
+	}
 	adapter->restart_queue = restart_queue;
 	adapter->tx_busy = tx_busy;
 	netdev->stats.tx_bytes = bytes;
@@ -6867,6 +6961,9 @@ static void ixgbe_fdir_reinit_subtask(struct ixgbe_adapter *adapter)
 		for (i = 0; i < adapter->num_tx_queues; i++)
 			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
 				&(adapter->tx_ring[i]->state));
+		for (i = 0; i < adapter->num_xdp_queues; i++)
+			set_bit(__IXGBE_TX_FDIR_INIT_DONE,
+				&adapter->xdp_ring[i]->state);
 		/* re-enable flow director interrupts */
 		IXGBE_WRITE_REG(hw, IXGBE_EIMS, IXGBE_EIMS_FLOW_DIR);
 	} else {
@@ -6900,6 +6997,8 @@ static void ixgbe_check_hang_subtask(struct ixgbe_adapter *adapter)
 	if (netif_carrier_ok(adapter->netdev)) {
 		for (i = 0; i < adapter->num_tx_queues; i++)
 			set_check_for_tx_hang(adapter->tx_ring[i]);
+		for (i = 0; i < adapter->num_xdp_queues; i++)
+			set_check_for_tx_hang(adapter->xdp_ring[i]);
 	}
 
 	if (!(adapter->flags & IXGBE_FLAG_MSIX_ENABLED)) {
@@ -7130,6 +7229,13 @@ static bool ixgbe_ring_tx_pending(struct ixgbe_adapter *adapter)
 			return true;
 	}
 
+	for (i = 0; i < adapter->num_xdp_queues; i++) {
+		struct ixgbe_ring *ring = adapter->xdp_ring[i];
+
+		if (ring->next_to_use != ring->next_to_clean)
+			return true;
+	}
+
 	return false;
 }
 
@@ -8090,6 +8196,69 @@ static u16 ixgbe_select_queue(struct net_device *dev, struct sk_buff *skb,
 #endif
 }
 
+static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
+			       struct xdp_buff *xdp)
+{
+	struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
+	struct ixgbe_tx_buffer *tx_buffer;
+	union ixgbe_adv_tx_desc *tx_desc;
+	u32 len, cmd_type;
+	dma_addr_t dma;
+	u16 i;
+
+	len = xdp->data_end - xdp->data;
+
+	if (unlikely(!ixgbe_desc_unused(ring)))
+		return IXGBE_XDP_CONSUMED;
+
+	dma = dma_map_single(ring->dev, xdp->data, len, DMA_TO_DEVICE);
+	if (dma_mapping_error(ring->dev, dma))
+		return IXGBE_XDP_CONSUMED;
+
+	/* record the location of the first descriptor for this packet */
+	tx_buffer = &ring->tx_buffer_info[ring->next_to_use];
+	tx_buffer->bytecount = len;
+	tx_buffer->gso_segs = 1;
+	tx_buffer->protocol = 0;
+
+	i = ring->next_to_use;
+	tx_desc = IXGBE_TX_DESC(ring, i);
+
+	dma_unmap_len_set(tx_buffer, len, len);
+	dma_unmap_addr_set(tx_buffer, dma, dma);
+	tx_buffer->data = xdp->data;
+	tx_desc->read.buffer_addr = cpu_to_le64(dma);
+
+	/* put descriptor type bits */
+	cmd_type = IXGBE_ADVTXD_DTYP_DATA |
+		   IXGBE_ADVTXD_DCMD_DEXT |
+		   IXGBE_ADVTXD_DCMD_IFCS;
+	cmd_type |= len | IXGBE_TXD_CMD;
+	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
+	tx_desc->read.olinfo_status =
+		cpu_to_le32(len << IXGBE_ADVTXD_PAYLEN_SHIFT);
+
+	/* Force memory writes to complete before letting h/w know there
+	 * are new descriptors to fetch.  (Only applicable for weak-ordered
+	 * memory model archs, such as IA-64).
+	 *
+	 * We also need this memory barrier to make certain all of the
+	 * status bits have been updated before next_to_watch is written.
+	 */
+	wmb();
+
+	/* set next_to_watch value indicating a packet is present */
+	i++;
+	if (i == ring->count)
+		i = 0;
+
+	tx_buffer->next_to_watch = tx_desc;
+	ring->next_to_use = i;
+
+	writel(i, ring->tail);
+	return IXGBE_XDP_TX;
+}
+
 netdev_tx_t ixgbe_xmit_frame_ring(struct sk_buff *skb,
 			  struct ixgbe_adapter *adapter,
 			  struct ixgbe_ring *tx_ring)
@@ -8381,6 +8550,23 @@ static void ixgbe_netpoll(struct net_device *netdev)
 
 #endif
 
+static void ixgbe_get_ring_stats64(struct rtnl_link_stats64 *stats,
+				   struct ixgbe_ring *ring)
+{
+	u64 bytes, packets;
+	unsigned int start;
+
+	if (ring) {
+		do {
+			start = u64_stats_fetch_begin_irq(&ring->syncp);
+			packets = ring->stats.packets;
+			bytes   = ring->stats.bytes;
+		} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
+		stats->tx_packets += packets;
+		stats->tx_bytes   += bytes;
+	}
+}
+
 static void ixgbe_get_stats64(struct net_device *netdev,
 			      struct rtnl_link_stats64 *stats)
 {
@@ -8406,18 +8592,13 @@ static void ixgbe_get_stats64(struct net_device *netdev,
 
 	for (i = 0; i < adapter->num_tx_queues; i++) {
 		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->tx_ring[i]);
-		u64 bytes, packets;
-		unsigned int start;
 
-		if (ring) {
-			do {
-				start = u64_stats_fetch_begin_irq(&ring->syncp);
-				packets = ring->stats.packets;
-				bytes   = ring->stats.bytes;
-			} while (u64_stats_fetch_retry_irq(&ring->syncp, start));
-			stats->tx_packets += packets;
-			stats->tx_bytes   += bytes;
-		}
+		ixgbe_get_ring_stats64(stats, ring);
+	}
+	for (i = 0; i < adapter->num_xdp_queues; i++) {
+		struct ixgbe_ring *ring = ACCESS_ONCE(adapter->xdp_ring[i]);
+
+		ixgbe_get_ring_stats64(stats, ring);
 	}
 	rcu_read_unlock();
 
@@ -9559,9 +9740,23 @@ static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
 			return -EINVAL;
 	}
 
+	if (nr_cpu_ids > MAX_XDP_QUEUES)
+		return -ENOMEM;
+
 	old_prog = xchg(&adapter->xdp_prog, prog);
-	for (i = 0; i < adapter->num_rx_queues; i++)
-		xchg(&adapter->rx_ring[i]->xdp_prog, adapter->xdp_prog);
+
+	/* If transitioning XDP modes reconfigure rings */
+	if (!!prog != !!old_prog) {
+		int err = ixgbe_setup_tc(dev, netdev_get_num_tc(dev));
+
+		if (err) {
+			rcu_assign_pointer(adapter->xdp_prog, old_prog);
+			return -EINVAL;
+		}
+	} else {
+		for (i = 0; i < adapter->num_rx_queues; i++)
+			xchg(&adapter->rx_ring[i]->xdp_prog, adapter->xdp_prog);
+	}
 
 	if (old_prog)
 		bpf_prog_put(old_prog);
@@ -10070,6 +10265,9 @@ static int ixgbe_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 	for (i = 0; i < adapter->num_tx_queues; i++)
 		u64_stats_init(&adapter->tx_ring[i]->syncp);
 
+	for (i = 0; i < adapter->num_xdp_queues; i++)
+		u64_stats_init(&adapter->xdp_ring[i]->syncp);
+
 	/* WOL not supported for all devices */
 	adapter->wol = 0;
 	hw->eeprom.ops.read(hw, 0x2c, &adapter->eeprom_cap);

Re: [PATCH 1/2] ixgbe: add XDP support for pass and drop actions

[Jakub Kicinski]

Hi!

On Sun, 23 Apr 2017 18:31:19 -0700, John Fastabend wrote:
> +static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
> +{
> +	int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
> +	struct ixgbe_adapter *adapter = netdev_priv(dev);
> +	struct bpf_prog *old_prog;
> +
> +	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
> +		return -EINVAL;
> +
> +	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
> +		return -EINVAL;
> +
> +	/* verify ixgbe ring attributes are sufficient for XDP */
> +	for (i = 0; i < adapter->num_rx_queues; i++) {
> +		struct ixgbe_ring *ring = adapter->rx_ring[i];
> +
> +		if (ring_is_rsc_enabled(ring))
> +			return -EINVAL;
> +
> +		if (frame_size > ixgbe_rx_bufsz(ring))
> +			return -EINVAL;
> +	}

I was just looking through the drivers, working on extended ack
reporting, trying to bring out the driver XDP error messages out
directly to iproute2.  It seems that multiple drivers are only 
checking that MTU/buffer size is appropriate in the XDP_SETUP 
function, and ignore XDP in case user tries to change MTU later.  
And I think it's the same story with LRO?

Re: [PATCH 1/2] ixgbe: add XDP support for pass and drop actions

[John Fastabend]

On 17-04-23 09:05 PM, Jakub Kicinski wrote:
> Hi!
> 
> On Sun, 23 Apr 2017 18:31:19 -0700, John Fastabend wrote:
>> +static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
>> +{
>> +	int i, frame_size = dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
>> +	struct ixgbe_adapter *adapter = netdev_priv(dev);
>> +	struct bpf_prog *old_prog;
>> +
>> +	if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
>> +		return -EINVAL;
>> +
>> +	if (adapter->flags & IXGBE_FLAG_DCB_ENABLED)
>> +		return -EINVAL;
>> +
>> +	/* verify ixgbe ring attributes are sufficient for XDP */
>> +	for (i = 0; i < adapter->num_rx_queues; i++) {
>> +		struct ixgbe_ring *ring = adapter->rx_ring[i];
>> +
>> +		if (ring_is_rsc_enabled(ring))
>> +			return -EINVAL;
>> +
>> +		if (frame_size > ixgbe_rx_bufsz(ring))
>> +			return -EINVAL;
>> +	}
> 
> I was just looking through the drivers, working on extended ack
> reporting, trying to bring out the driver XDP error messages out
> directly to iproute2.  It seems that multiple drivers are only 
> checking that MTU/buffer size is appropriate in the XDP_SETUP 
> function, and ignore XDP in case user tries to change MTU later.  
> And I think it's the same story with LRO?
> 

Agreed we need to harden the drivers to changes post XDP init. I'll submit
a few follow on patches for the ixgbe devices in the morning.

Thanks,
John

Re: [PATCH 2/2] ixgbe: add support for XDP_TX action

[Jesper Dangaard Brouer]

On Sun, 23 Apr 2017 18:31:36 -0700
John Fastabend <john.fastabend@gmail.com> wrote:

> +static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
> +			       struct xdp_buff *xdp)
> +{
> +	struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];

I was about to question whether is it always true that the array size
can match the number of CPUs in the system, but I can see later in
ixgbe_xdp_setup() that you reject XDP program if the system have more
CPUs that MAX_XDP_QUEUES.

> +	struct ixgbe_tx_buffer *tx_buffer;
> +	union ixgbe_adv_tx_desc *tx_desc;
> +	u32 len, cmd_type;
> +	dma_addr_t dma;
> +	u16 i;
> +
> +	len = xdp->data_end - xdp->data;
> +
> +	if (unlikely(!ixgbe_desc_unused(ring)))
> +		return IXGBE_XDP_CONSUMED;
> +
> +	dma = dma_map_single(ring->dev, xdp->data, len, DMA_TO_DEVICE);
> +	if (dma_mapping_error(ring->dev, dma))
> +		return IXGBE_XDP_CONSUMED;
> +
> +	/* record the location of the first descriptor for this packet */
> +	tx_buffer = &ring->tx_buffer_info[ring->next_to_use];
> +	tx_buffer->bytecount = len;
> +	tx_buffer->gso_segs = 1;
> +	tx_buffer->protocol = 0;
> +
> +	i = ring->next_to_use;
> +	tx_desc = IXGBE_TX_DESC(ring, i);
> +
> +	dma_unmap_len_set(tx_buffer, len, len);
> +	dma_unmap_addr_set(tx_buffer, dma, dma);
> +	tx_buffer->data = xdp->data;
> +	tx_desc->read.buffer_addr = cpu_to_le64(dma);
> +
> +	/* put descriptor type bits */
> +	cmd_type = IXGBE_ADVTXD_DTYP_DATA |
> +		   IXGBE_ADVTXD_DCMD_DEXT |
> +		   IXGBE_ADVTXD_DCMD_IFCS;
> +	cmd_type |= len | IXGBE_TXD_CMD;
> +	tx_desc->read.cmd_type_len = cpu_to_le32(cmd_type);
> +	tx_desc->read.olinfo_status =
> +		cpu_to_le32(len << IXGBE_ADVTXD_PAYLEN_SHIFT);
> +
> +	/* Force memory writes to complete before letting h/w know there
> +	 * are new descriptors to fetch.  (Only applicable for weak-ordered
> +	 * memory model archs, such as IA-64).
> +	 *
> +	 * We also need this memory barrier to make certain all of the
> +	 * status bits have been updated before next_to_watch is written.
> +	 */
> +	wmb();
> +
> +	/* set next_to_watch value indicating a packet is present */
> +	i++;
> +	if (i == ring->count)
> +		i = 0;
> +
> +	tx_buffer->next_to_watch = tx_desc;
> +	ring->next_to_use = i;
> +
> +	writel(i, ring->tail);

A tailptr write for every XDP_TX packet is not going be fast, but you
already mentioned that this is not optimal yet, so I guess you are aware.

> +	return IXGBE_XDP_TX;
> +}


On Sun, 23 Apr 2017 18:31:36 -0700
John Fastabend <john.fastabend@gmail.com> wrote:

> @@ -9559,9 +9740,23 @@ static int ixgbe_xdp_setup(struct net_device *dev, struct bpf_prog *prog)
>  			return -EINVAL;
>  	}
>  
> +	if (nr_cpu_ids > MAX_XDP_QUEUES)
> +		return -ENOMEM;
> +


-- 
Best regards,
  Jesper Dangaard Brouer
  MSc.CS, Principal Kernel Engineer at Red Hat
  LinkedIn: http://www.linkedin.com/in/brouer

Re: [PATCH 2/2] ixgbe: add support for XDP_TX action

[John Fastabend]

On 17-04-24 12:29 PM, Jesper Dangaard Brouer wrote:
> On Sun, 23 Apr 2017 18:31:36 -0700
> John Fastabend <john.fastabend@gmail.com> wrote:
> 
>> +static int ixgbe_xmit_xdp_ring(struct ixgbe_adapter *adapter,
>> +			       struct xdp_buff *xdp)
>> +{
>> +	struct ixgbe_ring *ring = adapter->xdp_ring[smp_processor_id()];
> 
> I was about to question whether is it always true that the array size
> can match the number of CPUs in the system, but I can see later in
> ixgbe_xdp_setup() that you reject XDP program if the system have more
> CPUs that MAX_XDP_QUEUES.

Yep.

[...]

>> +
>> +	tx_buffer->next_to_watch = tx_desc;
>> +	ring->next_to_use = i;
>> +
>> +	writel(i, ring->tail);
> 
> A tailptr write for every XDP_TX packet is not going be fast, but you
> already mentioned that this is not optimal yet, so I guess you are aware.
> 

There is another patch on Jeff's tree to only kick the tail ptr once per
receive path invocation.

https://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue.git/commit/?h=dev-queue&id=24043a662d11e048de903e12bf86059844c207e2

That patch brings packet rates up to near line rate @ 64 bytes.

Thanks,
John