Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next

Pull networking updates from David Miller:

 1) New ipset extensions for matching on destination MAC addresses, from
    Stefano Brivio.

 2) Add ipv4 ttl and tos, plus ipv6 flow label and hop limit offloads to
    nfp driver. From Stefano Brivio.

 3) Implement GRO for plain UDP sockets, from Paolo Abeni.

 4) Lots of work from Michał Mirosław to eliminate the VLAN_TAG_PRESENT
    bit so that we could support the entire vlan_tci value.

 5) Rework the IPSEC policy lookups to better optimize more usecases,
    from Florian Westphal.

 6) Infrastructure changes eliminating direct manipulation of SKB lists
    wherever possible, and to always use the appropriate SKB list
    helpers. This work is still ongoing...

 7) Lots of PHY driver and state machine improvements and
    simplifications, from Heiner Kallweit.

 8) Various TSO deferral refinements, from Eric Dumazet.

 9) Add ntuple filter support to aquantia driver, from Dmitry Bogdanov.

10) Batch dropping of XDP packets in tuntap, from Jason Wang.

11) Lots of cleanups and improvements to the r8169 driver from Heiner
    Kallweit, including support for ->xmit_more. This driver has been
    getting some much needed love since he started working on it.

12) Lots of new forwarding selftests from Petr Machata.

13) Enable VXLAN learning in mlxsw driver, from Ido Schimmel.

14) Packed ring support for virtio, from Tiwei Bie.

15) Add new Aquantia AQtion USB driver, from Dmitry Bezrukov.

16) Add XDP support to dpaa2-eth driver, from Ioana Ciocoi Radulescu.

17) Implement coalescing on TCP backlog queue, from Eric Dumazet.

18) Implement carrier change in tun driver, from Nicolas Dichtel.

19) Support msg_zerocopy in UDP, from Willem de Bruijn.

20) Significantly improve garbage collection of neighbor objects when
    the table has many PERMANENT entries, from David Ahern.

21) Remove egdev usage from nfp and mlx5, and remove the facility
    completely from the tree as it no longer has any users. From Oz
    Shlomo and others.

22) Add a NETDEV_PRE_CHANGEADDR so that drivers can veto the change and
    therefore abort the operation before the commit phase (which is the
    NETDEV_CHANGEADDR event). From Petr Machata.

23) Add indirect call wrappers to avoid retpoline overhead, and use them
    in the GRO code paths. From Paolo Abeni.

24) Add support for netlink FDB get operations, from Roopa Prabhu.

25) Support bloom filter in mlxsw driver, from Nir Dotan.

26) Add SKB extension infrastructure. This consolidates the handling of
    the auxiliary SKB data used by IPSEC and bridge netfilter, and is
    designed to support the needs to MPTCP which could be integrated in
    the future.

27) Lots of XDP TX optimizations in mlx5 from Tariq Toukan.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (1845 commits)
  net: dccp: fix kernel crash on module load
  drivers/net: appletalk/cops: remove redundant if statement and mask
  bnx2x: Fix NULL pointer dereference in bnx2x_del_all_vlans() on some hw
  net/net_namespace: Check the return value of register_pernet_subsys()
  net/netlink_compat: Fix a missing check of nla_parse_nested
  ieee802154: lowpan_header_create check must check daddr
  net/mlx4_core: drop useless LIST_HEAD
  mlxsw: spectrum: drop useless LIST_HEAD
  net/mlx5e: drop useless LIST_HEAD
  iptunnel: Set tun_flags in the iptunnel_metadata_reply from src
  net/mlx5e: fix semicolon.cocci warnings
  staging: octeon: fix build failure with XFRM enabled
  net: Revert recent Spectre-v1 patches.
  can: af_can: Fix Spectre v1 vulnerability
  packet: validate address length if non-zero
  nfc: af_nfc: Fix Spectre v1 vulnerability
  phonet: af_phonet: Fix Spectre v1 vulnerability
  net: core: Fix Spectre v1 vulnerability
  net: minor cleanup in skb_ext_add()
  net: drop the unused helper skb_ext_get()
  ...

1 files changed, 399 insertions, 55 deletions

diff --git a/drivers/net/ethernet/freescale/dpaa2/dpaa2-eth.c b/drivers/net/ethernet/freescale/dpaa2/dpaa2-eth.c
index 88f7acce38dc..1ca9a18139ec 100644
--- a/drivers/net/ethernet/freescale/dpaa2/dpaa2-eth.c
+++ b/drivers/net/ethernet/freescale/dpaa2/dpaa2-eth.c
@@ -13,7 +13,8 @@
 #include <linux/iommu.h>
 #include <linux/net_tstamp.h>
 #include <linux/fsl/mc.h>
-
+#include <linux/bpf.h>
+#include <linux/bpf_trace.h>
 #include <net/sock.h>
 
 #include "dpaa2-eth.h"
@@ -86,7 +87,7 @@ static void free_rx_fd(struct dpaa2_eth_priv *priv,
 		addr = dpaa2_sg_get_addr(&sgt[i]);
 		sg_vaddr = dpaa2_iova_to_virt(priv->iommu_domain, addr);
 		dma_unmap_single(dev, addr, DPAA2_ETH_RX_BUF_SIZE,
-				 DMA_FROM_DEVICE);
+				 DMA_BIDIRECTIONAL);
 
 		skb_free_frag(sg_vaddr);
 		if (dpaa2_sg_is_final(&sgt[i]))
@@ -144,7 +145,7 @@ static struct sk_buff *build_frag_skb(struct dpaa2_eth_priv *priv,
 		sg_addr = dpaa2_sg_get_addr(sge);
 		sg_vaddr = dpaa2_iova_to_virt(priv->iommu_domain, sg_addr);
 		dma_unmap_single(dev, sg_addr, DPAA2_ETH_RX_BUF_SIZE,
-				 DMA_FROM_DEVICE);
+				 DMA_BIDIRECTIONAL);
 
 		sg_length = dpaa2_sg_get_len(sge);
 
@@ -199,12 +200,148 @@ static struct sk_buff *build_frag_skb(struct dpaa2_eth_priv *priv,
 	return skb;
 }
 
+/* Free buffers acquired from the buffer pool or which were meant to
+ * be released in the pool
+ */
+static void free_bufs(struct dpaa2_eth_priv *priv, u64 *buf_array, int count)
+{
+	struct device *dev = priv->net_dev->dev.parent;
+	void *vaddr;
+	int i;
+
+	for (i = 0; i < count; i++) {
+		vaddr = dpaa2_iova_to_virt(priv->iommu_domain, buf_array[i]);
+		dma_unmap_single(dev, buf_array[i], DPAA2_ETH_RX_BUF_SIZE,
+				 DMA_BIDIRECTIONAL);
+		skb_free_frag(vaddr);
+	}
+}
+
+static void xdp_release_buf(struct dpaa2_eth_priv *priv,
+			    struct dpaa2_eth_channel *ch,
+			    dma_addr_t addr)
+{
+	int err;
+
+	ch->xdp.drop_bufs[ch->xdp.drop_cnt++] = addr;
+	if (ch->xdp.drop_cnt < DPAA2_ETH_BUFS_PER_CMD)
+		return;
+
+	while ((err = dpaa2_io_service_release(ch->dpio, priv->bpid,
+					       ch->xdp.drop_bufs,
+					       ch->xdp.drop_cnt)) == -EBUSY)
+		cpu_relax();
+
+	if (err) {
+		free_bufs(priv, ch->xdp.drop_bufs, ch->xdp.drop_cnt);
+		ch->buf_count -= ch->xdp.drop_cnt;
+	}
+
+	ch->xdp.drop_cnt = 0;
+}
+
+static int xdp_enqueue(struct dpaa2_eth_priv *priv, struct dpaa2_fd *fd,
+		       void *buf_start, u16 queue_id)
+{
+	struct dpaa2_eth_fq *fq;
+	struct dpaa2_faead *faead;
+	u32 ctrl, frc;
+	int i, err;
+
+	/* Mark the egress frame hardware annotation area as valid */
+	frc = dpaa2_fd_get_frc(fd);
+	dpaa2_fd_set_frc(fd, frc | DPAA2_FD_FRC_FAEADV);
+	dpaa2_fd_set_ctrl(fd, DPAA2_FD_CTRL_ASAL);
+
+	/* Instruct hardware to release the FD buffer directly into
+	 * the buffer pool once transmission is completed, instead of
+	 * sending a Tx confirmation frame to us
+	 */
+	ctrl = DPAA2_FAEAD_A4V | DPAA2_FAEAD_A2V | DPAA2_FAEAD_EBDDV;
+	faead = dpaa2_get_faead(buf_start, false);
+	faead->ctrl = cpu_to_le32(ctrl);
+	faead->conf_fqid = 0;
+
+	fq = &priv->fq[queue_id];
+	for (i = 0; i < DPAA2_ETH_ENQUEUE_RETRIES; i++) {
+		err = dpaa2_io_service_enqueue_qd(fq->channel->dpio,
+						  priv->tx_qdid, 0,
+						  fq->tx_qdbin, fd);
+		if (err != -EBUSY)
+			break;
+	}
+
+	return err;
+}
+
+static u32 run_xdp(struct dpaa2_eth_priv *priv,
+		   struct dpaa2_eth_channel *ch,
+		   struct dpaa2_eth_fq *rx_fq,
+		   struct dpaa2_fd *fd, void *vaddr)
+{
+	dma_addr_t addr = dpaa2_fd_get_addr(fd);
+	struct rtnl_link_stats64 *percpu_stats;
+	struct bpf_prog *xdp_prog;
+	struct xdp_buff xdp;
+	u32 xdp_act = XDP_PASS;
+	int err;
+
+	percpu_stats = this_cpu_ptr(priv->percpu_stats);
+
+	rcu_read_lock();
+
+	xdp_prog = READ_ONCE(ch->xdp.prog);
+	if (!xdp_prog)
+		goto out;
+
+	xdp.data = vaddr + dpaa2_fd_get_offset(fd);
+	xdp.data_end = xdp.data + dpaa2_fd_get_len(fd);
+	xdp.data_hard_start = xdp.data - XDP_PACKET_HEADROOM;
+	xdp_set_data_meta_invalid(&xdp);
+
+	xdp_act = bpf_prog_run_xdp(xdp_prog, &xdp);
+
+	/* xdp.data pointer may have changed */
+	dpaa2_fd_set_offset(fd, xdp.data - vaddr);
+	dpaa2_fd_set_len(fd, xdp.data_end - xdp.data);
+
+	switch (xdp_act) {
+	case XDP_PASS:
+		break;
+	case XDP_TX:
+		err = xdp_enqueue(priv, fd, vaddr, rx_fq->flowid);
+		if (err) {
+			xdp_release_buf(priv, ch, addr);
+			percpu_stats->tx_errors++;
+			ch->stats.xdp_tx_err++;
+		} else {
+			percpu_stats->tx_packets++;
+			percpu_stats->tx_bytes += dpaa2_fd_get_len(fd);
+			ch->stats.xdp_tx++;
+		}
+		break;
+	default:
+		bpf_warn_invalid_xdp_action(xdp_act);
+		/* fall through */
+	case XDP_ABORTED:
+		trace_xdp_exception(priv->net_dev, xdp_prog, xdp_act);
+		/* fall through */
+	case XDP_DROP:
+		xdp_release_buf(priv, ch, addr);
+		ch->stats.xdp_drop++;
+		break;
+	}
+
+out:
+	rcu_read_unlock();
+	return xdp_act;
+}
+
 /* Main Rx frame processing routine */
 static void dpaa2_eth_rx(struct dpaa2_eth_priv *priv,
 			 struct dpaa2_eth_channel *ch,
 			 const struct dpaa2_fd *fd,
-			 struct napi_struct *napi,
-			 u16 queue_id)
+			 struct dpaa2_eth_fq *fq)
 {
 	dma_addr_t addr = dpaa2_fd_get_addr(fd);
 	u8 fd_format = dpaa2_fd_get_format(fd);
@@ -216,12 +353,14 @@ static void dpaa2_eth_rx(struct dpaa2_eth_priv *priv,
 	struct dpaa2_fas *fas;
 	void *buf_data;
 	u32 status = 0;
+	u32 xdp_act;
 
 	/* Tracing point */
 	trace_dpaa2_rx_fd(priv->net_dev, fd);
 
 	vaddr = dpaa2_iova_to_virt(priv->iommu_domain, addr);
-	dma_unmap_single(dev, addr, DPAA2_ETH_RX_BUF_SIZE, DMA_FROM_DEVICE);
+	dma_sync_single_for_cpu(dev, addr, DPAA2_ETH_RX_BUF_SIZE,
+				DMA_BIDIRECTIONAL);
 
 	fas = dpaa2_get_fas(vaddr, false);
 	prefetch(fas);
@@ -232,8 +371,21 @@ static void dpaa2_eth_rx(struct dpaa2_eth_priv *priv,
 	percpu_extras = this_cpu_ptr(priv->percpu_extras);
 
 	if (fd_format == dpaa2_fd_single) {
+		xdp_act = run_xdp(priv, ch, fq, (struct dpaa2_fd *)fd, vaddr);
+		if (xdp_act != XDP_PASS) {
+			percpu_stats->rx_packets++;
+			percpu_stats->rx_bytes += dpaa2_fd_get_len(fd);
+			return;
+		}
+
+		dma_unmap_single(dev, addr, DPAA2_ETH_RX_BUF_SIZE,
+				 DMA_BIDIRECTIONAL);
 		skb = build_linear_skb(ch, fd, vaddr);
 	} else if (fd_format == dpaa2_fd_sg) {
+		WARN_ON(priv->xdp_prog);
+
+		dma_unmap_single(dev, addr, DPAA2_ETH_RX_BUF_SIZE,
+				 DMA_BIDIRECTIONAL);
 		skb = build_frag_skb(priv, ch, buf_data);
 		skb_free_frag(vaddr);
 		percpu_extras->rx_sg_frames++;
@@ -267,12 +419,12 @@ static void dpaa2_eth_rx(struct dpaa2_eth_priv *priv,
 	}
 
 	skb->protocol = eth_type_trans(skb, priv->net_dev);
-	skb_record_rx_queue(skb, queue_id);
+	skb_record_rx_queue(skb, fq->flowid);
 
 	percpu_stats->rx_packets++;
 	percpu_stats->rx_bytes += dpaa2_fd_get_len(fd);
 
-	napi_gro_receive(napi, skb);
+	napi_gro_receive(&ch->napi, skb);
 
 	return;
 
@@ -289,7 +441,7 @@ err_frame_format:
  * Observance of NAPI budget is not our concern, leaving that to the caller.
  */
 static int consume_frames(struct dpaa2_eth_channel *ch,
-			  enum dpaa2_eth_fq_type *type)
+			  struct dpaa2_eth_fq **src)
 {
 	struct dpaa2_eth_priv *priv = ch->priv;
 	struct dpaa2_eth_fq *fq = NULL;
@@ -312,7 +464,7 @@ static int consume_frames(struct dpaa2_eth_channel *ch,
 		fd = dpaa2_dq_fd(dq);
 		fq = (struct dpaa2_eth_fq *)(uintptr_t)dpaa2_dq_fqd_ctx(dq);
 
-		fq->consume(priv, ch, fd, &ch->napi, fq->flowid);
+		fq->consume(priv, ch, fd, fq);
 		cleaned++;
 	} while (!is_last);
 
@@ -320,13 +472,12 @@ static int consume_frames(struct dpaa2_eth_channel *ch,
 		return 0;
 
 	fq->stats.frames += cleaned;
-	ch->stats.frames += cleaned;
 
 	/* A dequeue operation only pulls frames from a single queue
-	 * into the store. Return the frame queue type as an out param.
+	 * into the store. Return the frame queue as an out param.
 	 */
-	if (type)
-		*type = fq->type;
+	if (src)
+		*src = fq;
 
 	return cleaned;
 }
@@ -571,8 +722,10 @@ static netdev_tx_t dpaa2_eth_tx(struct sk_buff *skb, struct net_device *net_dev)
 	struct rtnl_link_stats64 *percpu_stats;
 	struct dpaa2_eth_drv_stats *percpu_extras;
 	struct dpaa2_eth_fq *fq;
+	struct netdev_queue *nq;
 	u16 queue_mapping;
 	unsigned int needed_headroom;
+	u32 fd_len;
 	int err, i;
 
 	percpu_stats = this_cpu_ptr(priv->percpu_stats);
@@ -644,8 +797,12 @@ static netdev_tx_t dpaa2_eth_tx(struct sk_buff *skb, struct net_device *net_dev)
 		/* Clean up everything, including freeing the skb */
 		free_tx_fd(priv, &fd);
 	} else {
+		fd_len = dpaa2_fd_get_len(&fd);
 		percpu_stats->tx_packets++;
-		percpu_stats->tx_bytes += dpaa2_fd_get_len(&fd);
+		percpu_stats->tx_bytes += fd_len;
+
+		nq = netdev_get_tx_queue(net_dev, queue_mapping);
+		netdev_tx_sent_queue(nq, fd_len);
 	}
 
 	return NETDEV_TX_OK;
@@ -661,11 +818,11 @@ err_alloc_headroom:
 static void dpaa2_eth_tx_conf(struct dpaa2_eth_priv *priv,
 			      struct dpaa2_eth_channel *ch __always_unused,
 			      const struct dpaa2_fd *fd,
-			      struct napi_struct *napi __always_unused,
-			      u16 queue_id __always_unused)
+			      struct dpaa2_eth_fq *fq)
 {
 	struct rtnl_link_stats64 *percpu_stats;
 	struct dpaa2_eth_drv_stats *percpu_extras;
+	u32 fd_len = dpaa2_fd_get_len(fd);
 	u32 fd_errors;
 
 	/* Tracing point */
@@ -673,7 +830,10 @@ static void dpaa2_eth_tx_conf(struct dpaa2_eth_priv *priv,
 
 	percpu_extras = this_cpu_ptr(priv->percpu_extras);
 	percpu_extras->tx_conf_frames++;
-	percpu_extras->tx_conf_bytes += dpaa2_fd_get_len(fd);
+	percpu_extras->tx_conf_bytes += fd_len;
+
+	fq->dq_frames++;
+	fq->dq_bytes += fd_len;
 
 	/* Check frame errors in the FD field */
 	fd_errors = dpaa2_fd_get_ctrl(fd) & DPAA2_FD_TX_ERR_MASK;
@@ -735,23 +895,6 @@ static int set_tx_csum(struct dpaa2_eth_priv *priv, bool enable)
 	return 0;
 }
 
-/* Free buffers acquired from the buffer pool or which were meant to
- * be released in the pool
- */
-static void free_bufs(struct dpaa2_eth_priv *priv, u64 *buf_array, int count)
-{
-	struct device *dev = priv->net_dev->dev.parent;
-	void *vaddr;
-	int i;
-
-	for (i = 0; i < count; i++) {
-		vaddr = dpaa2_iova_to_virt(priv->iommu_domain, buf_array[i]);
-		dma_unmap_single(dev, buf_array[i], DPAA2_ETH_RX_BUF_SIZE,
-				 DMA_FROM_DEVICE);
-		skb_free_frag(vaddr);
-	}
-}
-
 /* Perform a single release command to add buffers
  * to the specified buffer pool
  */
@@ -775,7 +918,7 @@ static int add_bufs(struct dpaa2_eth_priv *priv,
 		buf = PTR_ALIGN(buf, priv->rx_buf_align);
 
 		addr = dma_map_single(dev, buf, DPAA2_ETH_RX_BUF_SIZE,
-				      DMA_FROM_DEVICE);
+				      DMA_BIDIRECTIONAL);
 		if (unlikely(dma_mapping_error(dev, addr)))
 			goto err_map;
 
@@ -934,8 +1077,9 @@ static int dpaa2_eth_poll(struct napi_struct *napi, int budget)
 	struct dpaa2_eth_channel *ch;
 	struct dpaa2_eth_priv *priv;
 	int rx_cleaned = 0, txconf_cleaned = 0;
-	enum dpaa2_eth_fq_type type = 0;
-	int store_cleaned;
+	struct dpaa2_eth_fq *fq, *txc_fq = NULL;
+	struct netdev_queue *nq;
+	int store_cleaned, work_done;
 	int err;
 
 	ch = container_of(napi, struct dpaa2_eth_channel, napi);
@@ -949,18 +1093,25 @@ static int dpaa2_eth_poll(struct napi_struct *napi, int budget)
 		/* Refill pool if appropriate */
 		refill_pool(priv, ch, priv->bpid);
 
-		store_cleaned = consume_frames(ch, &type);
-		if (type == DPAA2_RX_FQ)
+		store_cleaned = consume_frames(ch, &fq);
+		if (!store_cleaned)
+			break;
+		if (fq->type == DPAA2_RX_FQ) {
 			rx_cleaned += store_cleaned;
-		else
+		} else {
 			txconf_cleaned += store_cleaned;
+			/* We have a single Tx conf FQ on this channel */
+			txc_fq = fq;
+		}
 
 		/* If we either consumed the whole NAPI budget with Rx frames
 		 * or we reached the Tx confirmations threshold, we're done.
 		 */
 		if (rx_cleaned >= budget ||
-		    txconf_cleaned >= DPAA2_ETH_TXCONF_PER_NAPI)
-			return budget;
+		    txconf_cleaned >= DPAA2_ETH_TXCONF_PER_NAPI) {
+			work_done = budget;
+			goto out;
+		}
 	} while (store_cleaned);
 
 	/* We didn't consume the entire budget, so finish napi and
@@ -974,7 +1125,18 @@ static int dpaa2_eth_poll(struct napi_struct *napi, int budget)
 	WARN_ONCE(err, "CDAN notifications rearm failed on core %d",
 		  ch->nctx.desired_cpu);
 
-	return max(rx_cleaned, 1);
+	work_done = max(rx_cleaned, 1);
+
+out:
+	if (txc_fq) {
+		nq = netdev_get_tx_queue(priv->net_dev, txc_fq->flowid);
+		netdev_tx_completed_queue(nq, txc_fq->dq_frames,
+					  txc_fq->dq_bytes);
+		txc_fq->dq_frames = 0;
+		txc_fq->dq_bytes = 0;
+	}
+
+	return work_done;
 }
 
 static void enable_ch_napi(struct dpaa2_eth_priv *priv)
@@ -1400,6 +1562,174 @@ static int dpaa2_eth_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
 	return -EINVAL;
 }
 
+static bool xdp_mtu_valid(struct dpaa2_eth_priv *priv, int mtu)
+{
+	int mfl, linear_mfl;
+
+	mfl = DPAA2_ETH_L2_MAX_FRM(mtu);
+	linear_mfl = DPAA2_ETH_RX_BUF_SIZE - DPAA2_ETH_RX_HWA_SIZE -
+		     dpaa2_eth_rx_head_room(priv) - XDP_PACKET_HEADROOM;
+
+	if (mfl > linear_mfl) {
+		netdev_warn(priv->net_dev, "Maximum MTU for XDP is %d\n",
+			    linear_mfl - VLAN_ETH_HLEN);
+		return false;
+	}
+
+	return true;
+}
+
+static int set_rx_mfl(struct dpaa2_eth_priv *priv, int mtu, bool has_xdp)
+{
+	int mfl, err;
+
+	/* We enforce a maximum Rx frame length based on MTU only if we have
+	 * an XDP program attached (in order to avoid Rx S/G frames).
+	 * Otherwise, we accept all incoming frames as long as they are not
+	 * larger than maximum size supported in hardware
+	 */
+	if (has_xdp)
+		mfl = DPAA2_ETH_L2_MAX_FRM(mtu);
+	else
+		mfl = DPAA2_ETH_MFL;
+
+	err = dpni_set_max_frame_length(priv->mc_io, 0, priv->mc_token, mfl);
+	if (err) {
+		netdev_err(priv->net_dev, "dpni_set_max_frame_length failed\n");
+		return err;
+	}
+
+	return 0;
+}
+
+static int dpaa2_eth_change_mtu(struct net_device *dev, int new_mtu)
+{
+	struct dpaa2_eth_priv *priv = netdev_priv(dev);
+	int err;
+
+	if (!priv->xdp_prog)
+		goto out;
+
+	if (!xdp_mtu_valid(priv, new_mtu))
+		return -EINVAL;
+
+	err = set_rx_mfl(priv, new_mtu, true);
+	if (err)
+		return err;
+
+out:
+	dev->mtu = new_mtu;
+	return 0;
+}
+
+static int update_rx_buffer_headroom(struct dpaa2_eth_priv *priv, bool has_xdp)
+{
+	struct dpni_buffer_layout buf_layout = {0};
+	int err;
+
+	err = dpni_get_buffer_layout(priv->mc_io, 0, priv->mc_token,
+				     DPNI_QUEUE_RX, &buf_layout);
+	if (err) {
+		netdev_err(priv->net_dev, "dpni_get_buffer_layout failed\n");
+		return err;
+	}
+
+	/* Reserve extra headroom for XDP header size changes */
+	buf_layout.data_head_room = dpaa2_eth_rx_head_room(priv) +
+				    (has_xdp ? XDP_PACKET_HEADROOM : 0);
+	buf_layout.options = DPNI_BUF_LAYOUT_OPT_DATA_HEAD_ROOM;
+	err = dpni_set_buffer_layout(priv->mc_io, 0, priv->mc_token,
+				     DPNI_QUEUE_RX, &buf_layout);
+	if (err) {
+		netdev_err(priv->net_dev, "dpni_set_buffer_layout failed\n");
+		return err;
+	}
+
+	return 0;
+}
+
+static int setup_xdp(struct net_device *dev, struct bpf_prog *prog)
+{
+	struct dpaa2_eth_priv *priv = netdev_priv(dev);
+	struct dpaa2_eth_channel *ch;
+	struct bpf_prog *old;
+	bool up, need_update;
+	int i, err;
+
+	if (prog && !xdp_mtu_valid(priv, dev->mtu))
+		return -EINVAL;
+
+	if (prog) {
+		prog = bpf_prog_add(prog, priv->num_channels);
+		if (IS_ERR(prog))
+			return PTR_ERR(prog);
+	}
+
+	up = netif_running(dev);
+	need_update = (!!priv->xdp_prog != !!prog);
+
+	if (up)
+		dpaa2_eth_stop(dev);
+
+	/* While in xdp mode, enforce a maximum Rx frame size based on MTU.
+	 * Also, when switching between xdp/non-xdp modes we need to reconfigure
+	 * our Rx buffer layout. Buffer pool was drained on dpaa2_eth_stop,
+	 * so we are sure no old format buffers will be used from now on.
+	 */
+	if (need_update) {
+		err = set_rx_mfl(priv, dev->mtu, !!prog);
+		if (err)
+			goto out_err;
+		err = update_rx_buffer_headroom(priv, !!prog);
+		if (err)
+			goto out_err;
+	}
+
+	old = xchg(&priv->xdp_prog, prog);
+	if (old)
+		bpf_prog_put(old);
+
+	for (i = 0; i < priv->num_channels; i++) {
+		ch = priv->channel[i];
+		old = xchg(&ch->xdp.prog, prog);
+		if (old)
+			bpf_prog_put(old);
+	}
+
+	if (up) {
+		err = dpaa2_eth_open(dev);
+		if (err)
+			return err;
+	}
+
+	return 0;
+
+out_err:
+	if (prog)
+		bpf_prog_sub(prog, priv->num_channels);
+	if (up)
+		dpaa2_eth_open(dev);
+
+	return err;
+}
+
+static int dpaa2_eth_xdp(struct net_device *dev, struct netdev_bpf *xdp)
+{
+	struct dpaa2_eth_priv *priv = netdev_priv(dev);
+
+	switch (xdp->command) {
+	case XDP_SETUP_PROG:
+		return setup_xdp(dev, xdp->prog);
+	case XDP_QUERY_PROG:
+		xdp->prog_id = priv->xdp_prog ? priv->xdp_prog->aux->id : 0;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
 static const struct net_device_ops dpaa2_eth_ops = {
 	.ndo_open = dpaa2_eth_open,
 	.ndo_start_xmit = dpaa2_eth_tx,
@@ -1409,6 +1739,8 @@ static const struct net_device_ops dpaa2_eth_ops = {
 	.ndo_set_rx_mode = dpaa2_eth_set_rx_mode,
 	.ndo_set_features = dpaa2_eth_set_features,
 	.ndo_do_ioctl = dpaa2_eth_ioctl,
+	.ndo_change_mtu = dpaa2_eth_change_mtu,
+	.ndo_bpf = dpaa2_eth_xdp,
 };
 
 static void cdan_cb(struct dpaa2_io_notification_ctx *ctx)
@@ -1434,8 +1766,11 @@ static struct fsl_mc_device *setup_dpcon(struct dpaa2_eth_priv *priv)
 	err = fsl_mc_object_allocate(to_fsl_mc_device(dev),
 				     FSL_MC_POOL_DPCON, &dpcon);
 	if (err) {
-		dev_info(dev, "Not enough DPCONs, will go on as-is\n");
-		return NULL;
+		if (err == -ENXIO)
+			err = -EPROBE_DEFER;
+		else
+			dev_info(dev, "Not enough DPCONs, will go on as-is\n");
+		return ERR_PTR(err);
 	}
 
 	err = dpcon_open(priv->mc_io, 0, dpcon->obj_desc.id, &dpcon->mc_handle);
@@ -1493,8 +1828,10 @@ alloc_channel(struct dpaa2_eth_priv *priv)
 		return NULL;
 
 	channel->dpcon = setup_dpcon(priv);
-	if (!channel->dpcon)
+	if (IS_ERR_OR_NULL(channel->dpcon)) {
+		err = PTR_ERR(channel->dpcon);
 		goto err_setup;
+	}
 
 	err = dpcon_get_attributes(priv->mc_io, 0, channel->dpcon->mc_handle,
 				   &attr);
@@ -1513,7 +1850,7 @@ err_get_attr:
 	free_dpcon(priv, channel->dpcon);
 err_setup:
 	kfree(channel);
-	return NULL;
+	return ERR_PTR(err);
 }
 
 static void free_channel(struct dpaa2_eth_priv *priv,
@@ -1547,10 +1884,11 @@ static int setup_dpio(struct dpaa2_eth_priv *priv)
 	for_each_online_cpu(i) {
 		/* Try to allocate a channel */
 		channel = alloc_channel(priv);
-		if (!channel) {
-			dev_info(dev,
-				 "No affine channel for cpu %d and above\n", i);
-			err = -ENODEV;
+		if (IS_ERR_OR_NULL(channel)) {
+			err = PTR_ERR(channel);
+			if (err != -EPROBE_DEFER)
+				dev_info(dev,
+					 "No affine channel for cpu %d and above\n", i);
 			goto err_alloc_ch;
 		}
 
@@ -1597,7 +1935,7 @@ static int setup_dpio(struct dpaa2_eth_priv *priv)
 		/* Stop if we already have enough channels to accommodate all
 		 * RX and TX conf queues
 		 */
-		if (priv->num_channels == dpaa2_eth_queue_count(priv))
+		if (priv->num_channels == priv->dpni_attrs.num_queues)
 			break;
 	}
 
@@ -1608,9 +1946,12 @@ err_set_cdan:
 err_service_reg:
 	free_channel(priv, channel);
 err_alloc_ch:
+	if (err == -EPROBE_DEFER)
+		return err;
+
 	if (cpumask_empty(&priv->dpio_cpumask)) {
 		dev_err(dev, "No cpu with an affine DPIO/DPCON\n");
-		return err;
+		return -ENODEV;
 	}
 
 	dev_info(dev, "Cores %*pbl available for processing ingress traffic\n",
@@ -1732,7 +2073,10 @@ static int setup_dpbp(struct dpaa2_eth_priv *priv)
 	err = fsl_mc_object_allocate(to_fsl_mc_device(dev), FSL_MC_POOL_DPBP,
 				     &dpbp_dev);
 	if (err) {
-		dev_err(dev, "DPBP device allocation failed\n");
+		if (err == -ENXIO)
+			err = -EPROBE_DEFER;
+		else
+			dev_err(dev, "DPBP device allocation failed\n");
 		return err;
 	}