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

Pull networking updates from David Miller:
 "Highlights:

   1) Support AES128-CCM ciphers in kTLS, from Vakul Garg.

   2) Add fib_sync_mem to control the amount of dirty memory we allow to
      queue up between synchronize RCU calls, from David Ahern.

   3) Make flow classifier more lockless, from Vlad Buslov.

   4) Add PHY downshift support to aquantia driver, from Heiner
      Kallweit.

   5) Add SKB cache for TCP rx and tx, from Eric Dumazet. This reduces
      contention on SLAB spinlocks in heavy RPC workloads.

   6) Partial GSO offload support in XFRM, from Boris Pismenny.

   7) Add fast link down support to ethtool, from Heiner Kallweit.

   8) Use siphash for IP ID generator, from Eric Dumazet.

   9) Pull nexthops even further out from ipv4/ipv6 routes and FIB
      entries, from David Ahern.

  10) Move skb->xmit_more into a per-cpu variable, from Florian
      Westphal.

  11) Improve eBPF verifier speed and increase maximum program size,
      from Alexei Starovoitov.

  12) Eliminate per-bucket spinlocks in rhashtable, and instead use bit
      spinlocks. From Neil Brown.

  13) Allow tunneling with GUE encap in ipvs, from Jacky Hu.

  14) Improve link partner cap detection in generic PHY code, from
      Heiner Kallweit.

  15) Add layer 2 encap support to bpf_skb_adjust_room(), from Alan
      Maguire.

  16) Remove SKB list implementation assumptions in SCTP, your's truly.

  17) Various cleanups, optimizations, and simplifications in r8169
      driver. From Heiner Kallweit.

  18) Add memory accounting on TX and RX path of SCTP, from Xin Long.

  19) Switch PHY drivers over to use dynamic featue detection, from
      Heiner Kallweit.

  20) Support flow steering without masking in dpaa2-eth, from Ioana
      Ciocoi.

  21) Implement ndo_get_devlink_port in netdevsim driver, from Jiri
      Pirko.

  22) Increase the strict parsing of current and future netlink
      attributes, also export such policies to userspace. From Johannes
      Berg.

  23) Allow DSA tag drivers to be modular, from Andrew Lunn.

  24) Remove legacy DSA probing support, also from Andrew Lunn.

  25) Allow ll_temac driver to be used on non-x86 platforms, from Esben
      Haabendal.

  26) Add a generic tracepoint for TX queue timeouts to ease debugging,
      from Cong Wang.

  27) More indirect call optimizations, from Paolo Abeni"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (1763 commits)
  cxgb4: Fix error path in cxgb4_init_module
  net: phy: improve pause mode reporting in phy_print_status
  dt-bindings: net: Fix a typo in the phy-mode list for ethernet bindings
  net: macb: Change interrupt and napi enable order in open
  net: ll_temac: Improve error message on error IRQ
  net/sched: remove block pointer from common offload structure
  net: ethernet: support of_get_mac_address new ERR_PTR error
  net: usb: smsc: fix warning reported by kbuild test robot
  staging: octeon-ethernet: Fix of_get_mac_address ERR_PTR check
  net: dsa: support of_get_mac_address new ERR_PTR error
  net: dsa: sja1105: Fix status initialization in sja1105_get_ethtool_stats
  vrf: sit mtu should not be updated when vrf netdev is the link
  net: dsa: Fix error cleanup path in dsa_init_module
  l2tp: Fix possible NULL pointer dereference
  taprio: add null check on sched_nest to avoid potential null pointer dereference
  net: mvpp2: cls: fix less than zero check on a u32 variable
  net_sched: sch_fq: handle non connected flows
  net_sched: sch_fq: do not assume EDT packets are ordered
  net: hns3: use devm_kcalloc when allocating desc_cb
  net: hns3: some cleanup for struct hns3_enet_ring
  ...

37 files changed, 2327 insertions, 485 deletions

diff --git a/drivers/net/ethernet/netronome/Kconfig b/drivers/net/ethernet/netronome/Kconfig
index 549898d5d450..f0d0e09f60e2 100644
--- a/drivers/net/ethernet/netronome/Kconfig
+++ b/drivers/net/ethernet/netronome/Kconfig
@@ -19,6 +19,7 @@ config NFP
 	tristate "Netronome(R) NFP4000/NFP6000 NIC driver"
 	depends on PCI && PCI_MSI
 	depends on VXLAN || VXLAN=n
+	select NET_DEVLINK
 	---help---
 	  This driver supports the Netronome(R) NFP4000/NFP6000 based
 	  cards working as a advanced Ethernet NIC.  It works with both
diff --git a/drivers/net/ethernet/netronome/nfp/Makefile b/drivers/net/ethernet/netronome/nfp/Makefile
index 47c708f08ade..87bf784f8e8f 100644
--- a/drivers/net/ethernet/netronome/nfp/Makefile
+++ b/drivers/net/ethernet/netronome/nfp/Makefile
@@ -15,6 +15,7 @@ nfp-objs := \
 	    nfpcore/nfp_resource.o \
 	    nfpcore/nfp_rtsym.o \
 	    nfpcore/nfp_target.o \
+	    ccm.o \
 	    nfp_asm.o \
 	    nfp_app.o \
 	    nfp_app_nic.o \
@@ -42,7 +43,8 @@ nfp-objs += \
 	    flower/match.o \
 	    flower/metadata.o \
 	    flower/offload.o \
-	    flower/tunnel_conf.o
+	    flower/tunnel_conf.o \
+	    flower/qos_conf.o
 endif
 
 ifeq ($(CONFIG_BPF_SYSCALL),y)
diff --git a/drivers/net/ethernet/netronome/nfp/abm/ctrl.c b/drivers/net/ethernet/netronome/nfp/abm/ctrl.c
index 9584f03f3efa..69e84ff7f2e5 100644
--- a/drivers/net/ethernet/netronome/nfp/abm/ctrl.c
+++ b/drivers/net/ethernet/netronome/nfp/abm/ctrl.c
@@ -261,10 +261,15 @@ int nfp_abm_ctrl_qm_disable(struct nfp_abm *abm)
 
 int nfp_abm_ctrl_prio_map_update(struct nfp_abm_link *alink, u32 *packed)
 {
+	const u32 cmd = NFP_NET_CFG_MBOX_CMD_PCI_DSCP_PRIOMAP_SET;
 	struct nfp_net *nn = alink->vnic;
 	unsigned int i;
 	int err;
 
+	err = nfp_net_mbox_lock(nn, alink->abm->prio_map_len);
+	if (err)
+		return err;
+
 	/* Write data_len and wipe reserved */
 	nn_writeq(nn, nn->tlv_caps.mbox_off + NFP_NET_ABM_MBOX_DATALEN,
 		  alink->abm->prio_map_len);
@@ -273,8 +278,7 @@ int nfp_abm_ctrl_prio_map_update(struct nfp_abm_link *alink, u32 *packed)
 		nn_writel(nn, nn->tlv_caps.mbox_off + NFP_NET_ABM_MBOX_DATA + i,
 			  packed[i / sizeof(u32)]);
 
-	err = nfp_net_reconfig_mbox(nn,
-				    NFP_NET_CFG_MBOX_CMD_PCI_DSCP_PRIOMAP_SET);
+	err = nfp_net_mbox_reconfig_and_unlock(nn, cmd);
 	if (err)
 		nfp_err(alink->abm->app->cpp,
 			"setting DSCP -> VQ map failed with error %d\n", err);
diff --git a/drivers/net/ethernet/netronome/nfp/abm/main.c b/drivers/net/ethernet/netronome/nfp/abm/main.c
index 4d4ff5844c47..9183b3e85d21 100644
--- a/drivers/net/ethernet/netronome/nfp/abm/main.c
+++ b/drivers/net/ethernet/netronome/nfp/abm/main.c
@@ -53,7 +53,8 @@ nfp_abm_setup_tc(struct nfp_app *app, struct net_device *netdev,
 	}
 }
 
-static struct net_device *nfp_abm_repr_get(struct nfp_app *app, u32 port_id)
+static struct net_device *
+nfp_abm_repr_get(struct nfp_app *app, u32 port_id, bool *redir_egress)
 {
 	enum nfp_repr_type rtype;
 	struct nfp_reprs *reprs;
@@ -549,5 +550,5 @@ const struct nfp_app_type app_abm = {
 	.eswitch_mode_get	= nfp_abm_eswitch_mode_get,
 	.eswitch_mode_set	= nfp_abm_eswitch_mode_set,
 
-	.repr_get	= nfp_abm_repr_get,
+	.dev_get	= nfp_abm_repr_get,
 };
diff --git a/drivers/net/ethernet/netronome/nfp/bpf/cmsg.c b/drivers/net/ethernet/netronome/nfp/bpf/cmsg.c
index 9b6cfa697879..bc9850e4ec5e 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/cmsg.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/cmsg.c
@@ -6,48 +6,13 @@
 #include <linux/bug.h>
 #include <linux/jiffies.h>
 #include <linux/skbuff.h>
-#include <linux/wait.h>
 
+#include "../ccm.h"
 #include "../nfp_app.h"
 #include "../nfp_net.h"
 #include "fw.h"
 #include "main.h"
 
-#define NFP_BPF_TAG_ALLOC_SPAN	(U16_MAX / 4)
-
-static bool nfp_bpf_all_tags_busy(struct nfp_app_bpf *bpf)
-{
-	u16 used_tags;
-
-	used_tags = bpf->tag_alloc_next - bpf->tag_alloc_last;
-
-	return used_tags > NFP_BPF_TAG_ALLOC_SPAN;
-}
-
-static int nfp_bpf_alloc_tag(struct nfp_app_bpf *bpf)
-{
-	/* All FW communication for BPF is request-reply.  To make sure we
-	 * don't reuse the message ID too early after timeout - limit the
-	 * number of requests in flight.
-	 */
-	if (nfp_bpf_all_tags_busy(bpf)) {
-		cmsg_warn(bpf, "all FW request contexts busy!\n");
-		return -EAGAIN;
-	}
-
-	WARN_ON(__test_and_set_bit(bpf->tag_alloc_next, bpf->tag_allocator));
-	return bpf->tag_alloc_next++;
-}
-
-static void nfp_bpf_free_tag(struct nfp_app_bpf *bpf, u16 tag)
-{
-	WARN_ON(!__test_and_clear_bit(tag, bpf->tag_allocator));
-
-	while (!test_bit(bpf->tag_alloc_last, bpf->tag_allocator) &&
-	       bpf->tag_alloc_last != bpf->tag_alloc_next)
-		bpf->tag_alloc_last++;
-}
-
 static struct sk_buff *
 nfp_bpf_cmsg_alloc(struct nfp_app_bpf *bpf, unsigned int size)
 {
@@ -87,149 +52,6 @@ nfp_bpf_cmsg_map_reply_size(struct nfp_app_bpf *bpf, unsigned int n)
 	return size;
 }
 
-static u8 nfp_bpf_cmsg_get_type(struct sk_buff *skb)
-{
-	struct cmsg_hdr *hdr;
-
-	hdr = (struct cmsg_hdr *)skb->data;
-
-	return hdr->type;
-}
-
-static unsigned int nfp_bpf_cmsg_get_tag(struct sk_buff *skb)
-{
-	struct cmsg_hdr *hdr;
-
-	hdr = (struct cmsg_hdr *)skb->data;
-
-	return be16_to_cpu(hdr->tag);
-}
-
-static struct sk_buff *__nfp_bpf_reply(struct nfp_app_bpf *bpf, u16 tag)
-{
-	unsigned int msg_tag;
-	struct sk_buff *skb;
-
-	skb_queue_walk(&bpf->cmsg_replies, skb) {
-		msg_tag = nfp_bpf_cmsg_get_tag(skb);
-		if (msg_tag == tag) {
-			nfp_bpf_free_tag(bpf, tag);
-			__skb_unlink(skb, &bpf->cmsg_replies);
-			return skb;
-		}
-	}
-
-	return NULL;
-}
-
-static struct sk_buff *nfp_bpf_reply(struct nfp_app_bpf *bpf, u16 tag)
-{
-	struct sk_buff *skb;
-
-	nfp_ctrl_lock(bpf->app->ctrl);
-	skb = __nfp_bpf_reply(bpf, tag);
-	nfp_ctrl_unlock(bpf->app->ctrl);
-
-	return skb;
-}
-
-static struct sk_buff *nfp_bpf_reply_drop_tag(struct nfp_app_bpf *bpf, u16 tag)
-{
-	struct sk_buff *skb;
-
-	nfp_ctrl_lock(bpf->app->ctrl);
-	skb = __nfp_bpf_reply(bpf, tag);
-	if (!skb)
-		nfp_bpf_free_tag(bpf, tag);
-	nfp_ctrl_unlock(bpf->app->ctrl);
-
-	return skb;
-}
-
-static struct sk_buff *
-nfp_bpf_cmsg_wait_reply(struct nfp_app_bpf *bpf, enum nfp_bpf_cmsg_type type,
-			int tag)
-{
-	struct sk_buff *skb;
-	int i, err;
-
-	for (i = 0; i < 50; i++) {
-		udelay(4);
-		skb = nfp_bpf_reply(bpf, tag);
-		if (skb)
-			return skb;
-	}
-
-	err = wait_event_interruptible_timeout(bpf->cmsg_wq,
-					       skb = nfp_bpf_reply(bpf, tag),
-					       msecs_to_jiffies(5000));
-	/* We didn't get a response - try last time and atomically drop
-	 * the tag even if no response is matched.
-	 */
-	if (!skb)
-		skb = nfp_bpf_reply_drop_tag(bpf, tag);
-	if (err < 0) {
-		cmsg_warn(bpf, "%s waiting for response to 0x%02x: %d\n",
-			  err == ERESTARTSYS ? "interrupted" : "error",
-			  type, err);
-		return ERR_PTR(err);
-	}
-	if (!skb) {
-		cmsg_warn(bpf, "timeout waiting for response to 0x%02x\n",
-			  type);
-		return ERR_PTR(-ETIMEDOUT);
-	}
-
-	return skb;
-}
-
-static struct sk_buff *
-nfp_bpf_cmsg_communicate(struct nfp_app_bpf *bpf, struct sk_buff *skb,
-			 enum nfp_bpf_cmsg_type type, unsigned int reply_size)
-{
-	struct cmsg_hdr *hdr;
-	int tag;
-
-	nfp_ctrl_lock(bpf->app->ctrl);
-	tag = nfp_bpf_alloc_tag(bpf);
-	if (tag < 0) {
-		nfp_ctrl_unlock(bpf->app->ctrl);
-		dev_kfree_skb_any(skb);
-		return ERR_PTR(tag);
-	}
-
-	hdr = (void *)skb->data;
-	hdr->ver = CMSG_MAP_ABI_VERSION;
-	hdr->type = type;
-	hdr->tag = cpu_to_be16(tag);
-
-	__nfp_app_ctrl_tx(bpf->app, skb);
-
-	nfp_ctrl_unlock(bpf->app->ctrl);
-
-	skb = nfp_bpf_cmsg_wait_reply(bpf, type, tag);
-	if (IS_ERR(skb))
-		return skb;
-
-	hdr = (struct cmsg_hdr *)skb->data;
-	if (hdr->type != __CMSG_REPLY(type)) {
-		cmsg_warn(bpf, "cmsg drop - wrong type 0x%02x != 0x%02lx!\n",
-			  hdr->type, __CMSG_REPLY(type));
-		goto err_free;
-	}
-	/* 0 reply_size means caller will do the validation */
-	if (reply_size && skb->len != reply_size) {
-		cmsg_warn(bpf, "cmsg drop - type 0x%02x wrong size %d != %d!\n",
-			  type, skb->len, reply_size);
-		goto err_free;
-	}
-
-	return skb;
-err_free:
-	dev_kfree_skb_any(skb);
-	return ERR_PTR(-EIO);
-}
-
 static int
 nfp_bpf_ctrl_rc_to_errno(struct nfp_app_bpf *bpf,
 			 struct cmsg_reply_map_simple *reply)
@@ -275,8 +97,8 @@ nfp_bpf_ctrl_alloc_map(struct nfp_app_bpf *bpf, struct bpf_map *map)
 	req->map_type = cpu_to_be32(map->map_type);
 	req->map_flags = 0;
 
-	skb = nfp_bpf_cmsg_communicate(bpf, skb, CMSG_TYPE_MAP_ALLOC,
-				       sizeof(*reply));
+	skb = nfp_ccm_communicate(&bpf->ccm, skb, NFP_CCM_TYPE_BPF_MAP_ALLOC,
+				  sizeof(*reply));
 	if (IS_ERR(skb))
 		return PTR_ERR(skb);
 
@@ -310,8 +132,8 @@ void nfp_bpf_ctrl_free_map(struct nfp_app_bpf *bpf, struct nfp_bpf_map *nfp_map)
 	req = (void *)skb->data;
 	req->tid = cpu_to_be32(nfp_map->tid);
 
-	skb = nfp_bpf_cmsg_communicate(bpf, skb, CMSG_TYPE_MAP_FREE,
-				       sizeof(*reply));
+	skb = nfp_ccm_communicate(&bpf->ccm, skb, NFP_CCM_TYPE_BPF_MAP_FREE,
+				  sizeof(*reply));
 	if (IS_ERR(skb)) {
 		cmsg_warn(bpf, "leaking map - I/O error\n");
 		return;
@@ -354,8 +176,7 @@ nfp_bpf_ctrl_reply_val(struct nfp_app_bpf *bpf, struct cmsg_reply_map_op *reply,
 }
 
 static int
-nfp_bpf_ctrl_entry_op(struct bpf_offloaded_map *offmap,
-		      enum nfp_bpf_cmsg_type op,
+nfp_bpf_ctrl_entry_op(struct bpf_offloaded_map *offmap, enum nfp_ccm_type op,
 		      u8 *key, u8 *value, u64 flags, u8 *out_key, u8 *out_value)
 {
 	struct nfp_bpf_map *nfp_map = offmap->dev_priv;
@@ -386,8 +207,8 @@ nfp_bpf_ctrl_entry_op(struct bpf_offloaded_map *offmap,
 		memcpy(nfp_bpf_ctrl_req_val(bpf, req, 0), value,
 		       map->value_size);
 
-	skb = nfp_bpf_cmsg_communicate(bpf, skb, op,
-				       nfp_bpf_cmsg_map_reply_size(bpf, 1));
+	skb = nfp_ccm_communicate(&bpf->ccm, skb, op,
+				  nfp_bpf_cmsg_map_reply_size(bpf, 1));
 	if (IS_ERR(skb))
 		return PTR_ERR(skb);
 
@@ -415,34 +236,34 @@ err_free:
 int nfp_bpf_ctrl_update_entry(struct bpf_offloaded_map *offmap,
 			      void *key, void *value, u64 flags)
 {
-	return nfp_bpf_ctrl_entry_op(offmap, CMSG_TYPE_MAP_UPDATE,
+	return nfp_bpf_ctrl_entry_op(offmap, NFP_CCM_TYPE_BPF_MAP_UPDATE,
 				     key, value, flags, NULL, NULL);
 }
 
 int nfp_bpf_ctrl_del_entry(struct bpf_offloaded_map *offmap, void *key)
 {
-	return nfp_bpf_ctrl_entry_op(offmap, CMSG_TYPE_MAP_DELETE,
+	return nfp_bpf_ctrl_entry_op(offmap, NFP_CCM_TYPE_BPF_MAP_DELETE,
 				     key, NULL, 0, NULL, NULL);
 }
 
 int nfp_bpf_ctrl_lookup_entry(struct bpf_offloaded_map *offmap,
 			      void *key, void *value)
 {
-	return nfp_bpf_ctrl_entry_op(offmap, CMSG_TYPE_MAP_LOOKUP,
+	return nfp_bpf_ctrl_entry_op(offmap, NFP_CCM_TYPE_BPF_MAP_LOOKUP,
 				     key, NULL, 0, NULL, value);
 }
 
 int nfp_bpf_ctrl_getfirst_entry(struct bpf_offloaded_map *offmap,
 				void *next_key)
 {
-	return nfp_bpf_ctrl_entry_op(offmap, CMSG_TYPE_MAP_GETFIRST,
+	return nfp_bpf_ctrl_entry_op(offmap, NFP_CCM_TYPE_BPF_MAP_GETFIRST,
 				     NULL, NULL, 0, next_key, NULL);
 }
 
 int nfp_bpf_ctrl_getnext_entry(struct bpf_offloaded_map *offmap,
 			       void *key, void *next_key)
 {
-	return nfp_bpf_ctrl_entry_op(offmap, CMSG_TYPE_MAP_GETNEXT,
+	return nfp_bpf_ctrl_entry_op(offmap, NFP_CCM_TYPE_BPF_MAP_GETNEXT,
 				     key, NULL, 0, next_key, NULL);
 }
 
@@ -456,54 +277,35 @@ unsigned int nfp_bpf_ctrl_cmsg_mtu(struct nfp_app_bpf *bpf)
 void nfp_bpf_ctrl_msg_rx(struct nfp_app *app, struct sk_buff *skb)
 {
 	struct nfp_app_bpf *bpf = app->priv;
-	unsigned int tag;
 
 	if (unlikely(skb->len < sizeof(struct cmsg_reply_map_simple))) {
 		cmsg_warn(bpf, "cmsg drop - too short %d!\n", skb->len);
-		goto err_free;
+		dev_kfree_skb_any(skb);
+		return;
 	}
 
-	if (nfp_bpf_cmsg_get_type(skb) == CMSG_TYPE_BPF_EVENT) {
+	if (nfp_ccm_get_type(skb) == NFP_CCM_TYPE_BPF_BPF_EVENT) {
 		if (!nfp_bpf_event_output(bpf, skb->data, skb->len))
 			dev_consume_skb_any(skb);
 		else
 			dev_kfree_skb_any(skb);
-		return;
 	}
 
-	nfp_ctrl_lock(bpf->app->ctrl);
-
-	tag = nfp_bpf_cmsg_get_tag(skb);
-	if (unlikely(!test_bit(tag, bpf->tag_allocator))) {
-		cmsg_warn(bpf, "cmsg drop - no one is waiting for tag %u!\n",
-			  tag);
-		goto err_unlock;
-	}
-
-	__skb_queue_tail(&bpf->cmsg_replies, skb);
-	wake_up_interruptible_all(&bpf->cmsg_wq);
-
-	nfp_ctrl_unlock(bpf->app->ctrl);
-
-	return;
-err_unlock:
-	nfp_ctrl_unlock(bpf->app->ctrl);
-err_free:
-	dev_kfree_skb_any(skb);
+	nfp_ccm_rx(&bpf->ccm, skb);
 }
 
 void
 nfp_bpf_ctrl_msg_rx_raw(struct nfp_app *app, const void *data, unsigned int len)
 {
+	const struct nfp_ccm_hdr *hdr = data;
 	struct nfp_app_bpf *bpf = app->priv;
-	const struct cmsg_hdr *hdr = data;
 
 	if (unlikely(len < sizeof(struct cmsg_reply_map_simple))) {
 		cmsg_warn(bpf, "cmsg drop - too short %d!\n", len);
 		return;
 	}
 
-	if (hdr->type == CMSG_TYPE_BPF_EVENT)
+	if (hdr->type == NFP_CCM_TYPE_BPF_BPF_EVENT)
 		nfp_bpf_event_output(bpf, data, len);
 	else
 		cmsg_warn(bpf, "cmsg drop - msg type %d with raw buffer!\n",
diff --git a/drivers/net/ethernet/netronome/nfp/bpf/fw.h b/drivers/net/ethernet/netronome/nfp/bpf/fw.h
index 721921bcf120..06c4286bd79e 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/fw.h
+++ b/drivers/net/ethernet/netronome/nfp/bpf/fw.h
@@ -6,6 +6,7 @@
 
 #include <linux/bitops.h>
 #include <linux/types.h>
+#include "../ccm.h"
 
 /* Kernel's enum bpf_reg_type is not uABI so people may change it breaking
  * our FW ABI.  In that case we will do translation in the driver.
@@ -52,22 +53,6 @@ struct nfp_bpf_cap_tlv_maps {
 /*
  * Types defined for map related control messages
  */
-#define CMSG_MAP_ABI_VERSION		1
-
-enum nfp_bpf_cmsg_type {
-	CMSG_TYPE_MAP_ALLOC	= 1,
-	CMSG_TYPE_MAP_FREE	= 2,
-	CMSG_TYPE_MAP_LOOKUP	= 3,
-	CMSG_TYPE_MAP_UPDATE	= 4,
-	CMSG_TYPE_MAP_DELETE	= 5,
-	CMSG_TYPE_MAP_GETNEXT	= 6,
-	CMSG_TYPE_MAP_GETFIRST	= 7,
-	CMSG_TYPE_BPF_EVENT	= 8,
-	__CMSG_TYPE_MAP_MAX,
-};
-
-#define CMSG_TYPE_MAP_REPLY_BIT		7
-#define __CMSG_REPLY(req)		(BIT(CMSG_TYPE_MAP_REPLY_BIT) | (req))
 
 /* BPF ABIv2 fixed-length control message fields */
 #define CMSG_MAP_KEY_LW			16
@@ -84,19 +69,13 @@ enum nfp_bpf_cmsg_status {
 	CMSG_RC_ERR_MAP_E2BIG		= 7,
 };
 
-struct cmsg_hdr {
-	u8 type;
-	u8 ver;
-	__be16 tag;
-};
-
 struct cmsg_reply_map_simple {
-	struct cmsg_hdr hdr;
+	struct nfp_ccm_hdr hdr;
 	__be32 rc;
 };
 
 struct cmsg_req_map_alloc_tbl {
-	struct cmsg_hdr hdr;
+	struct nfp_ccm_hdr hdr;
 	__be32 key_size;		/* in bytes */
 	__be32 value_size;		/* in bytes */
 	__be32 max_entries;
@@ -110,7 +89,7 @@ struct cmsg_reply_map_alloc_tbl {
 };
 
 struct cmsg_req_map_free_tbl {
-	struct cmsg_hdr hdr;
+	struct nfp_ccm_hdr hdr;
 	__be32 tid;
 };
 
@@ -120,7 +99,7 @@ struct cmsg_reply_map_free_tbl {
 };
 
 struct cmsg_req_map_op {
-	struct cmsg_hdr hdr;
+	struct nfp_ccm_hdr hdr;
 	__be32 tid;
 	__be32 count;
 	__be32 flags;
@@ -135,7 +114,7 @@ struct cmsg_reply_map_op {
 };
 
 struct cmsg_bpf_event {
-	struct cmsg_hdr hdr;
+	struct nfp_ccm_hdr hdr;
 	__be32 cpu_id;
 	__be64 map_ptr;
 	__be32 data_size;
diff --git a/drivers/net/ethernet/netronome/nfp/bpf/main.c b/drivers/net/ethernet/netronome/nfp/bpf/main.c
index 275de9f4c61c..9c136da25221 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/main.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/main.c
@@ -442,14 +442,16 @@ static int nfp_bpf_init(struct nfp_app *app)
 	bpf->app = app;
 	app->priv = bpf;
 
-	skb_queue_head_init(&bpf->cmsg_replies);
-	init_waitqueue_head(&bpf->cmsg_wq);
 	INIT_LIST_HEAD(&bpf->map_list);
 
-	err = rhashtable_init(&bpf->maps_neutral, &nfp_bpf_maps_neutral_params);
+	err = nfp_ccm_init(&bpf->ccm, app);
 	if (err)
 		goto err_free_bpf;
 
+	err = rhashtable_init(&bpf->maps_neutral, &nfp_bpf_maps_neutral_params);
+	if (err)
+		goto err_clean_ccm;
+
 	nfp_bpf_init_capabilities(bpf);
 
 	err = nfp_bpf_parse_capabilities(app);
@@ -474,6 +476,8 @@ static int nfp_bpf_init(struct nfp_app *app)
 
 err_free_neutral_maps:
 	rhashtable_destroy(&bpf->maps_neutral);
+err_clean_ccm:
+	nfp_ccm_clean(&bpf->ccm);
 err_free_bpf:
 	kfree(bpf);
 	return err;
@@ -484,7 +488,7 @@ static void nfp_bpf_clean(struct nfp_app *app)
 	struct nfp_app_bpf *bpf = app->priv;
 
 	bpf_offload_dev_destroy(bpf->bpf_dev);
-	WARN_ON(!skb_queue_empty(&bpf->cmsg_replies));
+	nfp_ccm_clean(&bpf->ccm);
 	WARN_ON(!list_empty(&bpf->map_list));
 	WARN_ON(bpf->maps_in_use || bpf->map_elems_in_use);
 	rhashtable_free_and_destroy(&bpf->maps_neutral,
diff --git a/drivers/net/ethernet/netronome/nfp/bpf/main.h b/drivers/net/ethernet/netronome/nfp/bpf/main.h
index b25a48218bcf..e54d1ac84df2 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/main.h
+++ b/drivers/net/ethernet/netronome/nfp/bpf/main.h
@@ -14,6 +14,7 @@
 #include <linux/types.h>
 #include <linux/wait.h>
 
+#include "../ccm.h"
 #include "../nfp_asm.h"
 #include "fw.h"
 
@@ -84,16 +85,10 @@ enum pkt_vec {
 /**
  * struct nfp_app_bpf - bpf app priv structure
  * @app:		backpointer to the app
+ * @ccm:		common control message handler data
  *
  * @bpf_dev:		BPF offload device handle
  *
- * @tag_allocator:	bitmap of control message tags in use
- * @tag_alloc_next:	next tag bit to allocate
- * @tag_alloc_last:	next tag bit to be freed
- *
- * @cmsg_replies:	received cmsg replies waiting to be consumed
- * @cmsg_wq:		work queue for waiting for cmsg replies
- *
  * @cmsg_key_sz:	size of key in cmsg element array
  * @cmsg_val_sz:	size of value in cmsg element array
  *
@@ -132,16 +127,10 @@ enum pkt_vec {
  */
 struct nfp_app_bpf {
 	struct nfp_app *app;
+	struct nfp_ccm ccm;
 
 	struct bpf_offload_dev *bpf_dev;
 
-	DECLARE_BITMAP(tag_allocator, U16_MAX + 1);
-	u16 tag_alloc_next;
-	u16 tag_alloc_last;
-
-	struct sk_buff_head cmsg_replies;
-	struct wait_queue_head cmsg_wq;
-
 	unsigned int cmsg_key_sz;
 	unsigned int cmsg_val_sz;
 
diff --git a/drivers/net/ethernet/netronome/nfp/bpf/offload.c b/drivers/net/ethernet/netronome/nfp/bpf/offload.c
index 15dce97650a5..39c9fec222b4 100644
--- a/drivers/net/ethernet/netronome/nfp/bpf/offload.c
+++ b/drivers/net/ethernet/netronome/nfp/bpf/offload.c
@@ -22,6 +22,7 @@
 #include <net/tc_act/tc_mirred.h>
 
 #include "main.h"
+#include "../ccm.h"
 #include "../nfp_app.h"
 #include "../nfp_net_ctrl.h"
 #include "../nfp_net.h"
@@ -452,7 +453,7 @@ int nfp_bpf_event_output(struct nfp_app_bpf *bpf, const void *data,
 
 	if (len < sizeof(struct cmsg_bpf_event) + pkt_size + data_size)
 		return -EINVAL;
-	if (cbe->hdr.ver != CMSG_MAP_ABI_VERSION)
+	if (cbe->hdr.ver != NFP_CCM_ABI_VERSION)
 		return -EINVAL;
 
 	rcu_read_lock();
diff --git a/drivers/net/ethernet/netronome/nfp/ccm.c b/drivers/net/ethernet/netronome/nfp/ccm.c
new file mode 100644
index 000000000000..94476e41e261
--- /dev/null
+++ b/drivers/net/ethernet/netronome/nfp/ccm.c
@@ -0,0 +1,220 @@
+// SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+/* Copyright (C) 2016-2019 Netronome Systems, Inc. */
+
+#include <linux/bitops.h>
+
+#include "ccm.h"
+#include "nfp_app.h"
+#include "nfp_net.h"
+
+#define NFP_CCM_TYPE_REPLY_BIT		7
+#define __NFP_CCM_REPLY(req)		(BIT(NFP_CCM_TYPE_REPLY_BIT) | (req))
+
+#define ccm_warn(app, msg...)	nn_dp_warn(&(app)->ctrl->dp, msg)
+
+#define NFP_CCM_TAG_ALLOC_SPAN	(U16_MAX / 4)
+
+static bool nfp_ccm_all_tags_busy(struct nfp_ccm *ccm)
+{
+	u16 used_tags;
+
+	used_tags = ccm->tag_alloc_next - ccm->tag_alloc_last;
+
+	return used_tags > NFP_CCM_TAG_ALLOC_SPAN;
+}
+
+static int nfp_ccm_alloc_tag(struct nfp_ccm *ccm)
+{
+	/* CCM is for FW communication which is request-reply.  To make sure
+	 * we don't reuse the message ID too early after timeout - limit the
+	 * number of requests in flight.
+	 */
+	if (unlikely(nfp_ccm_all_tags_busy(ccm))) {
+		ccm_warn(ccm->app, "all FW request contexts busy!\n");
+		return -EAGAIN;
+	}
+
+	WARN_ON(__test_and_set_bit(ccm->tag_alloc_next, ccm->tag_allocator));
+	return ccm->tag_alloc_next++;
+}
+
+static void nfp_ccm_free_tag(struct nfp_ccm *ccm, u16 tag)
+{
+	WARN_ON(!__test_and_clear_bit(tag, ccm->tag_allocator));
+
+	while (!test_bit(ccm->tag_alloc_last, ccm->tag_allocator) &&
+	       ccm->tag_alloc_last != ccm->tag_alloc_next)
+		ccm->tag_alloc_last++;
+}
+
+static struct sk_buff *__nfp_ccm_reply(struct nfp_ccm *ccm, u16 tag)
+{
+	unsigned int msg_tag;
+	struct sk_buff *skb;
+
+	skb_queue_walk(&ccm->replies, skb) {
+		msg_tag = nfp_ccm_get_tag(skb);
+		if (msg_tag == tag) {
+			nfp_ccm_free_tag(ccm, tag);
+			__skb_unlink(skb, &ccm->replies);
+			return skb;
+		}
+	}
+
+	return NULL;
+}
+
+static struct sk_buff *
+nfp_ccm_reply(struct nfp_ccm *ccm, struct nfp_app *app, u16 tag)
+{
+	struct sk_buff *skb;
+
+	nfp_ctrl_lock(app->ctrl);
+	skb = __nfp_ccm_reply(ccm, tag);
+	nfp_ctrl_unlock(app->ctrl);
+
+	return skb;
+}
+
+static struct sk_buff *
+nfp_ccm_reply_drop_tag(struct nfp_ccm *ccm, struct nfp_app *app, u16 tag)
+{
+	struct sk_buff *skb;
+
+	nfp_ctrl_lock(app->ctrl);
+	skb = __nfp_ccm_reply(ccm, tag);
+	if (!skb)
+		nfp_ccm_free_tag(ccm, tag);
+	nfp_ctrl_unlock(app->ctrl);
+
+	return skb;
+}
+
+static struct sk_buff *
+nfp_ccm_wait_reply(struct nfp_ccm *ccm, struct nfp_app *app,
+		   enum nfp_ccm_type type, int tag)
+{
+	struct sk_buff *skb;
+	int i, err;
+
+	for (i = 0; i < 50; i++) {
+		udelay(4);
+		skb = nfp_ccm_reply(ccm, app, tag);
+		if (skb)
+			return skb;
+	}
+
+	err = wait_event_interruptible_timeout(ccm->wq,
+					       skb = nfp_ccm_reply(ccm, app,
+								   tag),
+					       msecs_to_jiffies(5000));
+	/* We didn't get a response - try last time and atomically drop
+	 * the tag even if no response is matched.
+	 */
+	if (!skb)
+		skb = nfp_ccm_reply_drop_tag(ccm, app, tag);
+	if (err < 0) {
+		ccm_warn(app, "%s waiting for response to 0x%02x: %d\n",
+			 err == ERESTARTSYS ? "interrupted" : "error",
+			 type, err);
+		return ERR_PTR(err);
+	}
+	if (!skb) {
+		ccm_warn(app, "timeout waiting for response to 0x%02x\n", type);
+		return ERR_PTR(-ETIMEDOUT);
+	}
+
+	return skb;
+}
+
+struct sk_buff *
+nfp_ccm_communicate(struct nfp_ccm *ccm, struct sk_buff *skb,
+		    enum nfp_ccm_type type, unsigned int reply_size)
+{
+	struct nfp_app *app = ccm->app;
+	struct nfp_ccm_hdr *hdr;
+	int reply_type, tag;
+
+	nfp_ctrl_lock(app->ctrl);
+	tag = nfp_ccm_alloc_tag(ccm);
+	if (tag < 0) {
+		nfp_ctrl_unlock(app->ctrl);
+		dev_kfree_skb_any(skb);
+		return ERR_PTR(tag);
+	}
+
+	hdr = (void *)skb->data;
+	hdr->ver = NFP_CCM_ABI_VERSION;
+	hdr->type = type;
+	hdr->tag = cpu_to_be16(tag);
+
+	__nfp_app_ctrl_tx(app, skb);
+
+	nfp_ctrl_unlock(app->ctrl);
+
+	skb = nfp_ccm_wait_reply(ccm, app, type, tag);
+	if (IS_ERR(skb))
+		return skb;
+
+	reply_type = nfp_ccm_get_type(skb);
+	if (reply_type != __NFP_CCM_REPLY(type)) {
+		ccm_warn(app, "cmsg drop - wrong type 0x%02x != 0x%02lx!\n",
+			 reply_type, __NFP_CCM_REPLY(type));
+		goto err_free;
+	}
+	/* 0 reply_size means caller will do the validation */
+	if (reply_size && skb->len != reply_size) {
+		ccm_warn(app, "cmsg drop - type 0x%02x wrong size %d != %d!\n",
+			 type, skb->len, reply_size);
+		goto err_free;
+	}
+
+	return skb;
+err_free:
+	dev_kfree_skb_any(skb);
+	return ERR_PTR(-EIO);
+}
+
+void nfp_ccm_rx(struct nfp_ccm *ccm, struct sk_buff *skb)
+{
+	struct nfp_app *app = ccm->app;
+	unsigned int tag;
+
+	if (unlikely(skb->len < sizeof(struct nfp_ccm_hdr))) {
+		ccm_warn(app, "cmsg drop - too short %d!\n", skb->len);
+		goto err_free;
+	}
+
+	nfp_ctrl_lock(app->ctrl);
+
+	tag = nfp_ccm_get_tag(skb);
+	if (unlikely(!test_bit(tag, ccm->tag_allocator))) {
+		ccm_warn(app, "cmsg drop - no one is waiting for tag %u!\n",
+			 tag);
+		goto err_unlock;
+	}
+
+	__skb_queue_tail(&ccm->replies, skb);
+	wake_up_interruptible_all(&ccm->wq);
+
+	nfp_ctrl_unlock(app->ctrl);
+	return;
+
+err_unlock:
+	nfp_ctrl_unlock(app->ctrl);
+err_free:
+	dev_kfree_skb_any(skb);
+}
+
+int nfp_ccm_init(struct nfp_ccm *ccm, struct nfp_app *app)
+{
+	ccm->app = app;
+	skb_queue_head_init(&ccm->replies);
+	init_waitqueue_head(&ccm->wq);
+	return 0;
+}
+
+void nfp_ccm_clean(struct nfp_ccm *ccm)
+{
+	WARN_ON(!skb_queue_empty(&ccm->replies));
+}
diff --git a/drivers/net/ethernet/netronome/nfp/ccm.h b/drivers/net/ethernet/netronome/nfp/ccm.h
new file mode 100644
index 000000000000..e2fe4b867958
--- /dev/null
+++ b/drivers/net/ethernet/netronome/nfp/ccm.h
@@ -0,0 +1,81 @@
+/* SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause) */
+/* Copyright (C) 2016-2019 Netronome Systems, Inc. */
+
+#ifndef NFP_CCM_H
+#define NFP_CCM_H 1
+
+#include <linux/bitmap.h>
+#include <linux/skbuff.h>
+#include <linux/wait.h>
+
+struct nfp_app;
+
+/* Firmware ABI */
+
+enum nfp_ccm_type {
+	NFP_CCM_TYPE_BPF_MAP_ALLOC	= 1,
+	NFP_CCM_TYPE_BPF_MAP_FREE	= 2,
+	NFP_CCM_TYPE_BPF_MAP_LOOKUP	= 3,
+	NFP_CCM_TYPE_BPF_MAP_UPDATE	= 4,
+	NFP_CCM_TYPE_BPF_MAP_DELETE	= 5,
+	NFP_CCM_TYPE_BPF_MAP_GETNEXT	= 6,
+	NFP_CCM_TYPE_BPF_MAP_GETFIRST	= 7,
+	NFP_CCM_TYPE_BPF_BPF_EVENT	= 8,
+	__NFP_CCM_TYPE_MAX,
+};
+
+#define NFP_CCM_ABI_VERSION		1
+
+struct nfp_ccm_hdr {
+	u8 type;
+	u8 ver;
+	__be16 tag;
+};
+
+static inline u8 nfp_ccm_get_type(struct sk_buff *skb)
+{
+	struct nfp_ccm_hdr *hdr;
+
+	hdr = (struct nfp_ccm_hdr *)skb->data;
+
+	return hdr->type;
+}
+
+static inline unsigned int nfp_ccm_get_tag(struct sk_buff *skb)
+{
+	struct nfp_ccm_hdr *hdr;
+
+	hdr = (struct nfp_ccm_hdr *)skb->data;
+
+	return be16_to_cpu(hdr->tag);
+}
+
+/* Implementation */
+
+/**
+ * struct nfp_ccm - common control message handling
+ * @tag_allocator:	bitmap of control message tags in use
+ * @tag_alloc_next:	next tag bit to allocate
+ * @tag_alloc_last:	next tag bit to be freed
+ *
+ * @replies:		received cmsg replies waiting to be consumed
+ * @wq:			work queue for waiting for cmsg replies
+ */
+struct nfp_ccm {
+	struct nfp_app *app;
+
+	DECLARE_BITMAP(tag_allocator, U16_MAX + 1);
+	u16 tag_alloc_next;
+	u16 tag_alloc_last;
+
+	struct sk_buff_head replies;
+	struct wait_queue_head wq;
+};
+
+int nfp_ccm_init(struct nfp_ccm *ccm, struct nfp_app *app);
+void nfp_ccm_clean(struct nfp_ccm *ccm);
+void nfp_ccm_rx(struct nfp_ccm *ccm, struct sk_buff *skb);
+struct sk_buff *
+nfp_ccm_communicate(struct nfp_ccm *ccm, struct sk_buff *skb,
+		    enum nfp_ccm_type type, unsigned int reply_size);
+#endif
diff --git a/drivers/net/ethernet/netronome/nfp/flower/action.c b/drivers/net/ethernet/netronome/nfp/flower/action.c
index e336f6ee94f5..c56e31d9f8a4 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/action.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/action.c
@@ -160,9 +160,9 @@ nfp_fl_get_tun_from_act_l4_port(struct nfp_app *app,
 	struct nfp_flower_priv *priv = app->priv;
 
 	switch (tun->key.tp_dst) {
-	case htons(NFP_FL_VXLAN_PORT):
+	case htons(IANA_VXLAN_UDP_PORT):
 		return NFP_FL_TUNNEL_VXLAN;
-	case htons(NFP_FL_GENEVE_PORT):
+	case htons(GENEVE_UDP_PORT):
 		if (priv->flower_ext_feats & NFP_FL_FEATS_GENEVE)
 			return NFP_FL_TUNNEL_GENEVE;
 		/* FALLTHROUGH */
@@ -582,60 +582,23 @@ static u32 nfp_fl_csum_l4_to_flag(u8 ip_proto)
 	}
 }
 
-static int
-nfp_fl_pedit(const struct flow_action_entry *act,
-	     struct tc_cls_flower_offload *flow,
-	     char *nfp_action, int *a_len, u32 *csum_updated)
-{
-	struct flow_rule *rule = tc_cls_flower_offload_flow_rule(flow);
+struct nfp_flower_pedit_acts {
 	struct nfp_fl_set_ipv6_addr set_ip6_dst, set_ip6_src;
 	struct nfp_fl_set_ipv6_tc_hl_fl set_ip6_tc_hl_fl;
 	struct nfp_fl_set_ip4_ttl_tos set_ip_ttl_tos;
 	struct nfp_fl_set_ip4_addrs set_ip_addr;
-	enum flow_action_mangle_base htype;
 	struct nfp_fl_set_tport set_tport;
 	struct nfp_fl_set_eth set_eth;
+};
+
+static int
+nfp_fl_commit_mangle(struct tc_cls_flower_offload *flow, char *nfp_action,
+		     int *a_len, struct nfp_flower_pedit_acts *set_act,
+		     u32 *csum_updated)
+{
+	struct flow_rule *rule = tc_cls_flower_offload_flow_rule(flow);
 	size_t act_size = 0;
 	u8 ip_proto = 0;
-	u32 offset;
-	int err;
-
-	memset(&set_ip6_tc_hl_fl, 0, sizeof(set_ip6_tc_hl_fl));
-	memset(&set_ip_ttl_tos, 0, sizeof(set_ip_ttl_tos));
-	memset(&set_ip6_dst, 0, sizeof(set_ip6_dst));
-	memset(&set_ip6_src, 0, sizeof(set_ip6_src));
-	memset(&set_ip_addr, 0, sizeof(set_ip_addr));
-	memset(&set_tport, 0, sizeof(set_tport));
-	memset(&set_eth, 0, sizeof(set_eth));
-
-	htype = act->mangle.htype;
-	offset = act->mangle.offset;
-
-	switch (htype) {
-	case TCA_PEDIT_KEY_EX_HDR_TYPE_ETH:
-		err = nfp_fl_set_eth(act, offset, &set_eth);
-		break;
-	case TCA_PEDIT_KEY_EX_HDR_TYPE_IP4:
-		err = nfp_fl_set_ip4(act, offset, &set_ip_addr,
-				     &set_ip_ttl_tos);
-		break;
-	case TCA_PEDIT_KEY_EX_HDR_TYPE_IP6:
-		err = nfp_fl_set_ip6(act, offset, &set_ip6_dst,
-				     &set_ip6_src, &set_ip6_tc_hl_fl);
-		break;
-	case TCA_PEDIT_KEY_EX_HDR_TYPE_TCP:
-		err = nfp_fl_set_tport(act, offset, &set_tport,
-				       NFP_FL_ACTION_OPCODE_SET_TCP);
-		break;
-	case TCA_PEDIT_KEY_EX_HDR_TYPE_UDP:
-		err = nfp_fl_set_tport(act, offset, &set_tport,
-				       NFP_FL_ACTION_OPCODE_SET_UDP);
-		break;
-	default:
-		return -EOPNOTSUPP;
-	}
-	if (err)
-		return err;
 
 	if (flow_rule_match_key(rule, FLOW_DISSECTOR_KEY_BASIC)) {
 		struct flow_match_basic match;
@@ -644,77 +607,82 @@ nfp_fl_pedit(const struct flow_action_entry *act,
 		ip_proto = match.key->ip_proto;
 	}
 
-	if (set_eth.head.len_lw) {
-		act_size = sizeof(set_eth);
-		memcpy(nfp_action, &set_eth, act_size);
+	if (set_act->set_eth.head.len_lw) {
+		act_size = sizeof(set_act->set_eth);
+		memcpy(nfp_action, &set_act->set_eth, act_size);
 		*a_len += act_size;
 	}
-	if (set_ip_ttl_tos.head.len_lw) {
+
+	if (set_act->set_ip_ttl_tos.head.len_lw) {
 		nfp_action += act_size;
-		act_size = sizeof(set_ip_ttl_tos);
-		memcpy(nfp_action, &set_ip_ttl_tos, act_size);
+		act_size = sizeof(set_act->set_ip_ttl_tos);
+		memcpy(nfp_action, &set_act->set_ip_ttl_tos, act_size);
 		*a_len += act_size;
 
 		/* Hardware will automatically fix IPv4 and TCP/UDP checksum. */
 		*csum_updated |= TCA_CSUM_UPDATE_FLAG_IPV4HDR |
 				nfp_fl_csum_l4_to_flag(ip_proto);
 	}
-	if (set_ip_addr.head.len_lw) {
+
+	if (set_act->set_ip_addr.head.len_lw) {
 		nfp_action += act_size;
-		act_size = sizeof(set_ip_addr);
-		memcpy(nfp_action, &set_ip_addr, act_size);
+		act_size = sizeof(set_act->set_ip_addr);
+		memcpy(nfp_action, &set_act->set_ip_addr, act_size);
 		*a_len += act_size;
 
 		/* Hardware will automatically fix IPv4 and TCP/UDP checksum. */
 		*csum_updated |= TCA_CSUM_UPDATE_FLAG_IPV4HDR |
 				nfp_fl_csum_l4_to_flag(ip_proto);
 	}
-	if (set_ip6_tc_hl_fl.head.len_lw) {
+
+	if (set_act->set_ip6_tc_hl_fl.head.len_lw) {
 		nfp_action += act_size;
-		act_size = sizeof(set_ip6_tc_hl_fl);
-		memcpy(nfp_action, &set_ip6_tc_hl_fl, act_size);
+		act_size = sizeof(set_act->set_ip6_tc_hl_fl);
+		memcpy(nfp_action, &set_act->set_ip6_tc_hl_fl, act_size);
 		*a_len += act_size;
 
 		/* Hardware will automatically fix TCP/UDP checksum. */
 		*csum_updated |= nfp_fl_csum_l4_to_flag(ip_proto);
 	}
-	if (set_ip6_dst.head.len_lw && set_ip6_src.head.len_lw) {
+
+	if (set_act->set_ip6_dst.head.len_lw &&
+	    set_act->set_ip6_src.head.len_lw) {
 		/* TC compiles set src and dst IPv6 address as a single action,
 		 * the hardware requires this to be 2 separate actions.
 		 */
 		nfp_action += act_size;
-		act_size = sizeof(set_ip6_src);
-		memcpy(nfp_action, &set_ip6_src, act_size);
+		act_size = sizeof(set_act->set_ip6_src);
+		memcpy(nfp_action, &set_act->set_ip6_src, act_size);
 		*a_len += act_size;
 
-		act_size = sizeof(set_ip6_dst);
-		memcpy(&nfp_action[sizeof(set_ip6_src)], &set_ip6_dst,
-		       act_size);
+		act_size = sizeof(set_act->set_ip6_dst);
+		memcpy(&nfp_action[sizeof(set_act->set_ip6_src)],
+		       &set_act->set_ip6_dst, act_size);
 		*a_len += act_size;
 
 		/* Hardware will automatically fix TCP/UDP checksum. */
 		*csum_updated |= nfp_fl_csum_l4_to_flag(ip_proto);
-	} else if (set_ip6_dst.head.len_lw) {
+	} else if (set_act->set_ip6_dst.head.len_lw) {
 		nfp_action += act_size;
-		act_size = sizeof(set_ip6_dst);
-		memcpy(nfp_action, &set_ip6_dst, act_size);
+		act_size = sizeof(set_act->set_ip6_dst);
+		memcpy(nfp_action, &set_act->set_ip6_dst, act_size);
 		*a_len += act_size;
 
 		/* Hardware will automatically fix TCP/UDP checksum. */
 		*csum_updated |= nfp_fl_csum_l4_to_flag(ip_proto);
-	} else if (set_ip6_src.head.len_lw) {
+	} else if (set_act->set_ip6_src.head.len_lw) {
 		nfp_action += act_size;
-		act_size = sizeof(set_ip6_src);
-		memcpy(nfp_action, &set_ip6_src, act_size);
+		act_size = sizeof(set_act->set_ip6_src);
+		memcpy(nfp_action, &set_act->set_ip6_src, act_size);
 		*a_len += act_size;
 
 		/* Hardware will automatically fix TCP/UDP checksum. */
 		*csum_updated |= nfp_fl_csum_l4_to_flag(ip_proto);
 	}
-	if (set_tport.head.len_lw) {
+	if (set_act->set_tport.head.len_lw) {
 		nfp_action += act_size;
-		act_size = sizeof(set_tport);
-		memcpy(nfp_action, &set_tport, act_size);
+		act_size = sizeof(set_act->set_tport);
+		memcpy(nfp_action, &set_act->set_tport, act_size);
 		*a_len += act_size;
 
 		/* Hardware will automatically fix TCP/UDP checksum. */
@@ -725,7 +693,40 @@ nfp_fl_pedit(const struct flow_action_entry *act,
 }
 
 static int
-nfp_flower_output_action(struct nfp_app *app, const struct flow_action_entry *act,
+nfp_fl_pedit(const struct flow_action_entry *act,
+	     struct tc_cls_flower_offload *flow, char *nfp_action, int *a_len,
+	     u32 *csum_updated, struct nfp_flower_pedit_acts *set_act)
+{
+	enum flow_action_mangle_base htype;
+	u32 offset;
+
+	htype = act->mangle.htype;
+	offset = act->mangle.offset;
+
+	switch (htype) {
+	case TCA_PEDIT_KEY_EX_HDR_TYPE_ETH:
+		return nfp_fl_set_eth(act, offset, &set_act->set_eth);
+	case TCA_PEDIT_KEY_EX_HDR_TYPE_IP4:
+		return nfp_fl_set_ip4(act, offset, &set_act->set_ip_addr,
+				      &set_act->set_ip_ttl_tos);
+	case TCA_PEDIT_KEY_EX_HDR_TYPE_IP6:
+		return nfp_fl_set_ip6(act, offset, &set_act->set_ip6_dst,
+				      &set_act->set_ip6_src,
+				      &set_act->set_ip6_tc_hl_fl);
+	case TCA_PEDIT_KEY_EX_HDR_TYPE_TCP:
+		return nfp_fl_set_tport(act, offset, &set_act->set_tport,
+					NFP_FL_ACTION_OPCODE_SET_TCP);
+	case TCA_PEDIT_KEY_EX_HDR_TYPE_UDP:
+		return nfp_fl_set_tport(act, offset, &set_act->set_tport,
+					NFP_FL_ACTION_OPCODE_SET_UDP);
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
+static int
+nfp_flower_output_action(struct nfp_app *app,
+			 const struct flow_action_entry *act,
 			 struct nfp_fl_payload *nfp_fl, int *a_len,
 			 struct net_device *netdev, bool last,
 			 enum nfp_flower_tun_type *tun_type, int *tun_out_cnt,
@@ -775,7 +776,8 @@ nfp_flower_loop_action(struct nfp_app *app, const struct flow_action_entry *act,
 		       struct nfp_fl_payload *nfp_fl, int *a_len,
 		       struct net_device *netdev,
 		       enum nfp_flower_tun_type *tun_type, int *tun_out_cnt,
-		       int *out_cnt, u32 *csum_updated)
+		       int *out_cnt, u32 *csum_updated,
+		       struct nfp_flower_pedit_acts *set_act)
 {
 	struct nfp_fl_set_ipv4_udp_tun *set_tun;
 	struct nfp_fl_pre_tunnel *pre_tun;
@@ -860,7 +862,7 @@ nfp_flower_loop_action(struct nfp_app *app, const struct flow_action_entry *act,
 		return 0;
 	case FLOW_ACTION_MANGLE:
 		if (nfp_fl_pedit(act, flow, &nfp_fl->action_data[*a_len],
-				 a_len, csum_updated))
+				 a_len, csum_updated, set_act))
 			return -EOPNOTSUPP;
 		break;
 	case FLOW_ACTION_CSUM:
@@ -880,12 +882,49 @@ nfp_flower_loop_action(struct nfp_app *app, const struct flow_action_entry *act,
 	return 0;
 }
 
+static bool nfp_fl_check_mangle_start(struct flow_action *flow_act,
+				      int current_act_idx)
+{
+	struct flow_action_entry current_act;
+	struct flow_action_entry prev_act;
+
+	current_act = flow_act->entries[current_act_idx];
+	if (current_act.id != FLOW_ACTION_MANGLE)
+		return false;
+
+	if (current_act_idx == 0)
+		return true;
+
+	prev_act = flow_act->entries[current_act_idx - 1];
+
+	return prev_act.id != FLOW_ACTION_MANGLE;
+}
+
+static bool nfp_fl_check_mangle_end(struct flow_action *flow_act,
+				    int current_act_idx)
+{
+	struct flow_action_entry current_act;
+	struct flow_action_entry next_act;
+
+	current_act = flow_act->entries[current_act_idx];
+	if (current_act.id != FLOW_ACTION_MANGLE)
+		return false;
+
+	if (current_act_idx == flow_act->num_entries)
+		return true;
+
+	next_act = flow_act->entries[current_act_idx + 1];
+
+	return next_act.id != FLOW_ACTION_MANGLE;
+}
+
 int nfp_flower_compile_action(struct nfp_app *app,
 			      struct tc_cls_flower_offload *flow,
 			      struct net_device *netdev,
 			      struct nfp_fl_payload *nfp_flow)
 {
 	int act_len, act_cnt, err, tun_out_cnt, out_cnt, i;
+	struct nfp_flower_pedit_acts set_act;
 	enum nfp_flower_tun_type tun_type;
 	struct flow_action_entry *act;
 	u32 csum_updated = 0;
@@ -899,12 +938,18 @@ int nfp_flower_compile_action(struct nfp_app *app,
 	out_cnt = 0;
 
 	flow_action_for_each(i, act, &flow->rule->action) {
+		if (nfp_fl_check_mangle_start(&flow->rule->action, i))
+			memset(&set_act, 0, sizeof(set_act));
 		err = nfp_flower_loop_action(app, act, flow, nfp_flow, &act_len,
 					     netdev, &tun_type, &tun_out_cnt,
-					     &out_cnt, &csum_updated);
+					     &out_cnt, &csum_updated, &set_act);
 		if (err)
 			return err;
 		act_cnt++;
+		if (nfp_fl_check_mangle_end(&flow->rule->action, i))
+			nfp_fl_commit_mangle(flow,
+					     &nfp_flow->action_data[act_len],
+					     &act_len, &set_act, &csum_updated);
 	}
 
 	/* We optimise when the action list is small, this can unfortunately
diff --git a/drivers/net/ethernet/netronome/nfp/flower/cmsg.c b/drivers/net/ethernet/netronome/nfp/flower/cmsg.c
index cf9e1118ee8f..d5bbe3d6048b 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/cmsg.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/cmsg.c
@@ -159,7 +159,7 @@ nfp_flower_cmsg_portmod_rx(struct nfp_app *app, struct sk_buff *skb)
 
 	rtnl_lock();
 	rcu_read_lock();
-	netdev = nfp_app_repr_get(app, be32_to_cpu(msg->portnum));
+	netdev = nfp_app_dev_get(app, be32_to_cpu(msg->portnum), NULL);
 	rcu_read_unlock();
 	if (!netdev) {
 		nfp_flower_cmsg_warn(app, "ctrl msg for unknown port 0x%08x\n",
@@ -192,7 +192,7 @@ nfp_flower_cmsg_portreify_rx(struct nfp_app *app, struct sk_buff *skb)
 	msg = nfp_flower_cmsg_get_data(skb);
 
 	rcu_read_lock();
-	exists = !!nfp_app_repr_get(app, be32_to_cpu(msg->portnum));
+	exists = !!nfp_app_dev_get(app, be32_to_cpu(msg->portnum), NULL);
 	rcu_read_unlock();
 	if (!exists) {
 		nfp_flower_cmsg_warn(app, "ctrl msg for unknown port 0x%08x\n",
@@ -205,6 +205,50 @@ nfp_flower_cmsg_portreify_rx(struct nfp_app *app, struct sk_buff *skb)
 }
 
 static void
+nfp_flower_cmsg_merge_hint_rx(struct nfp_app *app, struct sk_buff *skb)
+{
+	unsigned int msg_len = nfp_flower_cmsg_get_data_len(skb);
+	struct nfp_flower_cmsg_merge_hint *msg;
+	struct nfp_fl_payload *sub_flows[2];
+	int err, i, flow_cnt;
+
+	msg = nfp_flower_cmsg_get_data(skb);
+	/* msg->count starts at 0 and always assumes at least 1 entry. */
+	flow_cnt = msg->count + 1;
+
+	if (msg_len < struct_size(msg, flow, flow_cnt)) {
+		nfp_flower_cmsg_warn(app, "Merge hint ctrl msg too short - %d bytes but expect %zd\n",
+				     msg_len, struct_size(msg, flow, flow_cnt));
+		return;
+	}
+
+	if (flow_cnt != 2) {
+		nfp_flower_cmsg_warn(app, "Merge hint contains %d flows - two are expected\n",
+				     flow_cnt);
+		return;
+	}
+
+	rtnl_lock();
+	for (i = 0; i < flow_cnt; i++) {
+		u32 ctx = be32_to_cpu(msg->flow[i].host_ctx);
+
+		sub_flows[i] = nfp_flower_get_fl_payload_from_ctx(app, ctx);
+		if (!sub_flows[i]) {
+			nfp_flower_cmsg_warn(app, "Invalid flow in merge hint\n");
+			goto err_rtnl_unlock;
+		}
+	}
+
+	err = nfp_flower_merge_offloaded_flows(app, sub_flows[0], sub_flows[1]);
+	/* Only warn on memory fail. Hint veto will not break functionality. */
+	if (err == -ENOMEM)
+		nfp_flower_cmsg_warn(app, "Flow merge memory fail.\n");
+
+err_rtnl_unlock:
+	rtnl_unlock();
+}
+
+static void
 nfp_flower_cmsg_process_one_rx(struct nfp_app *app, struct sk_buff *skb)
 {
 	struct nfp_flower_priv *app_priv = app->priv;
@@ -222,12 +266,21 @@ nfp_flower_cmsg_process_one_rx(struct nfp_app *app, struct sk_buff *skb)
 	case NFP_FLOWER_CMSG_TYPE_PORT_MOD:
 		nfp_flower_cmsg_portmod_rx(app, skb);
 		break;
+	case NFP_FLOWER_CMSG_TYPE_MERGE_HINT:
+		if (app_priv->flower_ext_feats & NFP_FL_FEATS_FLOW_MERGE) {
+			nfp_flower_cmsg_merge_hint_rx(app, skb);
+			break;
+		}
+		goto err_default;
 	case NFP_FLOWER_CMSG_TYPE_NO_NEIGH:
 		nfp_tunnel_request_route(app, skb);
 		break;
 	case NFP_FLOWER_CMSG_TYPE_ACTIVE_TUNS:
 		nfp_tunnel_keep_alive(app, skb);
 		break;
+	case NFP_FLOWER_CMSG_TYPE_QOS_STATS:
+		nfp_flower_stats_rlim_reply(app, skb);
+		break;
 	case NFP_FLOWER_CMSG_TYPE_LAG_CONFIG:
 		if (app_priv->flower_ext_feats & NFP_FL_FEATS_LAG) {
 			skb_stored = nfp_flower_lag_unprocessed_msg(app, skb);
@@ -235,6 +288,7 @@ nfp_flower_cmsg_process_one_rx(struct nfp_app *app, struct sk_buff *skb)
 		}
 		/* fall through */
 	default:
+err_default:
 		nfp_flower_cmsg_warn(app, "Cannot handle invalid repr control type %u\n",
 				     type);
 		goto out;
diff --git a/drivers/net/ethernet/netronome/nfp/flower/cmsg.h b/drivers/net/ethernet/netronome/nfp/flower/cmsg.h
index 0ed51e79db00..537f7fc19584 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/cmsg.h
+++ b/drivers/net/ethernet/netronome/nfp/flower/cmsg.h
@@ -402,11 +402,13 @@ struct nfp_flower_cmsg_hdr {
 /* Types defined for port related control messages  */
 enum nfp_flower_cmsg_type_port {
 	NFP_FLOWER_CMSG_TYPE_FLOW_ADD =		0,
+	NFP_FLOWER_CMSG_TYPE_FLOW_MOD =		1,
 	NFP_FLOWER_CMSG_TYPE_FLOW_DEL =		2,
 	NFP_FLOWER_CMSG_TYPE_LAG_CONFIG =	4,
 	NFP_FLOWER_CMSG_TYPE_PORT_REIFY =	6,
 	NFP_FLOWER_CMSG_TYPE_MAC_REPR =		7,
 	NFP_FLOWER_CMSG_TYPE_PORT_MOD =		8,
+	NFP_FLOWER_CMSG_TYPE_MERGE_HINT =	9,
 	NFP_FLOWER_CMSG_TYPE_NO_NEIGH =		10,
 	NFP_FLOWER_CMSG_TYPE_TUN_MAC =		11,
 	NFP_FLOWER_CMSG_TYPE_ACTIVE_TUNS =	12,
@@ -414,6 +416,9 @@ enum nfp_flower_cmsg_type_port {
 	NFP_FLOWER_CMSG_TYPE_TUN_IPS =		14,
 	NFP_FLOWER_CMSG_TYPE_FLOW_STATS =	15,
 	NFP_FLOWER_CMSG_TYPE_PORT_ECHO =	16,
+	NFP_FLOWER_CMSG_TYPE_QOS_MOD =		18,
+	NFP_FLOWER_CMSG_TYPE_QOS_DEL =		19,
+	NFP_FLOWER_CMSG_TYPE_QOS_STATS =	20,
 	NFP_FLOWER_CMSG_TYPE_MAX =		32,
 };
 
@@ -451,6 +456,16 @@ struct nfp_flower_cmsg_portreify {
 
 #define NFP_FLOWER_CMSG_PORTREIFY_INFO_EXIST	BIT(0)
 
+/* NFP_FLOWER_CMSG_TYPE_FLOW_MERGE_HINT */
+struct nfp_flower_cmsg_merge_hint {
+	u8 reserved[3];
+	u8 count;
+	struct {
+		__be32 host_ctx;
+		__be64 host_cookie;
+	} __packed flow[0];
+};
+
 enum nfp_flower_cmsg_port_type {
 	NFP_FLOWER_CMSG_PORT_TYPE_UNSPEC =	0x0,
 	NFP_FLOWER_CMSG_PORT_TYPE_PHYS_PORT =	0x1,
@@ -473,6 +488,13 @@ enum nfp_flower_cmsg_port_vnic_type {
 #define NFP_FLOWER_CMSG_PORT_PCIE_Q		GENMASK(5, 0)
 #define NFP_FLOWER_CMSG_PORT_PHYS_PORT_NUM	GENMASK(7, 0)
 
+static inline u32 nfp_flower_internal_port_get_port_id(u8 internal_port)
+{
+	return FIELD_PREP(NFP_FLOWER_CMSG_PORT_PHYS_PORT_NUM, internal_port) |
+		FIELD_PREP(NFP_FLOWER_CMSG_PORT_TYPE,
+			   NFP_FLOWER_CMSG_PORT_TYPE_OTHER_PORT);
+}
+
 static inline u32 nfp_flower_cmsg_phys_port(u8 phys_port)
 {
 	return FIELD_PREP(NFP_FLOWER_CMSG_PORT_PHYS_PORT_NUM, phys_port) |
diff --git a/drivers/net/ethernet/netronome/nfp/flower/main.c b/drivers/net/ethernet/netronome/nfp/flower/main.c
index 408089133599..eb846133943b 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/main.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/main.c
@@ -22,6 +22,9 @@
 
 #define NFP_FLOWER_ALLOWED_VER 0x0001000000010000UL
 
+#define NFP_MIN_INT_PORT_ID	1
+#define NFP_MAX_INT_PORT_ID	256
+
 static const char *nfp_flower_extra_cap(struct nfp_app *app, struct nfp_net *nn)
 {
 	return "FLOWER";
@@ -32,6 +35,113 @@ static enum devlink_eswitch_mode eswitch_mode_get(struct nfp_app *app)
 	return DEVLINK_ESWITCH_MODE_SWITCHDEV;
 }
 
+static int
+nfp_flower_lookup_internal_port_id(struct nfp_flower_priv *priv,
+				   struct net_device *netdev)
+{
+	struct net_device *entry;
+	int i, id = 0;
+
+	rcu_read_lock();
+	idr_for_each_entry(&priv->internal_ports.port_ids, entry, i)
+		if (entry == netdev) {
+			id = i;
+			break;
+		}
+	rcu_read_unlock();
+
+	return id;
+}
+
+static int
+nfp_flower_get_internal_port_id(struct nfp_app *app, struct net_device *netdev)
+{
+	struct nfp_flower_priv *priv = app->priv;
+	int id;
+
+	id = nfp_flower_lookup_internal_port_id(priv, netdev);
+	if (id > 0)
+		return id;
+
+	idr_preload(GFP_ATOMIC);
+	spin_lock_bh(&priv->internal_ports.lock);
+	id = idr_alloc(&priv->internal_ports.port_ids, netdev,
+		       NFP_MIN_INT_PORT_ID, NFP_MAX_INT_PORT_ID, GFP_ATOMIC);
+	spin_unlock_bh(&priv->internal_ports.lock);
+	idr_preload_end();
+
+	return id;
+}
+
+u32 nfp_flower_get_port_id_from_netdev(struct nfp_app *app,
+				       struct net_device *netdev)
+{
+	int ext_port;
+
+	if (nfp_netdev_is_nfp_repr(netdev)) {
+		return nfp_repr_get_port_id(netdev);
+	} else if (nfp_flower_internal_port_can_offload(app, netdev)) {
+		ext_port = nfp_flower_get_internal_port_id(app, netdev);
+		if (ext_port < 0)
+			return 0;
+
+		return nfp_flower_internal_port_get_port_id(ext_port);
+	}
+
+	return 0;
+}
+
+static struct net_device *
+nfp_flower_get_netdev_from_internal_port_id(struct nfp_app *app, int port_id)
+{
+	struct nfp_flower_priv *priv = app->priv;
+	struct net_device *netdev;
+
+	rcu_read_lock();
+	netdev = idr_find(&priv->internal_ports.port_ids, port_id);
+	rcu_read_unlock();
+
+	return netdev;
+}
+
+static void
+nfp_flower_free_internal_port_id(struct nfp_app *app, struct net_device *netdev)
+{
+	struct nfp_flower_priv *priv = app->priv;
+	int id;
+
+	id = nfp_flower_lookup_internal_port_id(priv, netdev);
+	if (!id)
+		return;
+
+	spin_lock_bh(&priv->internal_ports.lock);
+	idr_remove(&priv->internal_ports.port_ids, id);
+	spin_unlock_bh(&priv->internal_ports.lock);
+}
+
+static int
+nfp_flower_internal_port_event_handler(struct nfp_app *app,
+				       struct net_device *netdev,
+				       unsigned long event)
+{
+	if (event == NETDEV_UNREGISTER &&
+	    nfp_flower_internal_port_can_offload(app, netdev))
+		nfp_flower_free_internal_port_id(app, netdev);
+
+	return NOTIFY_OK;
+}
+
+static void nfp_flower_internal_port_init(struct nfp_flower_priv *priv)
+{
+	spin_lock_init(&priv->internal_ports.lock);
+	idr_init(&priv->internal_ports.port_ids);
+}
+
+static void nfp_flower_internal_port_cleanup(struct nfp_flower_priv *priv)
+{
+	idr_destroy(&priv->internal_ports.port_ids);
+}
+
 static struct nfp_flower_non_repr_priv *
 nfp_flower_non_repr_priv_lookup(struct nfp_app *app, struct net_device *netdev)
 {
@@ -119,12 +229,21 @@ nfp_flower_repr_get_type_and_port(struct nfp_app *app, u32 port_id, u8 *port)
 }
 
 static struct net_device *
-nfp_flower_repr_get(struct nfp_app *app, u32 port_id)
+nfp_flower_dev_get(struct nfp_app *app, u32 port_id, bool *redir_egress)
 {
 	enum nfp_repr_type repr_type;
 	struct nfp_reprs *reprs;
 	u8 port = 0;
 
+	/* Check if the port is internal. */
+	if (FIELD_GET(NFP_FLOWER_CMSG_PORT_TYPE, port_id) ==
+	    NFP_FLOWER_CMSG_PORT_TYPE_OTHER_PORT) {
+		if (redir_egress)
+			*redir_egress = true;
+		port = FIELD_GET(NFP_FLOWER_CMSG_PORT_PHYS_PORT_NUM, port_id);
+		return nfp_flower_get_netdev_from_internal_port_id(app, port);
+	}
+
 	repr_type = nfp_flower_repr_get_type_and_port(app, port_id, &port);
 	if (repr_type > NFP_REPR_TYPE_MAX)
 		return NULL;
@@ -641,11 +760,33 @@ static int nfp_flower_init(struct nfp_app *app)
 		goto err_cleanup_metadata;
 	}
 
+	if (app_priv->flower_ext_feats & NFP_FL_FEATS_FLOW_MOD) {
+		/* Tell the firmware that the driver supports flow merging. */
+		err = nfp_rtsym_write_le(app->pf->rtbl,
+					 "_abi_flower_merge_hint_enable", 1);
+		if (!err) {
+			app_priv->flower_ext_feats |= NFP_FL_FEATS_FLOW_MERGE;
+			nfp_flower_internal_port_init(app_priv);
+		} else if (err == -ENOENT) {
+			nfp_warn(app->cpp, "Flow merge not supported by FW.\n");
+		} else {
+			goto err_lag_clean;
+		}
+	} else {
+		nfp_warn(app->cpp, "Flow mod/merge not supported by FW.\n");
+	}
+
+	if (app_priv->flower_ext_feats & NFP_FL_FEATS_VF_RLIM)
+		nfp_flower_qos_init(app);
+
 	INIT_LIST_HEAD(&app_priv->indr_block_cb_priv);
 	INIT_LIST_HEAD(&app_priv->non_repr_priv);
 
 	return 0;
 
+err_lag_clean:
+	if (app_priv->flower_ext_feats & NFP_FL_FEATS_LAG)
+		nfp_flower_lag_cleanup(&app_priv->nfp_lag);
 err_cleanup_metadata:
 	nfp_flower_metadata_cleanup(app);
 err_free_app_priv:
@@ -661,9 +802,15 @@ static void nfp_flower_clean(struct nfp_app *app)
 	skb_queue_purge(&app_priv->cmsg_skbs_low);
 	flush_work(&app_priv->cmsg_work);
 
+	if (app_priv->flower_ext_feats & NFP_FL_FEATS_VF_RLIM)
+		nfp_flower_qos_cleanup(app);
+
 	if (app_priv->flower_ext_feats & NFP_FL_FEATS_LAG)
 		nfp_flower_lag_cleanup(&app_priv->nfp_lag);
 
+	if (app_priv->flower_ext_feats & NFP_FL_FEATS_FLOW_MERGE)
+		nfp_flower_internal_port_cleanup(app_priv);
+
 	nfp_flower_metadata_cleanup(app);
 	vfree(app->priv);
 	app->priv = NULL;
@@ -762,6 +909,10 @@ nfp_flower_netdev_event(struct nfp_app *app, struct net_device *netdev,
 	if (ret & NOTIFY_STOP_MASK)
 		return ret;
 
+	ret = nfp_flower_internal_port_event_handler(app, netdev, event);
+	if (ret & NOTIFY_STOP_MASK)
+		return ret;
+
 	return nfp_tunnel_mac_event_handler(app, netdev, event, ptr);
 }
 
@@ -800,7 +951,7 @@ const struct nfp_app_type app_flower = {
 	.sriov_disable	= nfp_flower_sriov_disable,
 
 	.eswitch_mode_get  = eswitch_mode_get,
-	.repr_get	= nfp_flower_repr_get,
+	.dev_get	= nfp_flower_dev_get,
 
 	.setup_tc	= nfp_flower_setup_tc,
 };
diff --git a/drivers/net/ethernet/netronome/nfp/flower/main.h b/drivers/net/ethernet/netronome/nfp/flower/main.h
index c0945a5fd1a4..40957a8dbfe6 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/main.h
+++ b/drivers/net/ethernet/netronome/nfp/flower/main.h
@@ -5,6 +5,7 @@
 #define __NFP_FLOWER_H__ 1
 
 #include "cmsg.h"
+#include "../nfp_net.h"
 
 #include <linux/circ_buf.h>
 #include <linux/hashtable.h>
@@ -34,14 +35,14 @@ struct nfp_app;
 #define NFP_FL_MASK_REUSE_TIME_NS	40000
 #define NFP_FL_MASK_ID_LOCATION		1
 
-#define NFP_FL_VXLAN_PORT		4789
-#define NFP_FL_GENEVE_PORT		6081
-
 /* Extra features bitmap. */
 #define NFP_FL_FEATS_GENEVE		BIT(0)
 #define NFP_FL_NBI_MTU_SETTING		BIT(1)
 #define NFP_FL_FEATS_GENEVE_OPT		BIT(2)
 #define NFP_FL_FEATS_VLAN_PCP		BIT(3)
+#define NFP_FL_FEATS_VF_RLIM		BIT(4)
+#define NFP_FL_FEATS_FLOW_MOD		BIT(5)
+#define NFP_FL_FEATS_FLOW_MERGE		BIT(30)
 #define NFP_FL_FEATS_LAG		BIT(31)
 
 struct nfp_fl_mask_id {
@@ -118,6 +119,16 @@ struct nfp_fl_lag {
 };
 
 /**
+ * struct nfp_fl_internal_ports - Flower APP priv data for additional ports
+ * @port_ids:	Assignment of ids to any additional ports
+ * @lock:	Lock for extra ports list
+ */
+struct nfp_fl_internal_ports {
+	struct idr port_ids;
+	spinlock_t lock;
+};
+
+/**
  * struct nfp_flower_priv - Flower APP per-vNIC priv data
  * @app:		Back pointer to app
  * @nn:			Pointer to vNIC
@@ -131,6 +142,7 @@ struct nfp_fl_lag {
  * @flow_table:		Hash table used to store flower rules
  * @stats:		Stored stats updates for flower rules
  * @stats_lock:		Lock for flower rule stats updates
+ * @stats_ctx_table:	Hash table to map stats contexts to its flow rule
  * @cmsg_work:		Workqueue for control messages processing
  * @cmsg_skbs_high:	List of higher priority skbs for control message
  *			processing
@@ -146,6 +158,10 @@ struct nfp_fl_lag {
  * @non_repr_priv:	List of offloaded non-repr ports and their priv data
  * @active_mem_unit:	Current active memory unit for flower rules
  * @total_mem_units:	Total number of available memory units for flower rules
+ * @internal_ports:	Internal port ids used in offloaded rules
+ * @qos_stats_work:	Workqueue for qos stats processing
+ * @qos_rate_limiters:	Current active qos rate limiters
+ * @qos_stats_lock:	Lock on qos stats updates
  */
 struct nfp_flower_priv {
 	struct nfp_app *app;
@@ -160,6 +176,7 @@ struct nfp_flower_priv {
 	struct rhashtable flow_table;
 	struct nfp_fl_stats *stats;
 	spinlock_t stats_lock; /* lock stats */
+	struct rhashtable stats_ctx_table;
 	struct work_struct cmsg_work;
 	struct sk_buff_head cmsg_skbs_high;
 	struct sk_buff_head cmsg_skbs_low;
@@ -172,6 +189,24 @@ struct nfp_flower_priv {
 	struct list_head non_repr_priv;
 	unsigned int active_mem_unit;
 	unsigned int total_mem_units;
+	struct nfp_fl_internal_ports internal_ports;
+	struct delayed_work qos_stats_work;
+	unsigned int qos_rate_limiters;
+	spinlock_t qos_stats_lock; /* Protect the qos stats */
+};
+
+/**
+ * struct nfp_fl_qos - Flower APP priv data for quality of service
+ * @netdev_port_id:	NFP port number of repr with qos info
+ * @curr_stats:		Currently stored stats updates for qos info
+ * @prev_stats:		Previously stored updates for qos info
+ * @last_update:	Stored time when last stats were updated
+ */
+struct nfp_fl_qos {
+	u32 netdev_port_id;
+	struct nfp_stat_pair curr_stats;
+	struct nfp_stat_pair prev_stats;
+	u64 last_update;
 };
 
 /**
@@ -180,14 +215,18 @@ struct nfp_flower_priv {
  * @lag_port_flags:	Extended port flags to record lag state of repr
  * @mac_offloaded:	Flag indicating a MAC address is offloaded for repr
  * @offloaded_mac_addr:	MAC address that has been offloaded for repr
+ * @block_shared:	Flag indicating if offload applies to shared blocks
  * @mac_list:		List entry of reprs that share the same offloaded MAC
+ * @qos_table:		Stored info on filters implementing qos
  */
 struct nfp_flower_repr_priv {
 	struct nfp_repr *nfp_repr;
 	unsigned long lag_port_flags;
 	bool mac_offloaded;
 	u8 offloaded_mac_addr[ETH_ALEN];
+	bool block_shared;
 	struct list_head mac_list;
+	struct nfp_fl_qos qos_table;
 };
 
 /**
@@ -239,6 +278,25 @@ struct nfp_fl_payload {
 	char *unmasked_data;
 	char *mask_data;
 	char *action_data;
+	struct list_head linked_flows;
+	bool in_hw;
+};
+
+struct nfp_fl_payload_link {
+	/* A link contains a pointer to a merge flow and an associated sub_flow.
+	 * Each merge flow will feature in 2 links to its underlying sub_flows.
+	 * A sub_flow will have at least 1 link to a merge flow or more if it
+	 * has been used to create multiple merge flows.
+	 *
+	 * For a merge flow, 'linked_flows' in its nfp_fl_payload struct lists
+	 * all links to sub_flows (sub_flow.flow) via merge.list.
+	 * For a sub_flow, 'linked_flows' gives all links to merge flows it has
+	 * formed (merge_flow.flow) via sub_flow.list.
+	 */
+	struct {
+		struct list_head list;
+		struct nfp_fl_payload *flow;
+	} merge_flow, sub_flow;
 };
 
 extern const struct rhashtable_params nfp_flower_table_params;
@@ -250,12 +308,40 @@ struct nfp_fl_stats_frame {
 	__be64 stats_cookie;
 };
 
+static inline bool
+nfp_flower_internal_port_can_offload(struct nfp_app *app,
+				     struct net_device *netdev)
+{
+	struct nfp_flower_priv *app_priv = app->priv;
+
+	if (!(app_priv->flower_ext_feats & NFP_FL_FEATS_FLOW_MERGE))
+		return false;
+	if (!netdev->rtnl_link_ops)
+		return false;
+	if (!strcmp(netdev->rtnl_link_ops->kind, "openvswitch"))
+		return true;
+
+	return false;
+}
+
+/* The address of the merged flow acts as its cookie.
+ * Cookies supplied to us by TC flower are also addresses to allocated
+ * memory and thus this scheme should not generate any collisions.
+ */
+static inline bool nfp_flower_is_merge_flow(struct nfp_fl_payload *flow_pay)
+{
+	return flow_pay->tc_flower_cookie == (unsigned long)flow_pay;
+}
+
 int nfp_flower_metadata_init(struct nfp_app *app, u64 host_ctx_count,
 			     unsigned int host_ctx_split);
 void nfp_flower_metadata_cleanup(struct nfp_app *app);
 
 int nfp_flower_setup_tc(struct nfp_app *app, struct net_device *netdev,
 			enum tc_setup_type type, void *type_data);
+int nfp_flower_merge_offloaded_flows(struct nfp_app *app,
+				     struct nfp_fl_payload *sub_flow1,
+				     struct nfp_fl_payload *sub_flow2);
 int nfp_flower_compile_flow_match(struct nfp_app *app,
 				  struct tc_cls_flower_offload *flow,
 				  struct nfp_fl_key_ls *key_ls,
@@ -270,6 +356,8 @@ int nfp_compile_flow_metadata(struct nfp_app *app,
 			      struct tc_cls_flower_offload *flow,
 			      struct nfp_fl_payload *nfp_flow,
 			      struct net_device *netdev);
+void __nfp_modify_flow_metadata(struct nfp_flower_priv *priv,
+				struct nfp_fl_payload *nfp_flow);
 int nfp_modify_flow_metadata(struct nfp_app *app,
 			     struct nfp_fl_payload *nfp_flow);
 
@@ -277,6 +365,8 @@ struct nfp_fl_payload *
 nfp_flower_search_fl_table(struct nfp_app *app, unsigned long tc_flower_cookie,
 			   struct net_device *netdev);
 struct nfp_fl_payload *
+nfp_flower_get_fl_payload_from_ctx(struct nfp_app *app, u32 ctx_id);
+struct nfp_fl_payload *
 nfp_flower_remove_fl_table(struct nfp_app *app, unsigned long tc_flower_cookie);
 
 void nfp_flower_rx_flow_stats(struct nfp_app *app, struct sk_buff *skb);
@@ -302,6 +392,11 @@ int nfp_flower_lag_populate_pre_action(struct nfp_app *app,
 				       struct nfp_fl_pre_lag *pre_act);
 int nfp_flower_lag_get_output_id(struct nfp_app *app,
 				 struct net_device *master);
+void nfp_flower_qos_init(struct nfp_app *app);
+void nfp_flower_qos_cleanup(struct nfp_app *app);
+int nfp_flower_setup_qos_offload(struct nfp_app *app, struct net_device *netdev,
+				 struct tc_cls_matchall_offload *flow);
+void nfp_flower_stats_rlim_reply(struct nfp_app *app, struct sk_buff *skb);
 int nfp_flower_reg_indir_block_handler(struct nfp_app *app,
 				       struct net_device *netdev,
 				       unsigned long event);
@@ -314,4 +409,6 @@ void
 __nfp_flower_non_repr_priv_put(struct nfp_flower_non_repr_priv *non_repr_priv);
 void
 nfp_flower_non_repr_priv_put(struct nfp_app *app, struct net_device *netdev);
+u32 nfp_flower_get_port_id_from_netdev(struct nfp_app *app,
+				       struct net_device *netdev);
 #endif
diff --git a/drivers/net/ethernet/netronome/nfp/flower/match.c b/drivers/net/ethernet/netronome/nfp/flower/match.c
index 9b8b843d0340..bfa4bf34911d 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/match.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/match.c
@@ -326,13 +326,12 @@ int nfp_flower_compile_flow_match(struct nfp_app *app,
 				  struct nfp_fl_payload *nfp_flow,
 				  enum nfp_flower_tun_type tun_type)
 {
-	u32 cmsg_port = 0;
+	u32 port_id;
 	int err;
 	u8 *ext;
 	u8 *msk;
 
-	if (nfp_netdev_is_nfp_repr(netdev))
-		cmsg_port = nfp_repr_get_port_id(netdev);
+	port_id = nfp_flower_get_port_id_from_netdev(app, netdev);
 
 	memset(nfp_flow->unmasked_data, 0, key_ls->key_size);
 	memset(nfp_flow->mask_data, 0, key_ls->key_size);
@@ -358,13 +357,13 @@ int nfp_flower_compile_flow_match(struct nfp_app *app,
 
 	/* Populate Exact Port data. */
 	err = nfp_flower_compile_port((struct nfp_flower_in_port *)ext,
-				      cmsg_port, false, tun_type);
+				      port_id, false, tun_type);
 	if (err)
 		return err;
 
 	/* Populate Mask Port Data. */
 	err = nfp_flower_compile_port((struct nfp_flower_in_port *)msk,
-				      cmsg_port, true, tun_type);
+				      port_id, true, tun_type);
 	if (err)
 		return err;
 
diff --git a/drivers/net/ethernet/netronome/nfp/flower/metadata.c b/drivers/net/ethernet/netronome/nfp/flower/metadata.c
index 492837b852b6..3d326efdc814 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/metadata.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/metadata.c
@@ -24,6 +24,18 @@ struct nfp_fl_flow_table_cmp_arg {
 	unsigned long cookie;
 };
 
+struct nfp_fl_stats_ctx_to_flow {
+	struct rhash_head ht_node;
+	u32 stats_cxt;
+	struct nfp_fl_payload *flow;
+};
+
+static const struct rhashtable_params stats_ctx_table_params = {
+	.key_offset	= offsetof(struct nfp_fl_stats_ctx_to_flow, stats_cxt),
+	.head_offset	= offsetof(struct nfp_fl_stats_ctx_to_flow, ht_node),
+	.key_len	= sizeof(u32),
+};
+
 static int nfp_release_stats_entry(struct nfp_app *app, u32 stats_context_id)
 {
 	struct nfp_flower_priv *priv = app->priv;
@@ -264,9 +276,6 @@ nfp_check_mask_remove(struct nfp_app *app, char *mask_data, u32 mask_len,
 	if (!mask_entry)
 		return false;
 
-	if (meta_flags)
-		*meta_flags &= ~NFP_FL_META_FLAG_MANAGE_MASK;
-
 	*mask_id = mask_entry->mask_id;
 	mask_entry->ref_cnt--;
 	if (!mask_entry->ref_cnt) {
@@ -285,25 +294,42 @@ int nfp_compile_flow_metadata(struct nfp_app *app,
 			      struct nfp_fl_payload *nfp_flow,
 			      struct net_device *netdev)
 {
+	struct nfp_fl_stats_ctx_to_flow *ctx_entry;
 	struct nfp_flower_priv *priv = app->priv;
 	struct nfp_fl_payload *check_entry;
 	u8 new_mask_id;
 	u32 stats_cxt;
+	int err;
 
-	if (nfp_get_stats_entry(app, &stats_cxt))
-		return -ENOENT;
+	err = nfp_get_stats_entry(app, &stats_cxt);
+	if (err)
+		return err;
 
 	nfp_flow->meta.host_ctx_id = cpu_to_be32(stats_cxt);
 	nfp_flow->meta.host_cookie = cpu_to_be64(flow->cookie);
 	nfp_flow->ingress_dev = netdev;
 
+	ctx_entry = kzalloc(sizeof(*ctx_entry), GFP_KERNEL);
+	if (!ctx_entry) {
+		err = -ENOMEM;
+		goto err_release_stats;
+	}
+
+	ctx_entry->stats_cxt = stats_cxt;
+	ctx_entry->flow = nfp_flow;
+
+	if (rhashtable_insert_fast(&priv->stats_ctx_table, &ctx_entry->ht_node,
+				   stats_ctx_table_params)) {
+		err = -ENOMEM;
+		goto err_free_ctx_entry;
+	}
+
 	new_mask_id = 0;
 	if (!nfp_check_mask_add(app, nfp_flow->mask_data,
 				nfp_flow->meta.mask_len,
 				&nfp_flow->meta.flags, &new_mask_id)) {
-		if (nfp_release_stats_entry(app, stats_cxt))
-			return -EINVAL;
-		return -ENOENT;
+		err = -ENOENT;
+		goto err_remove_rhash;
 	}
 
 	nfp_flow->meta.flow_version = cpu_to_be64(priv->flower_version);
@@ -317,43 +343,82 @@ int nfp_compile_flow_metadata(struct nfp_app *app,
 
 	check_entry = nfp_flower_search_fl_table(app, flow->cookie, netdev);
 	if (check_entry) {
-		if (nfp_release_stats_entry(app, stats_cxt))
-			return -EINVAL;
-
-		if (!nfp_check_mask_remove(app, nfp_flow->mask_data,
-					   nfp_flow->meta.mask_len,
-					   NULL, &new_mask_id))
-			return -EINVAL;
-
-		return -EEXIST;
+		err = -EEXIST;
+		goto err_remove_mask;
 	}
 
 	return 0;
+
+err_remove_mask:
+	nfp_check_mask_remove(app, nfp_flow->mask_data, nfp_flow->meta.mask_len,
+			      NULL, &new_mask_id);
+err_remove_rhash:
+	WARN_ON_ONCE(rhashtable_remove_fast(&priv->stats_ctx_table,
+					    &ctx_entry->ht_node,
+					    stats_ctx_table_params));
+err_free_ctx_entry:
+	kfree(ctx_entry);
+err_release_stats:
+	nfp_release_stats_entry(app, stats_cxt);
+
+	return err;
+}
+
+void __nfp_modify_flow_metadata(struct nfp_flower_priv *priv,
+				struct nfp_fl_payload *nfp_flow)
+{
+	nfp_flow->meta.flags &= ~NFP_FL_META_FLAG_MANAGE_MASK;
+	nfp_flow->meta.flow_version = cpu_to_be64(priv->flower_version);
+	priv->flower_version++;
 }
 
 int nfp_modify_flow_metadata(struct nfp_app *app,
 			     struct nfp_fl_payload *nfp_flow)
 {
+	struct nfp_fl_stats_ctx_to_flow *ctx_entry;
 	struct nfp_flower_priv *priv = app->priv;
 	u8 new_mask_id = 0;
 	u32 temp_ctx_id;
 
+	__nfp_modify_flow_metadata(priv, nfp_flow);
+
 	nfp_check_mask_remove(app, nfp_flow->mask_data,
 			      nfp_flow->meta.mask_len, &nfp_flow->meta.flags,
 			      &new_mask_id);
 
-	nfp_flow->meta.flow_version = cpu_to_be64(priv->flower_version);
-	priv->flower_version++;
-
 	/* Update flow payload with mask ids. */
 	nfp_flow->unmasked_data[NFP_FL_MASK_ID_LOCATION] = new_mask_id;
 
-	/* Release the stats ctx id. */
+	/* Release the stats ctx id and ctx to flow table entry. */
 	temp_ctx_id = be32_to_cpu(nfp_flow->meta.host_ctx_id);
 
+	ctx_entry = rhashtable_lookup_fast(&priv->stats_ctx_table, &temp_ctx_id,
+					   stats_ctx_table_params);
+	if (!ctx_entry)
+		return -ENOENT;
+
+	WARN_ON_ONCE(rhashtable_remove_fast(&priv->stats_ctx_table,
+					    &ctx_entry->ht_node,
+					    stats_ctx_table_params));
+	kfree(ctx_entry);
+
 	return nfp_release_stats_entry(app, temp_ctx_id);
 }
 
+struct nfp_fl_payload *
+nfp_flower_get_fl_payload_from_ctx(struct nfp_app *app, u32 ctx_id)
+{
+	struct nfp_fl_stats_ctx_to_flow *ctx_entry;
+	struct nfp_flower_priv *priv = app->priv;
+
+	ctx_entry = rhashtable_lookup_fast(&priv->stats_ctx_table, &ctx_id,
+					   stats_ctx_table_params);
+	if (!ctx_entry)
+		return NULL;
+
+	return ctx_entry->flow;
+}
+
 static int nfp_fl_obj_cmpfn(struct rhashtable_compare_arg *arg,
 			    const void *obj)
 {
@@ -403,6 +468,10 @@ int nfp_flower_metadata_init(struct nfp_app *app, u64 host_ctx_count,
 	if (err)
 		return err;
 
+	err = rhashtable_init(&priv->stats_ctx_table, &stats_ctx_table_params);
+	if (err)
+		goto err_free_flow_table;
+
 	get_random_bytes(&priv->mask_id_seed, sizeof(priv->mask_id_seed));
 
 	/* Init ring buffer and unallocated mask_ids. */
@@ -410,7 +479,7 @@ int nfp_flower_metadata_init(struct nfp_app *app, u64 host_ctx_count,
 		kmalloc_array(NFP_FLOWER_MASK_ENTRY_RS,
 			      NFP_FLOWER_MASK_ELEMENT_RS, GFP_KERNEL);
 	if (!priv->mask_ids.mask_id_free_list.buf)
-		goto err_free_flow_table;
+		goto err_free_stats_ctx_table;
 
 	priv->mask_ids.init_unallocated = NFP_FLOWER_MASK_ENTRY_RS - 1;
 
@@ -447,6 +516,8 @@ err_free_last_used:
 	kfree(priv->mask_ids.last_used);
 err_free_mask_id:
 	kfree(priv->mask_ids.mask_id_free_list.buf);
+err_free_stats_ctx_table:
+	rhashtable_destroy(&priv->stats_ctx_table);
 err_free_flow_table:
 	rhashtable_destroy(&priv->flow_table);
 	return -ENOMEM;
@@ -461,6 +532,8 @@ void nfp_flower_metadata_cleanup(struct nfp_app *app)
 
 	rhashtable_free_and_destroy(&priv->flow_table,
 				    nfp_check_rhashtable_empty, NULL);
+	rhashtable_free_and_destroy(&priv->stats_ctx_table,
+				    nfp_check_rhashtable_empty, NULL);
 	kvfree(priv->stats);
 	kfree(priv->mask_ids.mask_id_free_list.buf);
 	kfree(priv->mask_ids.last_used);
diff --git a/drivers/net/ethernet/netronome/nfp/flower/offload.c b/drivers/net/ethernet/netronome/nfp/flower/offload.c
index 450d7296fd57..1fbfeb43c538 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/offload.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/offload.c
@@ -55,6 +55,28 @@
 	 BIT(FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) | \
 	 BIT(FLOW_DISSECTOR_KEY_ENC_PORTS))
 
+#define NFP_FLOWER_MERGE_FIELDS \
+	(NFP_FLOWER_LAYER_PORT | \
+	 NFP_FLOWER_LAYER_MAC | \
+	 NFP_FLOWER_LAYER_TP | \
+	 NFP_FLOWER_LAYER_IPV4 | \
+	 NFP_FLOWER_LAYER_IPV6)
+
+struct nfp_flower_merge_check {
+	union {
+		struct {
+			__be16 tci;
+			struct nfp_flower_mac_mpls l2;
+			struct nfp_flower_tp_ports l4;
+			union {
+				struct nfp_flower_ipv4 ipv4;
+				struct nfp_flower_ipv6 ipv6;
+			};
+		};
+		unsigned long vals[8];
+	};
+};
+
 static int
 nfp_flower_xmit_flow(struct nfp_app *app, struct nfp_fl_payload *nfp_flow,
 		     u8 mtype)
@@ -195,7 +217,7 @@ nfp_flower_calculate_key_layers(struct nfp_app *app,
 			flow_rule_match_enc_opts(rule, &enc_op);
 
 		switch (enc_ports.key->dst) {
-		case htons(NFP_FL_VXLAN_PORT):
+		case htons(IANA_VXLAN_UDP_PORT):
 			*tun_type = NFP_FL_TUNNEL_VXLAN;
 			key_layer |= NFP_FLOWER_LAYER_VXLAN;
 			key_size += sizeof(struct nfp_flower_ipv4_udp_tun);
@@ -203,7 +225,7 @@ nfp_flower_calculate_key_layers(struct nfp_app *app,
 			if (enc_op.key)
 				return -EOPNOTSUPP;
 			break;
-		case htons(NFP_FL_GENEVE_PORT):
+		case htons(GENEVE_UDP_PORT):
 			if (!(priv->flower_ext_feats & NFP_FL_FEATS_GENEVE))
 				return -EOPNOTSUPP;
 			*tun_type = NFP_FL_TUNNEL_GENEVE;
@@ -326,7 +348,7 @@ nfp_flower_calculate_key_layers(struct nfp_app *app,
 				break;
 
 			case cpu_to_be16(ETH_P_IPV6):
-				key_layer |= NFP_FLOWER_LAYER_IPV6;
+					key_layer |= NFP_FLOWER_LAYER_IPV6;
 				key_size += sizeof(struct nfp_flower_ipv6);
 				break;
 
@@ -376,6 +398,8 @@ nfp_flower_allocate_new(struct nfp_fl_key_ls *key_layer)
 
 	flow_pay->nfp_tun_ipv4_addr = 0;
 	flow_pay->meta.flags = 0;
+	INIT_LIST_HEAD(&flow_pay->linked_flows);
+	flow_pay->in_hw = false;
 
 	return flow_pay;
 
@@ -388,6 +412,447 @@ err_free_flow:
 	return NULL;
 }
 
+static int
+nfp_flower_update_merge_with_actions(struct nfp_fl_payload *flow,
+				     struct nfp_flower_merge_check *merge,
+				     u8 *last_act_id, int *act_out)
+{
+	struct nfp_fl_set_ipv6_tc_hl_fl *ipv6_tc_hl_fl;
+	struct nfp_fl_set_ip4_ttl_tos *ipv4_ttl_tos;
+	struct nfp_fl_set_ip4_addrs *ipv4_add;
+	struct nfp_fl_set_ipv6_addr *ipv6_add;
+	struct nfp_fl_push_vlan *push_vlan;
+	struct nfp_fl_set_tport *tport;
+	struct nfp_fl_set_eth *eth;
+	struct nfp_fl_act_head *a;
+	unsigned int act_off = 0;
+	u8 act_id = 0;
+	u8 *ports;
+	int i;
+
+	while (act_off < flow->meta.act_len) {
+		a = (struct nfp_fl_act_head *)&flow->action_data[act_off];
+		act_id = a->jump_id;
+
+		switch (act_id) {
+		case NFP_FL_ACTION_OPCODE_OUTPUT:
+			if (act_out)
+				(*act_out)++;
+			break;
+		case NFP_FL_ACTION_OPCODE_PUSH_VLAN:
+			push_vlan = (struct nfp_fl_push_vlan *)a;
+			if (push_vlan->vlan_tci)
+				merge->tci = cpu_to_be16(0xffff);
+			break;
+		case NFP_FL_ACTION_OPCODE_POP_VLAN:
+			merge->tci = cpu_to_be16(0);
+			break;
+		case NFP_FL_ACTION_OPCODE_SET_IPV4_TUNNEL:
+			/* New tunnel header means l2 to l4 can be matched. */
+			eth_broadcast_addr(&merge->l2.mac_dst[0]);
+			eth_broadcast_addr(&merge->l2.mac_src[0]);
+			memset(&merge->l4, 0xff,
+			       sizeof(struct nfp_flower_tp_ports));
+			memset(&merge->ipv4, 0xff,
+			       sizeof(struct nfp_flower_ipv4));
+			break;
+		case NFP_FL_ACTION_OPCODE_SET_ETHERNET:
+			eth = (struct nfp_fl_set_eth *)a;
+			for (i = 0; i < ETH_ALEN; i++)
+				merge->l2.mac_dst[i] |= eth->eth_addr_mask[i];
+			for (i = 0; i < ETH_ALEN; i++)
+				merge->l2.mac_src[i] |=
+					eth->eth_addr_mask[ETH_ALEN + i];
+			break;
+		case NFP_FL_ACTION_OPCODE_SET_IPV4_ADDRS:
+			ipv4_add = (struct nfp_fl_set_ip4_addrs *)a;
+			merge->ipv4.ipv4_src |= ipv4_add->ipv4_src_mask;
+			merge->ipv4.ipv4_dst |= ipv4_add->ipv4_dst_mask;
+			break;
+		case NFP_FL_ACTION_OPCODE_SET_IPV4_TTL_TOS:
+			ipv4_ttl_tos = (struct nfp_fl_set_ip4_ttl_tos *)a;
+			merge->ipv4.ip_ext.ttl |= ipv4_ttl_tos->ipv4_ttl_mask;
+			merge->ipv4.ip_ext.tos |= ipv4_ttl_tos->ipv4_tos_mask;
+			break;
+		case NFP_FL_ACTION_OPCODE_SET_IPV6_SRC:
+			ipv6_add = (struct nfp_fl_set_ipv6_addr *)a;
+			for (i = 0; i < 4; i++)
+				merge->ipv6.ipv6_src.in6_u.u6_addr32[i] |=
+					ipv6_add->ipv6[i].mask;
+			break;
+		case NFP_FL_ACTION_OPCODE_SET_IPV6_DST:
+			ipv6_add = (struct nfp_fl_set_ipv6_addr *)a;
+			for (i = 0; i < 4; i++)
+				merge->ipv6.ipv6_dst.in6_u.u6_addr32[i] |=
+					ipv6_add->ipv6[i].mask;
+			break;
+		case NFP_FL_ACTION_OPCODE_SET_IPV6_TC_HL_FL:
+			ipv6_tc_hl_fl = (struct nfp_fl_set_ipv6_tc_hl_fl *)a;
+			merge->ipv6.ip_ext.ttl |=
+				ipv6_tc_hl_fl->ipv6_hop_limit_mask;
+			merge->ipv6.ip_ext.tos |= ipv6_tc_hl_fl->ipv6_tc_mask;
+			merge->ipv6.ipv6_flow_label_exthdr |=
+				ipv6_tc_hl_fl->ipv6_label_mask;
+			break;
+		case NFP_FL_ACTION_OPCODE_SET_UDP:
+		case NFP_FL_ACTION_OPCODE_SET_TCP:
+			tport = (struct nfp_fl_set_tport *)a;
+			ports = (u8 *)&merge->l4.port_src;
+			for (i = 0; i < 4; i++)
+				ports[i] |= tport->tp_port_mask[i];
+			break;
+		case NFP_FL_ACTION_OPCODE_PRE_TUNNEL:
+		case NFP_FL_ACTION_OPCODE_PRE_LAG:
+		case NFP_FL_ACTION_OPCODE_PUSH_GENEVE:
+			break;
+		default:
+			return -EOPNOTSUPP;
+		}
+
+		act_off += a->len_lw << NFP_FL_LW_SIZ;
+	}
+
+	if (last_act_id)
+		*last_act_id = act_id;
+
+	return 0;
+}
+
+static int
+nfp_flower_populate_merge_match(struct nfp_fl_payload *flow,
+				struct nfp_flower_merge_check *merge,
+				bool extra_fields)
+{
+	struct nfp_flower_meta_tci *meta_tci;
+	u8 *mask = flow->mask_data;
+	u8 key_layer, match_size;
+
+	memset(merge, 0, sizeof(struct nfp_flower_merge_check));
+
+	meta_tci = (struct nfp_flower_meta_tci *)mask;
+	key_layer = meta_tci->nfp_flow_key_layer;
+
+	if (key_layer & ~NFP_FLOWER_MERGE_FIELDS && !extra_fields)
+		return -EOPNOTSUPP;
+
+	merge->tci = meta_tci->tci;
+	mask += sizeof(struct nfp_flower_meta_tci);
+
+	if (key_layer & NFP_FLOWER_LAYER_EXT_META)
+		mask += sizeof(struct nfp_flower_ext_meta);
+
+	mask += sizeof(struct nfp_flower_in_port);
+
+	if (key_layer & NFP_FLOWER_LAYER_MAC) {
+		match_size = sizeof(struct nfp_flower_mac_mpls);
+		memcpy(&merge->l2, mask, match_size);
+		mask += match_size;
+	}
+
+	if (key_layer & NFP_FLOWER_LAYER_TP) {
+		match_size = sizeof(struct nfp_flower_tp_ports);
+		memcpy(&merge->l4, mask, match_size);
+		mask += match_size;
+	}
+
+	if (key_layer & NFP_FLOWER_LAYER_IPV4) {
+		match_size = sizeof(struct nfp_flower_ipv4);
+		memcpy(&merge->ipv4, mask, match_size);
+	}
+
+	if (key_layer & NFP_FLOWER_LAYER_IPV6) {
+		match_size = sizeof(struct nfp_flower_ipv6);
+		memcpy(&merge->ipv6, mask, match_size);
+	}
+
+	return 0;
+}
+
+static int
+nfp_flower_can_merge(struct nfp_fl_payload *sub_flow1,
+		     struct nfp_fl_payload *sub_flow2)
+{
+	/* Two flows can be merged if sub_flow2 only matches on bits that are
+	 * either matched by sub_flow1 or set by a sub_flow1 action. This
+	 * ensures that every packet that hits sub_flow1 and recirculates is
+	 * guaranteed to hit sub_flow2.
+	 */
+	struct nfp_flower_merge_check sub_flow1_merge, sub_flow2_merge;
+	int err, act_out = 0;
+	u8 last_act_id = 0;
+
+	err = nfp_flower_populate_merge_match(sub_flow1, &sub_flow1_merge,
+					      true);
+	if (err)
+		return err;
+
+	err = nfp_flower_populate_merge_match(sub_flow2, &sub_flow2_merge,
+					      false);
+	if (err)
+		return err;
+
+	err = nfp_flower_update_merge_with_actions(sub_flow1, &sub_flow1_merge,
+						   &last_act_id, &act_out);
+	if (err)
+		return err;
+
+	/* Must only be 1 output action and it must be the last in sequence. */
+	if (act_out != 1 || last_act_id != NFP_FL_ACTION_OPCODE_OUTPUT)
+		return -EOPNOTSUPP;
+
+	/* Reject merge if sub_flow2 matches on something that is not matched
+	 * on or set in an action by sub_flow1.
+	 */
+	err = bitmap_andnot(sub_flow2_merge.vals, sub_flow2_merge.vals,
+			    sub_flow1_merge.vals,
+			    sizeof(struct nfp_flower_merge_check) * 8);
+	if (err)
+		return -EINVAL;
+
+	return 0;
+}
+
+static unsigned int
+nfp_flower_copy_pre_actions(char *act_dst, char *act_src, int len,
+			    bool *tunnel_act)
+{
+	unsigned int act_off = 0, act_len;
+	struct nfp_fl_act_head *a;
+	u8 act_id = 0;
+
+	while (act_off < len) {
+		a = (struct nfp_fl_act_head *)&act_src[act_off];
+		act_len = a->len_lw << NFP_FL_LW_SIZ;
+		act_id = a->jump_id;
+
+		switch (act_id) {
+		case NFP_FL_ACTION_OPCODE_PRE_TUNNEL:
+			if (tunnel_act)
+				*tunnel_act = true;
+			/* fall through */
+		case NFP_FL_ACTION_OPCODE_PRE_LAG:
+			memcpy(act_dst + act_off, act_src + act_off, act_len);
+			break;
+		default:
+			return act_off;
+		}
+
+		act_off += act_len;
+	}
+
+	return act_off;
+}
+
+static int nfp_fl_verify_post_tun_acts(char *acts, int len)
+{
+	struct nfp_fl_act_head *a;
+	unsigned int act_off = 0;
+
+	while (act_off < len) {
+		a = (struct nfp_fl_act_head *)&acts[act_off];
+		if (a->jump_id != NFP_FL_ACTION_OPCODE_OUTPUT)
+			return -EOPNOTSUPP;
+
+		act_off += a->len_lw << NFP_FL_LW_SIZ;
+	}
+
+	return 0;
+}
+
+static int
+nfp_flower_merge_action(struct nfp_fl_payload *sub_flow1,
+			struct nfp_fl_payload *sub_flow2,
+			struct nfp_fl_payload *merge_flow)
+{
+	unsigned int sub1_act_len, sub2_act_len, pre_off1, pre_off2;
+	bool tunnel_act = false;
+	char *merge_act;
+	int err;
+
+	/* The last action of sub_flow1 must be output - do not merge this. */
+	sub1_act_len = sub_flow1->meta.act_len - sizeof(struct nfp_fl_output);
+	sub2_act_len = sub_flow2->meta.act_len;
+
+	if (!sub2_act_len)
+		return -EINVAL;
+
+	if (sub1_act_len + sub2_act_len > NFP_FL_MAX_A_SIZ)
+		return -EINVAL;
+
+	/* A shortcut can only be applied if there is a single action. */
+	if (sub1_act_len)
+		merge_flow->meta.shortcut = cpu_to_be32(NFP_FL_SC_ACT_NULL);
+	else
+		merge_flow->meta.shortcut = sub_flow2->meta.shortcut;
+
+	merge_flow->meta.act_len = sub1_act_len + sub2_act_len;
+	merge_act = merge_flow->action_data;
+
+	/* Copy any pre-actions to the start of merge flow action list. */
+	pre_off1 = nfp_flower_copy_pre_actions(merge_act,
+					       sub_flow1->action_data,
+					       sub1_act_len, &tunnel_act);
+	merge_act += pre_off1;
+	sub1_act_len -= pre_off1;
+	pre_off2 = nfp_flower_copy_pre_actions(merge_act,
+					       sub_flow2->action_data,
+					       sub2_act_len, NULL);
+	merge_act += pre_off2;
+	sub2_act_len -= pre_off2;
+
+	/* FW does a tunnel push when egressing, therefore, if sub_flow 1 pushes
+	 * a tunnel, sub_flow 2 can only have output actions for a valid merge.
+	 */
+	if (tunnel_act) {
+		char *post_tun_acts = &sub_flow2->action_data[pre_off2];
+
+		err = nfp_fl_verify_post_tun_acts(post_tun_acts, sub2_act_len);
+		if (err)
+			return err;
+	}
+
+	/* Copy remaining actions from sub_flows 1 and 2. */
+	memcpy(merge_act, sub_flow1->action_data + pre_off1, sub1_act_len);
+	merge_act += sub1_act_len;
+	memcpy(merge_act, sub_flow2->action_data + pre_off2, sub2_act_len);
+
+	return 0;
+}
+
+/* Flow link code should only be accessed under RTNL. */
+static void nfp_flower_unlink_flow(struct nfp_fl_payload_link *link)
+{
+	list_del(&link->merge_flow.list);
+	list_del(&link->sub_flow.list);
+	kfree(link);
+}
+
+static void nfp_flower_unlink_flows(struct nfp_fl_payload *merge_flow,
+				    struct nfp_fl_payload *sub_flow)
+{
+	struct nfp_fl_payload_link *link;
+
+	list_for_each_entry(link, &merge_flow->linked_flows, merge_flow.list)
+		if (link->sub_flow.flow == sub_flow) {
+			nfp_flower_unlink_flow(link);
+			return;
+		}
+}
+
+static int nfp_flower_link_flows(struct nfp_fl_payload *merge_flow,
+				 struct nfp_fl_payload *sub_flow)
+{
+	struct nfp_fl_payload_link *link;
+
+	link = kmalloc(sizeof(*link), GFP_KERNEL);
+	if (!link)
+		return -ENOMEM;
+
+	link->merge_flow.flow = merge_flow;
+	list_add_tail(&link->merge_flow.list, &merge_flow->linked_flows);
+	link->sub_flow.flow = sub_flow;
+	list_add_tail(&link->sub_flow.list, &sub_flow->linked_flows);
+
+	return 0;
+}
+
+/**
+ * nfp_flower_merge_offloaded_flows() - Merge 2 existing flows to single flow.
+ * @app:	Pointer to the APP handle
+ * @sub_flow1:	Initial flow matched to produce merge hint
+ * @sub_flow2:	Post recirculation flow matched in merge hint
+ *
+ * Combines 2 flows (if valid) to a single flow, removing the initial from hw
+ * and offloading the new, merged flow.
+ *
+ * Return: negative value on error, 0 in success.
+ */
+int nfp_flower_merge_offloaded_flows(struct nfp_app *app,
+				     struct nfp_fl_payload *sub_flow1,
+				     struct nfp_fl_payload *sub_flow2)
+{
+	struct tc_cls_flower_offload merge_tc_off;
+	struct nfp_flower_priv *priv = app->priv;
+	struct nfp_fl_payload *merge_flow;
+	struct nfp_fl_key_ls merge_key_ls;
+	int err;
+
+	ASSERT_RTNL();
+
+	if (sub_flow1 == sub_flow2 ||
+	    nfp_flower_is_merge_flow(sub_flow1) ||
+	    nfp_flower_is_merge_flow(sub_flow2))
+		return -EINVAL;
+
+	err = nfp_flower_can_merge(sub_flow1, sub_flow2);
+	if (err)
+		return err;
+
+	merge_key_ls.key_size = sub_flow1->meta.key_len;
+
+	merge_flow = nfp_flower_allocate_new(&merge_key_ls);
+	if (!merge_flow)
+		return -ENOMEM;
+
+	merge_flow->tc_flower_cookie = (unsigned long)merge_flow;
+	merge_flow->ingress_dev = sub_flow1->ingress_dev;
+
+	memcpy(merge_flow->unmasked_data, sub_flow1->unmasked_data,
+	       sub_flow1->meta.key_len);
+	memcpy(merge_flow->mask_data, sub_flow1->mask_data,
+	       sub_flow1->meta.mask_len);
+
+	err = nfp_flower_merge_action(sub_flow1, sub_flow2, merge_flow);
+	if (err)
+		goto err_destroy_merge_flow;
+
+	err = nfp_flower_link_flows(merge_flow, sub_flow1);
+	if (err)
+		goto err_destroy_merge_flow;
+
+	err = nfp_flower_link_flows(merge_flow, sub_flow2);
+	if (err)
+		goto err_unlink_sub_flow1;
+
+	merge_tc_off.cookie = merge_flow->tc_flower_cookie;
+	err = nfp_compile_flow_metadata(app, &merge_tc_off, merge_flow,
+					merge_flow->ingress_dev);
+	if (err)
+		goto err_unlink_sub_flow2;
+
+	err = rhashtable_insert_fast(&priv->flow_table, &merge_flow->fl_node,
+				     nfp_flower_table_params);
+	if (err)
+		goto err_release_metadata;
+
+	err = nfp_flower_xmit_flow(app, merge_flow,
+				   NFP_FLOWER_CMSG_TYPE_FLOW_MOD);
+	if (err)
+		goto err_remove_rhash;
+
+	merge_flow->in_hw = true;
+	sub_flow1->in_hw = false;
+
+	return 0;
+
+err_remove_rhash:
+	WARN_ON_ONCE(rhashtable_remove_fast(&priv->flow_table,
+					    &merge_flow->fl_node,
+					    nfp_flower_table_params));
+err_release_metadata:
+	nfp_modify_flow_metadata(app, merge_flow);
+err_unlink_sub_flow2:
+	nfp_flower_unlink_flows(merge_flow, sub_flow2);
+err_unlink_sub_flow1:
+	nfp_flower_unlink_flows(merge_flow, sub_flow1);
+err_destroy_merge_flow:
+	kfree(merge_flow->action_data);
+	kfree(merge_flow->mask_data);
+	kfree(merge_flow->unmasked_data);
+	kfree(merge_flow);
+	return err;
+}
+
 /**
  * nfp_flower_add_offload() - Adds a new flow to hardware.
  * @app:	Pointer to the APP handle
@@ -454,6 +919,8 @@ nfp_flower_add_offload(struct nfp_app *app, struct net_device *netdev,
 	if (port)
 		port->tc_offload_cnt++;
 
+	flow_pay->in_hw = true;
+
 	/* Deallocate flow payload when flower rule has been destroyed. */
 	kfree(key_layer);
 
@@ -475,6 +942,75 @@ err_free_key_ls:
 	return err;
 }
 
+static void
+nfp_flower_remove_merge_flow(struct nfp_app *app,
+			     struct nfp_fl_payload *del_sub_flow,
+			     struct nfp_fl_payload *merge_flow)
+{
+	struct nfp_flower_priv *priv = app->priv;
+	struct nfp_fl_payload_link *link, *temp;
+	struct nfp_fl_payload *origin;
+	bool mod = false;
+	int err;
+
+	link = list_first_entry(&merge_flow->linked_flows,
+				struct nfp_fl_payload_link, merge_flow.list);
+	origin = link->sub_flow.flow;
+
+	/* Re-add rule the merge had overwritten if it has not been deleted. */
+	if (origin != del_sub_flow)
+		mod = true;
+
+	err = nfp_modify_flow_metadata(app, merge_flow);
+	if (err) {
+		nfp_flower_cmsg_warn(app, "Metadata fail for merge flow delete.\n");
+		goto err_free_links;
+	}
+
+	if (!mod) {
+		err = nfp_flower_xmit_flow(app, merge_flow,
+					   NFP_FLOWER_CMSG_TYPE_FLOW_DEL);
+		if (err) {
+			nfp_flower_cmsg_warn(app, "Failed to delete merged flow.\n");
+			goto err_free_links;
+		}
+	} else {
+		__nfp_modify_flow_metadata(priv, origin);
+		err = nfp_flower_xmit_flow(app, origin,
+					   NFP_FLOWER_CMSG_TYPE_FLOW_MOD);
+		if (err)
+			nfp_flower_cmsg_warn(app, "Failed to revert merge flow.\n");
+		origin->in_hw = true;
+	}
+
+err_free_links:
+	/* Clean any links connected with the merged flow. */
+	list_for_each_entry_safe(link, temp, &merge_flow->linked_flows,
+				 merge_flow.list)
+		nfp_flower_unlink_flow(link);
+
+	kfree(merge_flow->action_data);
+	kfree(merge_flow->mask_data);
+	kfree(merge_flow->unmasked_data);
+	WARN_ON_ONCE(rhashtable_remove_fast(&priv->flow_table,
+					    &merge_flow->fl_node,
+					    nfp_flower_table_params));
+	kfree_rcu(merge_flow, rcu);
+}
+
+static void
+nfp_flower_del_linked_merge_flows(struct nfp_app *app,
+				  struct nfp_fl_payload *sub_flow)
+{
+	struct nfp_fl_payload_link *link, *temp;
+
+	/* Remove any merge flow formed from the deleted sub_flow. */
+	list_for_each_entry_safe(link, temp, &sub_flow->linked_flows,
+				 sub_flow.list)
+		nfp_flower_remove_merge_flow(app, sub_flow,
+					     link->merge_flow.flow);
+}
+
 /**
  * nfp_flower_del_offload() - Removes a flow from hardware.
  * @app:	Pointer to the APP handle
@@ -482,7 +1018,7 @@ err_free_key_ls:
  * @flow:	TC flower classifier offload structure
  *
  * Removes a flow from the repeated hash structure and clears the
- * action payload.
+ * action payload. Any flows merged from this are also deleted.
  *
  * Return: negative value on error, 0 if removed successfully.
  */
@@ -504,17 +1040,22 @@ nfp_flower_del_offload(struct nfp_app *app, struct net_device *netdev,
 
 	err = nfp_modify_flow_metadata(app, nfp_flow);
 	if (err)
-		goto err_free_flow;
+		goto err_free_merge_flow;
 
 	if (nfp_flow->nfp_tun_ipv4_addr)
 		nfp_tunnel_del_ipv4_off(app, nfp_flow->nfp_tun_ipv4_addr);
 
+	if (!nfp_flow->in_hw) {
+		err = 0;
+		goto err_free_merge_flow;
+	}
+
 	err = nfp_flower_xmit_flow(app, nfp_flow,
 				   NFP_FLOWER_CMSG_TYPE_FLOW_DEL);
-	if (err)
-		goto err_free_flow;
+	/* Fall through on error. */
 
-err_free_flow:
+err_free_merge_flow:
+	nfp_flower_del_linked_merge_flows(app, nfp_flow);
 	if (port)
 		port->tc_offload_cnt--;
 	kfree(nfp_flow->action_data);
@@ -527,6 +1068,52 @@ err_free_flow:
 	return err;
 }
 
+static void
+__nfp_flower_update_merge_stats(struct nfp_app *app,
+				struct nfp_fl_payload *merge_flow)
+{
+	struct nfp_flower_priv *priv = app->priv;
+	struct nfp_fl_payload_link *link;
+	struct nfp_fl_payload *sub_flow;
+	u64 pkts, bytes, used;
+	u32 ctx_id;
+
+	ctx_id = be32_to_cpu(merge_flow->meta.host_ctx_id);
+	pkts = priv->stats[ctx_id].pkts;
+	/* Do not cycle subflows if no stats to distribute. */
+	if (!pkts)
+		return;
+	bytes = priv->stats[ctx_id].bytes;
+	used = priv->stats[ctx_id].used;
+
+	/* Reset stats for the merge flow. */
+	priv->stats[ctx_id].pkts = 0;
+	priv->stats[ctx_id].bytes = 0;
+
+	/* The merge flow has received stats updates from firmware.
+	 * Distribute these stats to all subflows that form the merge.
+	 * The stats will collected from TC via the subflows.
+	 */
+	list_for_each_entry(link, &merge_flow->linked_flows, merge_flow.list) {
+		sub_flow = link->sub_flow.flow;
+		ctx_id = be32_to_cpu(sub_flow->meta.host_ctx_id);
+		priv->stats[ctx_id].pkts += pkts;
+		priv->stats[ctx_id].bytes += bytes;
+		max_t(u64, priv->stats[ctx_id].used, used);
+	}
+}
+
+static void
+nfp_flower_update_merge_stats(struct nfp_app *app,
+			      struct nfp_fl_payload *sub_flow)
+{
+	struct nfp_fl_payload_link *link;
+
+	/* Get merge flows that the subflow forms to distribute their stats. */
+	list_for_each_entry(link, &sub_flow->linked_flows, sub_flow.list)
+		__nfp_flower_update_merge_stats(app, link->merge_flow.flow);
+}
+
 /**
  * nfp_flower_get_stats() - Populates flow stats obtained from hardware.
  * @app:	Pointer to the APP handle
@@ -553,6 +1140,10 @@ nfp_flower_get_stats(struct nfp_app *app, struct net_device *netdev,
 	ctx_id = be32_to_cpu(nfp_flow->meta.host_ctx_id);
 
 	spin_lock_bh(&priv->stats_lock);
+	/* If request is for a sub_flow, update stats from merged flows. */
+	if (!list_empty(&nfp_flow->linked_flows))
+		nfp_flower_update_merge_stats(app, nfp_flow);
+
 	flow_stats_update(&flow->stats, priv->stats[ctx_id].bytes,
 			  priv->stats[ctx_id].pkts, priv->stats[ctx_id].used);
 
@@ -594,6 +1185,9 @@ static int nfp_flower_setup_tc_block_cb(enum tc_setup_type type,
 	case TC_SETUP_CLSFLOWER:
 		return nfp_flower_repr_offload(repr->app, repr->netdev,
 					       type_data);
+	case TC_SETUP_CLSMATCHALL:
+		return nfp_flower_setup_qos_offload(repr->app, repr->netdev,
+						    type_data);
 	default:
 		return -EOPNOTSUPP;
 	}
@@ -603,10 +1197,14 @@ static int nfp_flower_setup_tc_block(struct net_device *netdev,
 				     struct tc_block_offload *f)
 {
 	struct nfp_repr *repr = netdev_priv(netdev);
+	struct nfp_flower_repr_priv *repr_priv;
 
 	if (f->binder_type != TCF_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
 		return -EOPNOTSUPP;
 
+	repr_priv = repr->app_priv;
+	repr_priv->block_shared = tcf_block_shared(f->block);
+
 	switch (f->command) {
 	case TC_BLOCK_BIND:
 		return tcf_block_cb_register(f->block,
@@ -682,7 +1280,9 @@ nfp_flower_setup_indr_tc_block(struct net_device *netdev, struct nfp_app *app,
 	struct nfp_flower_priv *priv = app->priv;
 	int err;
 
-	if (f->binder_type != TCF_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
+	if (f->binder_type != TCF_BLOCK_BINDER_TYPE_CLSACT_INGRESS &&
+	    !(f->binder_type == TCF_BLOCK_BINDER_TYPE_CLSACT_EGRESS &&
+	      nfp_flower_internal_port_can_offload(app, netdev)))
 		return -EOPNOTSUPP;
 
 	switch (f->command) {
diff --git a/drivers/net/ethernet/netronome/nfp/flower/qos_conf.c b/drivers/net/ethernet/netronome/nfp/flower/qos_conf.c
new file mode 100644
index 000000000000..86e968cd5ffd
--- /dev/null
+++ b/drivers/net/ethernet/netronome/nfp/flower/qos_conf.c
@@ -0,0 +1,366 @@
+// SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+/* Copyright (C) 2019 Netronome Systems, Inc. */
+
+#include <linux/math64.h>
+#include <net/pkt_cls.h>
+#include <net/pkt_sched.h>
+
+#include "cmsg.h"
+#include "main.h"
+#include "../nfp_port.h"
+
+#define NFP_FL_QOS_UPDATE		msecs_to_jiffies(1000)
+
+struct nfp_police_cfg_head {
+	__be32 flags_opts;
+	__be32 port;
+};
+
+/* Police cmsg for configuring a trTCM traffic conditioner (8W/32B)
+ * See RFC 2698 for more details.
+ * ----------------------------------------------------------------
+ *    3                   2                   1
+ *  1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * |                          Flag options                         |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * |                          Port Ingress                         |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * |                        Token Bucket Peak                      |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * |                     Token Bucket Committed                    |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * |                         Peak Burst Size                       |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * |                      Committed Burst Size                     |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * |                      Peak Information Rate                    |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ * |                    Committed Information Rate                 |
+ * +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ */
+struct nfp_police_config {
+	struct nfp_police_cfg_head head;
+	__be32 bkt_tkn_p;
+	__be32 bkt_tkn_c;
+	__be32 pbs;
+	__be32 cbs;
+	__be32 pir;
+	__be32 cir;
+};
+
+struct nfp_police_stats_reply {
+	struct nfp_police_cfg_head head;
+	__be64 pass_bytes;
+	__be64 pass_pkts;
+	__be64 drop_bytes;
+	__be64 drop_pkts;
+};
+
+static int
+nfp_flower_install_rate_limiter(struct nfp_app *app, struct net_device *netdev,
+				struct tc_cls_matchall_offload *flow,
+				struct netlink_ext_ack *extack)
+{
+	struct flow_action_entry *action = &flow->rule->action.entries[0];
+	struct nfp_flower_priv *fl_priv = app->priv;
+	struct nfp_flower_repr_priv *repr_priv;
+	struct nfp_police_config *config;
+	struct nfp_repr *repr;
+	struct sk_buff *skb;
+	u32 netdev_port_id;
+	u64 burst, rate;
+
+	if (!nfp_netdev_is_nfp_repr(netdev)) {
+		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: qos rate limit offload not supported on higher level port");
+		return -EOPNOTSUPP;
+	}
+	repr = netdev_priv(netdev);
+	repr_priv = repr->app_priv;
+
+	if (repr_priv->block_shared) {
+		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: qos rate limit offload not supported on shared blocks");
+		return -EOPNOTSUPP;
+	}
+
+	if (repr->port->type != NFP_PORT_VF_PORT) {
+		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: qos rate limit offload not supported on non-VF ports");
+		return -EOPNOTSUPP;
+	}
+
+	if (!flow_offload_has_one_action(&flow->rule->action)) {
+		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: qos rate limit offload requires a single action");
+		return -EOPNOTSUPP;
+	}
+
+	if (flow->common.prio != (1 << 16)) {
+		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: qos rate limit offload requires highest priority");
+		return -EOPNOTSUPP;
+	}
+
+	if (action->id != FLOW_ACTION_POLICE) {
+		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: qos rate limit offload requires police action");
+		return -EOPNOTSUPP;
+	}
+
+	rate = action->police.rate_bytes_ps;
+	burst = div_u64(rate * PSCHED_NS2TICKS(action->police.burst),
+			PSCHED_TICKS_PER_SEC);
+	netdev_port_id = nfp_repr_get_port_id(netdev);
+
+	skb = nfp_flower_cmsg_alloc(repr->app, sizeof(struct nfp_police_config),
+				    NFP_FLOWER_CMSG_TYPE_QOS_MOD, GFP_KERNEL);
+	if (!skb)
+		return -ENOMEM;
+
+	config = nfp_flower_cmsg_get_data(skb);
+	memset(config, 0, sizeof(struct nfp_police_config));
+	config->head.port = cpu_to_be32(netdev_port_id);
+	config->bkt_tkn_p = cpu_to_be32(burst);
+	config->bkt_tkn_c = cpu_to_be32(burst);
+	config->pbs = cpu_to_be32(burst);
+	config->cbs = cpu_to_be32(burst);
+	config->pir = cpu_to_be32(rate);
+	config->cir = cpu_to_be32(rate);
+	nfp_ctrl_tx(repr->app->ctrl, skb);
+
+	repr_priv->qos_table.netdev_port_id = netdev_port_id;
+	fl_priv->qos_rate_limiters++;
+	if (fl_priv->qos_rate_limiters == 1)
+		schedule_delayed_work(&fl_priv->qos_stats_work,
+				      NFP_FL_QOS_UPDATE);
+
+	return 0;
+}
+
+static int
+nfp_flower_remove_rate_limiter(struct nfp_app *app, struct net_device *netdev,
+			       struct tc_cls_matchall_offload *flow,
+			       struct netlink_ext_ack *extack)
+{
+	struct nfp_flower_priv *fl_priv = app->priv;
+	struct nfp_flower_repr_priv *repr_priv;
+	struct nfp_police_config *config;
+	struct nfp_repr *repr;
+	struct sk_buff *skb;
+	u32 netdev_port_id;
+
+	if (!nfp_netdev_is_nfp_repr(netdev)) {
+		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: qos rate limit offload not supported on higher level port");
+		return -EOPNOTSUPP;
+	}
+	repr = netdev_priv(netdev);
+
+	netdev_port_id = nfp_repr_get_port_id(netdev);
+	repr_priv = repr->app_priv;
+
+	if (!repr_priv->qos_table.netdev_port_id) {
+		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: cannot remove qos entry that does not exist");
+		return -EOPNOTSUPP;
+	}
+
+	skb = nfp_flower_cmsg_alloc(repr->app, sizeof(struct nfp_police_config),
+				    NFP_FLOWER_CMSG_TYPE_QOS_DEL, GFP_KERNEL);
+	if (!skb)
+		return -ENOMEM;
+
+	/* Clear all qos associate data for this interface */
+	memset(&repr_priv->qos_table, 0, sizeof(struct nfp_fl_qos));
+	fl_priv->qos_rate_limiters--;
+	if (!fl_priv->qos_rate_limiters)
+		cancel_delayed_work_sync(&fl_priv->qos_stats_work);
+
+	config = nfp_flower_cmsg_get_data(skb);
+	memset(config, 0, sizeof(struct nfp_police_config));
+	config->head.port = cpu_to_be32(netdev_port_id);
+	nfp_ctrl_tx(repr->app->ctrl, skb);
+
+	return 0;
+}
+
+void nfp_flower_stats_rlim_reply(struct nfp_app *app, struct sk_buff *skb)
+{
+	struct nfp_flower_priv *fl_priv = app->priv;
+	struct nfp_flower_repr_priv *repr_priv;
+	struct nfp_police_stats_reply *msg;
+	struct nfp_stat_pair *curr_stats;
+	struct nfp_stat_pair *prev_stats;
+	struct net_device *netdev;
+	struct nfp_repr *repr;
+	u32 netdev_port_id;
+
+	msg = nfp_flower_cmsg_get_data(skb);
+	netdev_port_id = be32_to_cpu(msg->head.port);
+	rcu_read_lock();
+	netdev = nfp_app_dev_get(app, netdev_port_id, NULL);
+	if (!netdev)
+		goto exit_unlock_rcu;
+
+	repr = netdev_priv(netdev);
+	repr_priv = repr->app_priv;
+	curr_stats = &repr_priv->qos_table.curr_stats;
+	prev_stats = &repr_priv->qos_table.prev_stats;
+
+	spin_lock_bh(&fl_priv->qos_stats_lock);
+	curr_stats->pkts = be64_to_cpu(msg->pass_pkts) +
+			   be64_to_cpu(msg->drop_pkts);
+	curr_stats->bytes = be64_to_cpu(msg->pass_bytes) +
+			    be64_to_cpu(msg->drop_bytes);
+
+	if (!repr_priv->qos_table.last_update) {
+		prev_stats->pkts = curr_stats->pkts;
+		prev_stats->bytes = curr_stats->bytes;
+	}
+
+	repr_priv->qos_table.last_update = jiffies;
+	spin_unlock_bh(&fl_priv->qos_stats_lock);
+
+exit_unlock_rcu:
+	rcu_read_unlock();
+}
+
+static void
+nfp_flower_stats_rlim_request(struct nfp_flower_priv *fl_priv,
+			      u32 netdev_port_id)
+{
+	struct nfp_police_cfg_head *head;
+	struct sk_buff *skb;
+
+	skb = nfp_flower_cmsg_alloc(fl_priv->app,
+				    sizeof(struct nfp_police_cfg_head),
+				    NFP_FLOWER_CMSG_TYPE_QOS_STATS,
+				    GFP_ATOMIC);
+	if (!skb)
+		return;
+
+	head = nfp_flower_cmsg_get_data(skb);
+	memset(head, 0, sizeof(struct nfp_police_cfg_head));
+	head->port = cpu_to_be32(netdev_port_id);
+
+	nfp_ctrl_tx(fl_priv->app->ctrl, skb);
+}
+
+static void
+nfp_flower_stats_rlim_request_all(struct nfp_flower_priv *fl_priv)
+{
+	struct nfp_reprs *repr_set;
+	int i;
+
+	rcu_read_lock();
+	repr_set = rcu_dereference(fl_priv->app->reprs[NFP_REPR_TYPE_VF]);
+	if (!repr_set)
+		goto exit_unlock_rcu;
+
+	for (i = 0; i < repr_set->num_reprs; i++) {
+		struct net_device *netdev;
+
+		netdev = rcu_dereference(repr_set->reprs[i]);
+		if (netdev) {
+			struct nfp_repr *priv = netdev_priv(netdev);
+			struct nfp_flower_repr_priv *repr_priv;
+			u32 netdev_port_id;
+
+			repr_priv = priv->app_priv;
+			netdev_port_id = repr_priv->qos_table.netdev_port_id;
+			if (!netdev_port_id)
+				continue;
+
+			nfp_flower_stats_rlim_request(fl_priv, netdev_port_id);
+		}
+	}
+
+exit_unlock_rcu:
+	rcu_read_unlock();
+}
+
+static void update_stats_cache(struct work_struct *work)
+{
+	struct delayed_work *delayed_work;
+	struct nfp_flower_priv *fl_priv;
+
+	delayed_work = to_delayed_work(work);
+	fl_priv = container_of(delayed_work, struct nfp_flower_priv,
+			       qos_stats_work);
+
+	nfp_flower_stats_rlim_request_all(fl_priv);
+	schedule_delayed_work(&fl_priv->qos_stats_work, NFP_FL_QOS_UPDATE);
+}
+
+static int
+nfp_flower_stats_rate_limiter(struct nfp_app *app, struct net_device *netdev,
+			      struct tc_cls_matchall_offload *flow,
+			      struct netlink_ext_ack *extack)
+{
+	struct nfp_flower_priv *fl_priv = app->priv;
+	struct nfp_flower_repr_priv *repr_priv;
+	struct nfp_stat_pair *curr_stats;
+	struct nfp_stat_pair *prev_stats;
+	u64 diff_bytes, diff_pkts;
+	struct nfp_repr *repr;
+
+	if (!nfp_netdev_is_nfp_repr(netdev)) {
+		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: qos rate limit offload not supported on higher level port");
+		return -EOPNOTSUPP;
+	}
+	repr = netdev_priv(netdev);
+
+	repr_priv = repr->app_priv;
+	if (!repr_priv->qos_table.netdev_port_id) {
+		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: cannot find qos entry for stats update");
+		return -EOPNOTSUPP;
+	}
+
+	spin_lock_bh(&fl_priv->qos_stats_lock);
+	curr_stats = &repr_priv->qos_table.curr_stats;
+	prev_stats = &repr_priv->qos_table.prev_stats;
+	diff_pkts = curr_stats->pkts - prev_stats->pkts;
+	diff_bytes = curr_stats->bytes - prev_stats->bytes;
+	prev_stats->pkts = curr_stats->pkts;
+	prev_stats->bytes = curr_stats->bytes;
+	spin_unlock_bh(&fl_priv->qos_stats_lock);
+
+	flow_stats_update(&flow->stats, diff_bytes, diff_pkts,
+			  repr_priv->qos_table.last_update);
+	return 0;
+}
+
+void nfp_flower_qos_init(struct nfp_app *app)
+{
+	struct nfp_flower_priv *fl_priv = app->priv;
+
+	spin_lock_init(&fl_priv->qos_stats_lock);
+	INIT_DELAYED_WORK(&fl_priv->qos_stats_work, &update_stats_cache);
+}
+
+void nfp_flower_qos_cleanup(struct nfp_app *app)
+{
+	struct nfp_flower_priv *fl_priv = app->priv;
+
+	cancel_delayed_work_sync(&fl_priv->qos_stats_work);
+}
+
+int nfp_flower_setup_qos_offload(struct nfp_app *app, struct net_device *netdev,
+				 struct tc_cls_matchall_offload *flow)
+{
+	struct netlink_ext_ack *extack = flow->common.extack;
+	struct nfp_flower_priv *fl_priv = app->priv;
+
+	if (!(fl_priv->flower_ext_feats & NFP_FL_FEATS_VF_RLIM)) {
+		NL_SET_ERR_MSG_MOD(extack, "unsupported offload: loaded firmware does not support qos rate limit offload");
+		return -EOPNOTSUPP;
+	}
+
+	switch (flow->command) {
+	case TC_CLSMATCHALL_REPLACE:
+		return nfp_flower_install_rate_limiter(app, netdev, flow,
+						       extack);
+	case TC_CLSMATCHALL_DESTROY:
+		return nfp_flower_remove_rate_limiter(app, netdev, flow,
+						      extack);
+	case TC_CLSMATCHALL_STATS:
+		return nfp_flower_stats_rate_limiter(app, netdev, flow,
+						     extack);
+	default:
+		return -EOPNOTSUPP;
+	}
+}
diff --git a/drivers/net/ethernet/netronome/nfp/flower/tunnel_conf.c b/drivers/net/ethernet/netronome/nfp/flower/tunnel_conf.c
index 4d78be4ec4e9..faa06edf95ac 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/tunnel_conf.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/tunnel_conf.c
@@ -171,7 +171,7 @@ void nfp_tunnel_keep_alive(struct nfp_app *app, struct sk_buff *skb)
 	for (i = 0; i < count; i++) {
 		ipv4_addr = payload->tun_info[i].ipv4;
 		port = be32_to_cpu(payload->tun_info[i].egress_port);
-		netdev = nfp_app_repr_get(app, port);
+		netdev = nfp_app_dev_get(app, port, NULL);
 		if (!netdev)
 			continue;
 
@@ -270,9 +270,10 @@ nfp_tun_write_neigh(struct net_device *netdev, struct nfp_app *app,
 		    struct flowi4 *flow, struct neighbour *neigh, gfp_t flag)
 {
 	struct nfp_tun_neigh payload;
+	u32 port_id;
 
-	/* Only offload representor IPv4s for now. */
-	if (!nfp_netdev_is_nfp_repr(netdev))
+	port_id = nfp_flower_get_port_id_from_netdev(app, netdev);
+	if (!port_id)
 		return;
 
 	memset(&payload, 0, sizeof(struct nfp_tun_neigh));
@@ -290,7 +291,7 @@ nfp_tun_write_neigh(struct net_device *netdev, struct nfp_app *app,
 	payload.src_ipv4 = flow->saddr;
 	ether_addr_copy(payload.src_addr, netdev->dev_addr);
 	neigh_ha_snapshot(payload.dst_addr, neigh, netdev);
-	payload.port_id = cpu_to_be32(nfp_repr_get_port_id(netdev));
+	payload.port_id = cpu_to_be32(port_id);
 	/* Add destination of new route to NFP cache. */
 	nfp_tun_add_route_to_cache(app, payload.dst_ipv4);
 
@@ -366,7 +367,7 @@ void nfp_tunnel_request_route(struct nfp_app *app, struct sk_buff *skb)
 
 	payload = nfp_flower_cmsg_get_data(skb);
 
-	netdev = nfp_app_repr_get(app, be32_to_cpu(payload->ingress_port));
+	netdev = nfp_app_dev_get(app, be32_to_cpu(payload->ingress_port), NULL);
 	if (!netdev)
 		goto route_fail_warning;
 
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_app.h b/drivers/net/ethernet/netronome/nfp/nfp_app.h
index f8d422713705..76d13af46a7a 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_app.h
+++ b/drivers/net/ethernet/netronome/nfp/nfp_app.h
@@ -79,7 +79,7 @@ extern const struct nfp_app_type app_abm;
  * @eswitch_mode_set:    set SR-IOV eswitch mode (under pf->lock)
  * @sriov_enable: app-specific sriov initialisation
  * @sriov_disable: app-specific sriov clean-up
- * @repr_get:	get representor netdev
+ * @dev_get:	get representor or internal port representing netdev
  */
 struct nfp_app_type {
 	enum nfp_app_id id;
@@ -143,7 +143,8 @@ struct nfp_app_type {
 
 	enum devlink_eswitch_mode (*eswitch_mode_get)(struct nfp_app *app);
 	int (*eswitch_mode_set)(struct nfp_app *app, u16 mode);
-	struct net_device *(*repr_get)(struct nfp_app *app, u32 id);
+	struct net_device *(*dev_get)(struct nfp_app *app, u32 id,
+				      bool *redir_egress);
 };
 
 /**
@@ -397,12 +398,14 @@ static inline void nfp_app_sriov_disable(struct nfp_app *app)
 		app->type->sriov_disable(app);
 }
 
-static inline struct net_device *nfp_app_repr_get(struct nfp_app *app, u32 id)
+static inline
+struct net_device *nfp_app_dev_get(struct nfp_app *app, u32 id,
+				   bool *redir_egress)
 {
-	if (unlikely(!app || !app->type->repr_get))
+	if (unlikely(!app || !app->type->dev_get))
 		return NULL;
 
-	return app->type->repr_get(app, id);
+	return app->type->dev_get(app, id, redir_egress);
 }
 
 struct nfp_app *nfp_app_from_netdev(struct net_device *netdev);
@@ -433,6 +436,6 @@ int nfp_app_nic_vnic_alloc(struct nfp_app *app, struct nfp_net *nn,
 int nfp_app_nic_vnic_init_phy_port(struct nfp_pf *pf, struct nfp_app *app,
 				   struct nfp_net *nn, unsigned int id);
 
-struct devlink *nfp_devlink_get_devlink(struct net_device *netdev);
+struct devlink_port *nfp_devlink_get_devlink_port(struct net_device *netdev);
 
 #endif
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_devlink.c b/drivers/net/ethernet/netronome/nfp/nfp_devlink.c
index e9eca99cf493..c50fce42f473 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_devlink.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_devlink.c
@@ -144,7 +144,8 @@ nfp_devlink_sb_pool_get(struct devlink *devlink, unsigned int sb_index,
 static int
 nfp_devlink_sb_pool_set(struct devlink *devlink, unsigned int sb_index,
 			u16 pool_index,
-			u32 size, enum devlink_sb_threshold_type threshold_type)
+			u32 size, enum devlink_sb_threshold_type threshold_type,
+			struct netlink_ext_ack *extack)
 {
 	struct nfp_pf *pf = devlink_priv(devlink);
 
@@ -354,6 +355,8 @@ int nfp_devlink_port_register(struct nfp_app *app, struct nfp_port *port)
 {
 	struct nfp_eth_table_port eth_port;
 	struct devlink *devlink;
+	const u8 *serial;
+	int serial_len;
 	int ret;
 
 	rtnl_lock();
@@ -362,10 +365,10 @@ int nfp_devlink_port_register(struct nfp_app *app, struct nfp_port *port)
 	if (ret)
 		return ret;
 
-	devlink_port_type_eth_set(&port->dl_port, port->netdev);
+	serial_len = nfp_cpp_serial(port->app->cpp, &serial);
 	devlink_port_attrs_set(&port->dl_port, DEVLINK_PORT_FLAVOUR_PHYSICAL,
 			       eth_port.label_port, eth_port.is_split,
-			       eth_port.label_subport);
+			       eth_port.label_subport, serial, serial_len);
 
 	devlink = priv_to_devlink(app->pf);
 
@@ -377,13 +380,23 @@ void nfp_devlink_port_unregister(struct nfp_port *port)
 	devlink_port_unregister(&port->dl_port);
 }
 
-struct devlink *nfp_devlink_get_devlink(struct net_device *netdev)
+void nfp_devlink_port_type_eth_set(struct nfp_port *port)
+{
+	devlink_port_type_eth_set(&port->dl_port, port->netdev);
+}
+
+void nfp_devlink_port_type_clear(struct nfp_port *port)
 {
-	struct nfp_app *app;
+	devlink_port_type_clear(&port->dl_port);
+}
+
+struct devlink_port *nfp_devlink_get_devlink_port(struct net_device *netdev)
+{
+	struct nfp_port *port;
 
-	app = nfp_app_from_netdev(netdev);
-	if (!app)
+	port = nfp_port_from_netdev(netdev);
+	if (!port)
 		return NULL;
 
-	return priv_to_devlink(app->pf);
+	return &port->dl_port;
 }
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_main.c b/drivers/net/ethernet/netronome/nfp/nfp_main.c
index f4c8776e42b6..948d1a4b4643 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_main.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_main.c
@@ -294,6 +294,9 @@ static int nfp_pcie_sriov_disable(struct pci_dev *pdev)
 
 static int nfp_pcie_sriov_configure(struct pci_dev *pdev, int num_vfs)
 {
+	if (!pci_get_drvdata(pdev))
+		return -ENOENT;
+
 	if (num_vfs == 0)
 		return nfp_pcie_sriov_disable(pdev);
 	else
@@ -720,9 +723,13 @@ err_pci_disable:
 	return err;
 }
 
-static void nfp_pci_remove(struct pci_dev *pdev)
+static void __nfp_pci_shutdown(struct pci_dev *pdev, bool unload_fw)
 {
-	struct nfp_pf *pf = pci_get_drvdata(pdev);
+	struct nfp_pf *pf;
+
+	pf = pci_get_drvdata(pdev);
+	if (!pf)
+		return;
 
 	nfp_hwmon_unregister(pf);
 
@@ -733,7 +740,7 @@ static void nfp_pci_remove(struct pci_dev *pdev)
 	vfree(pf->dumpspec);
 	kfree(pf->rtbl);
 	nfp_mip_close(pf->mip);
-	if (pf->fw_loaded)
+	if (unload_fw && pf->fw_loaded)
 		nfp_fw_unload(pf);
 
 	destroy_workqueue(pf->wq);
@@ -749,11 +756,22 @@ static void nfp_pci_remove(struct pci_dev *pdev)
 	pci_disable_device(pdev);
 }
 
+static void nfp_pci_remove(struct pci_dev *pdev)
+{
+	__nfp_pci_shutdown(pdev, true);
+}
+
+static void nfp_pci_shutdown(struct pci_dev *pdev)
+{
+	__nfp_pci_shutdown(pdev, false);
+}
+
 static struct pci_driver nfp_pci_driver = {
 	.name			= nfp_driver_name,
 	.id_table		= nfp_pci_device_ids,
 	.probe			= nfp_pci_probe,
 	.remove			= nfp_pci_remove,
+	.shutdown		= nfp_pci_shutdown,
 	.sriov_configure	= nfp_pcie_sriov_configure,
 };
 
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net.h b/drivers/net/ethernet/netronome/nfp/nfp_net.h
index be37c2d6151c..df9aff2684ed 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_net.h
+++ b/drivers/net/ethernet/netronome/nfp/nfp_net.h
@@ -539,12 +539,17 @@ struct nfp_net_dp {
  * @shared_handler:     Handler for shared interrupts
  * @shared_name:        Name for shared interrupt
  * @me_freq_mhz:        ME clock_freq (MHz)
- * @reconfig_lock:	Protects HW reconfiguration request regs/machinery
+ * @reconfig_lock:	Protects @reconfig_posted, @reconfig_timer_active,
+ *			@reconfig_sync_present and HW reconfiguration request
+ *			regs/machinery from async requests (sync must take
+ *			@bar_lock)
  * @reconfig_posted:	Pending reconfig bits coming from async sources
  * @reconfig_timer_active:  Timer for reading reconfiguration results is pending
  * @reconfig_sync_present:  Some thread is performing synchronous reconfig
  * @reconfig_timer:	Timer for async reading of reconfig results
  * @reconfig_in_progress_update:	Update FW is processing now (debug only)
+ * @bar_lock:		vNIC config BAR access lock, protects: update,
+ *			mailbox area
  * @link_up:            Is the link up?
  * @link_status_lock:	Protects @link_* and ensures atomicity with BAR reading
  * @rx_coalesce_usecs:      RX interrupt moderation usecs delay parameter
@@ -615,6 +620,8 @@ struct nfp_net {
 	struct timer_list reconfig_timer;
 	u32 reconfig_in_progress_update;
 
+	struct mutex bar_lock;
+
 	u32 rx_coalesce_usecs;
 	u32 rx_coalesce_max_frames;
 	u32 tx_coalesce_usecs;
@@ -839,6 +846,16 @@ static inline void nfp_ctrl_unlock(struct nfp_net *nn)
 	spin_unlock_bh(&nn->r_vecs[0].lock);
 }
 
+static inline void nn_ctrl_bar_lock(struct nfp_net *nn)
+{
+	mutex_lock(&nn->bar_lock);
+}
+
+static inline void nn_ctrl_bar_unlock(struct nfp_net *nn)
+{
+	mutex_unlock(&nn->bar_lock);
+}
+
 /* Globals */
 extern const char nfp_driver_version[];
 
@@ -871,7 +888,9 @@ unsigned int nfp_net_rss_key_sz(struct nfp_net *nn);
 void nfp_net_rss_write_itbl(struct nfp_net *nn);
 void nfp_net_rss_write_key(struct nfp_net *nn);
 void nfp_net_coalesce_write_cfg(struct nfp_net *nn);
-int nfp_net_reconfig_mbox(struct nfp_net *nn, u32 mbox_cmd);
+int nfp_net_mbox_lock(struct nfp_net *nn, unsigned int data_size);
+int nfp_net_mbox_reconfig(struct nfp_net *nn, u32 mbox_cmd);
+int nfp_net_mbox_reconfig_and_unlock(struct nfp_net *nn, u32 mbox_cmd);
 
 unsigned int
 nfp_net_irqs_alloc(struct pci_dev *pdev, struct msix_entry *irq_entries,
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
index 6d1b8816552e..b82b684f52ce 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
@@ -23,6 +23,7 @@
 #include <linux/interrupt.h>
 #include <linux/ip.h>
 #include <linux/ipv6.h>
+#include <linux/lockdep.h>
 #include <linux/mm.h>
 #include <linux/overflow.h>
 #include <linux/page_ref.h>
@@ -137,20 +138,37 @@ static bool nfp_net_reconfig_check_done(struct nfp_net *nn, bool last_check)
 	return false;
 }
 
-static int nfp_net_reconfig_wait(struct nfp_net *nn, unsigned long deadline)
+static bool __nfp_net_reconfig_wait(struct nfp_net *nn, unsigned long deadline)
 {
 	bool timed_out = false;
+	int i;
+
+	/* Poll update field, waiting for NFP to ack the config.
+	 * Do an opportunistic wait-busy loop, afterward sleep.
+	 */
+	for (i = 0; i < 50; i++) {
+		if (nfp_net_reconfig_check_done(nn, false))
+			return false;
+		udelay(4);
+	}
 
-	/* Poll update field, waiting for NFP to ack the config */
 	while (!nfp_net_reconfig_check_done(nn, timed_out)) {
-		msleep(1);
+		usleep_range(250, 500);
 		timed_out = time_is_before_eq_jiffies(deadline);
 	}
 
+	return timed_out;
+}
+
+static int nfp_net_reconfig_wait(struct nfp_net *nn, unsigned long deadline)
+{
+	if (__nfp_net_reconfig_wait(nn, deadline))
+		return -EIO;
+
 	if (nn_readl(nn, NFP_NET_CFG_UPDATE) & NFP_NET_CFG_UPDATE_ERR)
 		return -EIO;
 
-	return timed_out ? -EIO : 0;
+	return 0;
 }
 
 static void nfp_net_reconfig_timer(struct timer_list *t)
@@ -243,7 +261,7 @@ static void nfp_net_reconfig_wait_posted(struct nfp_net *nn)
 }
 
 /**
- * nfp_net_reconfig() - Reconfigure the firmware
+ * __nfp_net_reconfig() - Reconfigure the firmware
  * @nn:      NFP Net device to reconfigure
  * @update:  The value for the update field in the BAR config
  *
@@ -253,10 +271,12 @@ static void nfp_net_reconfig_wait_posted(struct nfp_net *nn)
  *
  * Return: Negative errno on error, 0 on success
  */
-int nfp_net_reconfig(struct nfp_net *nn, u32 update)
+static int __nfp_net_reconfig(struct nfp_net *nn, u32 update)
 {
 	int ret;
 
+	lockdep_assert_held(&nn->bar_lock);
+
 	nfp_net_reconfig_sync_enter(nn);
 
 	nfp_net_reconfig_start(nn, update);
@@ -274,8 +294,31 @@ int nfp_net_reconfig(struct nfp_net *nn, u32 update)
 	return ret;
 }
 
+int nfp_net_reconfig(struct nfp_net *nn, u32 update)
+{
+	int ret;
+
+	nn_ctrl_bar_lock(nn);
+	ret = __nfp_net_reconfig(nn, update);
+	nn_ctrl_bar_unlock(nn);
+
+	return ret;
+}
+
+int nfp_net_mbox_lock(struct nfp_net *nn, unsigned int data_size)
+{
+	if (nn->tlv_caps.mbox_len < NFP_NET_CFG_MBOX_SIMPLE_VAL + data_size) {
+		nn_err(nn, "mailbox too small for %u of data (%u)\n",
+		       data_size, nn->tlv_caps.mbox_len);
+		return -EIO;
+	}
+
+	nn_ctrl_bar_lock(nn);
+	return 0;
+}
+
 /**
- * nfp_net_reconfig_mbox() - Reconfigure the firmware via the mailbox
+ * nfp_net_mbox_reconfig() - Reconfigure the firmware via the mailbox
  * @nn:        NFP Net device to reconfigure
  * @mbox_cmd:  The value for the mailbox command
  *
@@ -283,19 +326,15 @@ int nfp_net_reconfig(struct nfp_net *nn, u32 update)
  *
  * Return: Negative errno on error, 0 on success
  */
-int nfp_net_reconfig_mbox(struct nfp_net *nn, u32 mbox_cmd)
+int nfp_net_mbox_reconfig(struct nfp_net *nn, u32 mbox_cmd)
 {
 	u32 mbox = nn->tlv_caps.mbox_off;
 	int ret;
 
-	if (!nfp_net_has_mbox(&nn->tlv_caps)) {
-		nn_err(nn, "no mailbox present, command: %u\n", mbox_cmd);
-		return -EIO;
-	}
-
+	lockdep_assert_held(&nn->bar_lock);
 	nn_writeq(nn, mbox + NFP_NET_CFG_MBOX_SIMPLE_CMD, mbox_cmd);
 
-	ret = nfp_net_reconfig(nn, NFP_NET_CFG_UPDATE_MBOX);
+	ret = __nfp_net_reconfig(nn, NFP_NET_CFG_UPDATE_MBOX);
 	if (ret) {
 		nn_err(nn, "Mailbox update error\n");
 		return ret;
@@ -304,6 +343,15 @@ int nfp_net_reconfig_mbox(struct nfp_net *nn, u32 mbox_cmd)
 	return -nn_readl(nn, mbox + NFP_NET_CFG_MBOX_SIMPLE_RET);
 }
 
+int nfp_net_mbox_reconfig_and_unlock(struct nfp_net *nn, u32 mbox_cmd)
+{
+	int ret;
+
+	ret = nfp_net_mbox_reconfig(nn, mbox_cmd);
+	nn_ctrl_bar_unlock(nn);
+	return ret;
+}
+
 /* Interrupt configuration and handling
  */
 
@@ -909,7 +957,7 @@ static int nfp_net_tx(struct sk_buff *skb, struct net_device *netdev)
 		nfp_net_tx_ring_stop(nd_q, tx_ring);
 
 	tx_ring->wr_ptr_add += nr_frags + 1;
-	if (__netdev_tx_sent_queue(nd_q, txbuf->real_len, skb->xmit_more))
+	if (__netdev_tx_sent_queue(nd_q, txbuf->real_len, netdev_xmit_more()))
 		nfp_net_tx_xmit_more_flush(tx_ring);
 
 	return NETDEV_TX_OK;
@@ -1635,6 +1683,7 @@ static int nfp_net_rx(struct nfp_net_rx_ring *rx_ring, int budget)
 		struct nfp_net_rx_buf *rxbuf;
 		struct nfp_net_rx_desc *rxd;
 		struct nfp_meta_parsed meta;
+		bool redir_egress = false;
 		struct net_device *netdev;
 		dma_addr_t new_dma_addr;
 		u32 meta_len_xdp = 0;
@@ -1770,13 +1819,16 @@ static int nfp_net_rx(struct nfp_net_rx_ring *rx_ring, int budget)
 			struct nfp_net *nn;
 
 			nn = netdev_priv(dp->netdev);
-			netdev = nfp_app_repr_get(nn->app, meta.portid);
+			netdev = nfp_app_dev_get(nn->app, meta.portid,
+						 &redir_egress);
 			if (unlikely(!netdev)) {
 				nfp_net_rx_drop(dp, r_vec, rx_ring, rxbuf,
 						NULL);
 				continue;
 			}
-			nfp_repr_inc_rx_stats(netdev, pkt_len);
+
+			if (nfp_netdev_is_nfp_repr(netdev))
+				nfp_repr_inc_rx_stats(netdev, pkt_len);
 		}
 
 		skb = build_skb(rxbuf->frag, true_bufsz);
@@ -1811,7 +1863,13 @@ static int nfp_net_rx(struct nfp_net_rx_ring *rx_ring, int budget)
 		if (meta_len_xdp)
 			skb_metadata_set(skb, meta_len_xdp);
 
-		napi_gro_receive(&rx_ring->r_vec->napi, skb);
+		if (likely(!redir_egress)) {
+			napi_gro_receive(&rx_ring->r_vec->napi, skb);
+		} else {
+			skb->dev = netdev;
+			__skb_push(skb, ETH_HLEN);
+			dev_queue_xmit(skb);
+		}
 	}
 
 	if (xdp_prog) {
@@ -3111,7 +3169,9 @@ static int nfp_net_change_mtu(struct net_device *netdev, int new_mtu)
 static int
 nfp_net_vlan_rx_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
 {
+	const u32 cmd = NFP_NET_CFG_MBOX_CMD_CTAG_FILTER_ADD;
 	struct nfp_net *nn = netdev_priv(netdev);
+	int err;
 
 	/* Priority tagged packets with vlan id 0 are processed by the
 	 * NFP as untagged packets
@@ -3119,17 +3179,23 @@ nfp_net_vlan_rx_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
 	if (!vid)
 		return 0;
 
+	err = nfp_net_mbox_lock(nn, NFP_NET_CFG_VLAN_FILTER_SZ);
+	if (err)
+		return err;
+
 	nn_writew(nn, nn->tlv_caps.mbox_off + NFP_NET_CFG_VLAN_FILTER_VID, vid);
 	nn_writew(nn, nn->tlv_caps.mbox_off + NFP_NET_CFG_VLAN_FILTER_PROTO,
 		  ETH_P_8021Q);
 
-	return nfp_net_reconfig_mbox(nn, NFP_NET_CFG_MBOX_CMD_CTAG_FILTER_ADD);
+	return nfp_net_mbox_reconfig_and_unlock(nn, cmd);
 }
 
 static int
 nfp_net_vlan_rx_kill_vid(struct net_device *netdev, __be16 proto, u16 vid)
 {
+	const u32 cmd = NFP_NET_CFG_MBOX_CMD_CTAG_FILTER_KILL;
 	struct nfp_net *nn = netdev_priv(netdev);
+	int err;
 
 	/* Priority tagged packets with vlan id 0 are processed by the
 	 * NFP as untagged packets
@@ -3137,11 +3203,15 @@ nfp_net_vlan_rx_kill_vid(struct net_device *netdev, __be16 proto, u16 vid)
 	if (!vid)
 		return 0;
 
+	err = nfp_net_mbox_lock(nn, NFP_NET_CFG_VLAN_FILTER_SZ);
+	if (err)
+		return err;
+
 	nn_writew(nn, nn->tlv_caps.mbox_off + NFP_NET_CFG_VLAN_FILTER_VID, vid);
 	nn_writew(nn, nn->tlv_caps.mbox_off + NFP_NET_CFG_VLAN_FILTER_PROTO,
 		  ETH_P_8021Q);
 
-	return nfp_net_reconfig_mbox(nn, NFP_NET_CFG_MBOX_CMD_CTAG_FILTER_KILL);
+	return nfp_net_mbox_reconfig_and_unlock(nn, cmd);
 }
 
 static void nfp_net_stat64(struct net_device *netdev,
@@ -3324,8 +3394,11 @@ nfp_net_get_phys_port_name(struct net_device *netdev, char *name, size_t len)
 	struct nfp_net *nn = netdev_priv(netdev);
 	int n;
 
+	/* If port is defined, devlink_port is registered and devlink core
+	 * is taking care of name formatting.
+	 */
 	if (nn->port)
-		return nfp_port_get_phys_port_name(netdev, name, len);
+		return -EOPNOTSUPP;
 
 	if (nn->dp.is_vf || nn->vnic_no_name)
 		return -EOPNOTSUPP;
@@ -3517,6 +3590,7 @@ const struct net_device_ops nfp_net_netdev_ops = {
 	.ndo_set_vf_mac         = nfp_app_set_vf_mac,
 	.ndo_set_vf_vlan        = nfp_app_set_vf_vlan,
 	.ndo_set_vf_spoofchk    = nfp_app_set_vf_spoofchk,
+	.ndo_set_vf_trust	= nfp_app_set_vf_trust,
 	.ndo_get_vf_config	= nfp_app_get_vf_config,
 	.ndo_set_vf_link_state  = nfp_app_set_vf_link_state,
 	.ndo_setup_tc		= nfp_port_setup_tc,
@@ -3530,8 +3604,7 @@ const struct net_device_ops nfp_net_netdev_ops = {
 	.ndo_udp_tunnel_add	= nfp_net_add_vxlan_port,
 	.ndo_udp_tunnel_del	= nfp_net_del_vxlan_port,
 	.ndo_bpf		= nfp_net_xdp,
-	.ndo_get_port_parent_id	= nfp_port_get_port_parent_id,
-	.ndo_get_devlink	= nfp_devlink_get_devlink,
+	.ndo_get_devlink_port	= nfp_devlink_get_devlink_port,
 };
 
 /**
@@ -3548,7 +3621,7 @@ void nfp_net_info(struct nfp_net *nn)
 		nn->fw_ver.resv, nn->fw_ver.class,
 		nn->fw_ver.major, nn->fw_ver.minor,
 		nn->max_mtu);
-	nn_info(nn, "CAP: %#x %s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
+	nn_info(nn, "CAP: %#x %s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
 		nn->cap,
 		nn->cap & NFP_NET_CFG_CTRL_PROMISC  ? "PROMISC "  : "",
 		nn->cap & NFP_NET_CFG_CTRL_L2BC     ? "L2BCFILT " : "",
@@ -3564,7 +3637,6 @@ void nfp_net_info(struct nfp_net *nn)
 		nn->cap & NFP_NET_CFG_CTRL_RSS      ? "RSS1 "     : "",
 		nn->cap & NFP_NET_CFG_CTRL_RSS2     ? "RSS2 "     : "",
 		nn->cap & NFP_NET_CFG_CTRL_CTAG_FILTER ? "CTAG_FILTER " : "",
-		nn->cap & NFP_NET_CFG_CTRL_L2SWITCH ? "L2SWITCH " : "",
 		nn->cap & NFP_NET_CFG_CTRL_MSIXAUTO ? "AUTOMASK " : "",
 		nn->cap & NFP_NET_CFG_CTRL_IRQMOD   ? "IRQMOD "   : "",
 		nn->cap & NFP_NET_CFG_CTRL_VXLAN    ? "VXLAN "    : "",
@@ -3632,6 +3704,8 @@ nfp_net_alloc(struct pci_dev *pdev, void __iomem *ctrl_bar, bool needs_netdev,
 	nn->dp.txd_cnt = NFP_NET_TX_DESCS_DEFAULT;
 	nn->dp.rxd_cnt = NFP_NET_RX_DESCS_DEFAULT;
 
+	mutex_init(&nn->bar_lock);
+
 	spin_lock_init(&nn->reconfig_lock);
 	spin_lock_init(&nn->link_status_lock);
 
@@ -3659,6 +3733,9 @@ err_free_nn:
 void nfp_net_free(struct nfp_net *nn)
 {
 	WARN_ON(timer_pending(&nn->reconfig_timer) || nn->reconfig_posted);
+
+	mutex_destroy(&nn->bar_lock);
+
 	if (nn->dp.netdev)
 		free_netdev(nn->dp.netdev);
 	else
@@ -3920,9 +3997,6 @@ int nfp_net_init(struct nfp_net *nn)
 		nn->dp.ctrl |= NFP_NET_CFG_CTRL_IRQMOD;
 	}
 
-	if (nn->dp.netdev)
-		nfp_net_netdev_init(nn);
-
 	/* Stash the re-configuration queue away.  First odd queue in TX Bar */
 	nn->qcp_cfg = nn->tx_bar + NFP_QCP_QUEUE_ADDR_SZ;
 
@@ -3935,6 +4009,9 @@ int nfp_net_init(struct nfp_net *nn)
 	if (err)
 		return err;
 
+	if (nn->dp.netdev)
+		nfp_net_netdev_init(nn);
+
 	nfp_net_vecs_init(nn);
 
 	if (!nn->dp.netdev)
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_ctrl.h b/drivers/net/ethernet/netronome/nfp/nfp_net_ctrl.h
index 372adea10e14..25919e338071 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_net_ctrl.h
+++ b/drivers/net/ethernet/netronome/nfp/nfp_net_ctrl.h
@@ -104,8 +104,6 @@
 #define   NFP_NET_CFG_CTRL_RINGPRIO	  (0x1 << 19) /* Ring priorities */
 #define   NFP_NET_CFG_CTRL_MSIXAUTO	  (0x1 << 20) /* MSI-X auto-masking */
 #define   NFP_NET_CFG_CTRL_TXRWB	  (0x1 << 21) /* Write-back of TX ring*/
-#define   NFP_NET_CFG_CTRL_L2SWITCH	  (0x1 << 22) /* L2 Switch */
-#define   NFP_NET_CFG_CTRL_L2SWITCH_LOCAL (0x1 << 23) /* Switch to local */
 #define   NFP_NET_CFG_CTRL_VXLAN	  (0x1 << 24) /* VXLAN tunnel support */
 #define   NFP_NET_CFG_CTRL_NVGRE	  (0x1 << 25) /* NVGRE tunnel support */
 #define   NFP_NET_CFG_CTRL_BPF		  (0x1 << 27) /* BPF offload capable */
@@ -130,7 +128,6 @@
 #define   NFP_NET_CFG_UPDATE_TXRPRIO	  (0x1 <<  3) /* TX Ring prio change */
 #define   NFP_NET_CFG_UPDATE_RXRPRIO	  (0x1 <<  4) /* RX Ring prio change */
 #define   NFP_NET_CFG_UPDATE_MSIX	  (0x1 <<  5) /* MSI-X change */
-#define   NFP_NET_CFG_UPDATE_L2SWITCH	  (0x1 <<  6) /* Switch changes */
 #define   NFP_NET_CFG_UPDATE_RESET	  (0x1 <<  7) /* Update due to FLR */
 #define   NFP_NET_CFG_UPDATE_IRQMOD	  (0x1 <<  8) /* IRQ mod change */
 #define   NFP_NET_CFG_UPDATE_VXLAN	  (0x1 <<  9) /* VXLAN port change */
@@ -392,7 +389,6 @@
 #define NFP_NET_CFG_MBOX_SIMPLE_CMD	0x0
 #define NFP_NET_CFG_MBOX_SIMPLE_RET	0x4
 #define NFP_NET_CFG_MBOX_SIMPLE_VAL	0x8
-#define NFP_NET_CFG_MBOX_SIMPLE_LEN	12
 
 #define NFP_NET_CFG_MBOX_CMD_CTAG_FILTER_ADD 1
 #define NFP_NET_CFG_MBOX_CMD_CTAG_FILTER_KILL 2
@@ -498,10 +494,4 @@ struct nfp_net_tlv_caps {
 
 int nfp_net_tlv_caps_parse(struct device *dev, u8 __iomem *ctrl_mem,
 			   struct nfp_net_tlv_caps *caps);
-
-static inline bool nfp_net_has_mbox(struct nfp_net_tlv_caps *caps)
-{
-	return caps->mbox_len >= NFP_NET_CFG_MBOX_SIMPLE_LEN;
-}
-
 #endif /* _NFP_NET_CTRL_H_ */
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_ethtool.c b/drivers/net/ethernet/netronome/nfp/nfp_net_ethtool.c
index 690b62718dbb..851e31e0ba8e 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_net_ethtool.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_net_ethtool.c
@@ -18,6 +18,7 @@
 #include <linux/pci.h>
 #include <linux/ethtool.h>
 #include <linux/firmware.h>
+#include <linux/sfp.h>
 
 #include "nfpcore/nfp.h"
 #include "nfpcore/nfp_nsp.h"
@@ -152,6 +153,8 @@ static const struct nfp_et_stat nfp_mac_et_stats[] = {
 #define NN_RVEC_GATHER_STATS	9
 #define NN_RVEC_PER_Q_STATS	3
 
+#define SFP_SFF_REV_COMPLIANCE	1
+
 static void nfp_net_get_nspinfo(struct nfp_app *app, char *version)
 {
 	struct nfp_nsp *nsp;
@@ -1096,6 +1099,130 @@ nfp_app_get_dump_data(struct net_device *netdev, struct ethtool_dump *dump,
 					    buffer);
 }
 
+static int
+nfp_port_get_module_info(struct net_device *netdev,
+			 struct ethtool_modinfo *modinfo)
+{
+	struct nfp_eth_table_port *eth_port;
+	struct nfp_port *port;
+	unsigned int read_len;
+	struct nfp_nsp *nsp;
+	int err = 0;
+	u8 data;
+
+	port = nfp_port_from_netdev(netdev);
+	eth_port = nfp_port_get_eth_port(port);
+	if (!eth_port)
+		return -EOPNOTSUPP;
+
+	nsp = nfp_nsp_open(port->app->cpp);
+	if (IS_ERR(nsp)) {
+		err = PTR_ERR(nsp);
+		netdev_err(netdev, "Failed to access the NSP: %d\n", err);
+		return err;
+	}
+
+	if (!nfp_nsp_has_read_module_eeprom(nsp)) {
+		netdev_info(netdev, "reading module EEPROM not supported. Please update flash\n");
+		err = -EOPNOTSUPP;
+		goto exit_close_nsp;
+	}
+
+	switch (eth_port->interface) {
+	case NFP_INTERFACE_SFP:
+	case NFP_INTERFACE_SFP28:
+		err = nfp_nsp_read_module_eeprom(nsp, eth_port->eth_index,
+						 SFP_SFF8472_COMPLIANCE, &data,
+						 1, &read_len);
+		if (err < 0)
+			goto exit_close_nsp;
+
+		if (!data) {
+			modinfo->type = ETH_MODULE_SFF_8079;
+			modinfo->eeprom_len = ETH_MODULE_SFF_8079_LEN;
+		} else {
+			modinfo->type = ETH_MODULE_SFF_8472;
+			modinfo->eeprom_len = ETH_MODULE_SFF_8472_LEN;
+		}
+		break;
+	case NFP_INTERFACE_QSFP:
+		err = nfp_nsp_read_module_eeprom(nsp, eth_port->eth_index,
+						 SFP_SFF_REV_COMPLIANCE, &data,
+						 1, &read_len);
+		if (err < 0)
+			goto exit_close_nsp;
+
+		if (data < 0x3) {
+			modinfo->type = ETH_MODULE_SFF_8436;
+			modinfo->eeprom_len = ETH_MODULE_SFF_8436_LEN;
+		} else {
+			modinfo->type = ETH_MODULE_SFF_8636;
+			modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
+		}
+		break;
+	case NFP_INTERFACE_QSFP28:
+		modinfo->type = ETH_MODULE_SFF_8636;
+		modinfo->eeprom_len = ETH_MODULE_SFF_8636_LEN;
+		break;
+	default:
+		netdev_err(netdev, "Unsupported module 0x%x detected\n",
+			   eth_port->interface);
+		err = -EINVAL;
+	}
+
+exit_close_nsp:
+	nfp_nsp_close(nsp);
+	return err;
+}
+
+static int
+nfp_port_get_module_eeprom(struct net_device *netdev,
+			   struct ethtool_eeprom *eeprom, u8 *data)
+{
+	struct nfp_eth_table_port *eth_port;
+	struct nfp_port *port;
+	struct nfp_nsp *nsp;
+	int err;
+
+	port = nfp_port_from_netdev(netdev);
+	eth_port = __nfp_port_get_eth_port(port);
+	if (!eth_port)
+		return -EOPNOTSUPP;
+
+	nsp = nfp_nsp_open(port->app->cpp);
+	if (IS_ERR(nsp)) {
+		err = PTR_ERR(nsp);
+		netdev_err(netdev, "Failed to access the NSP: %d\n", err);
+		return err;
+	}
+
+	if (!nfp_nsp_has_read_module_eeprom(nsp)) {
+		netdev_info(netdev, "reading module EEPROM not supported. Please update flash\n");
+		err = -EOPNOTSUPP;
+		goto exit_close_nsp;
+	}
+
+	err = nfp_nsp_read_module_eeprom(nsp, eth_port->eth_index,
+					 eeprom->offset, data, eeprom->len,
+					 &eeprom->len);
+	if (err < 0) {
+		if (eeprom->len) {
+			netdev_warn(netdev,
+				    "Incomplete read from module EEPROM: %d\n",
+				     err);
+			err = 0;
+		} else {
+			netdev_err(netdev,
+				   "Reading from module EEPROM failed: %d\n",
+				   err);
+		}
+	}
+
+exit_close_nsp:
+	nfp_nsp_close(nsp);
+	return err;
+}
+
 static int nfp_net_set_coalesce(struct net_device *netdev,
 				struct ethtool_coalesce *ec)
 {
@@ -1253,6 +1380,8 @@ static const struct ethtool_ops nfp_net_ethtool_ops = {
 	.set_dump		= nfp_app_set_dump,
 	.get_dump_flag		= nfp_app_get_dump_flag,
 	.get_dump_data		= nfp_app_get_dump_data,
+	.get_module_info	= nfp_port_get_module_info,
+	.get_module_eeprom	= nfp_port_get_module_eeprom,
 	.get_coalesce           = nfp_net_get_coalesce,
 	.set_coalesce           = nfp_net_set_coalesce,
 	.get_channels		= nfp_net_get_channels,
@@ -1272,6 +1401,8 @@ const struct ethtool_ops nfp_port_ethtool_ops = {
 	.set_dump		= nfp_app_set_dump,
 	.get_dump_flag		= nfp_app_get_dump_flag,
 	.get_dump_data		= nfp_app_get_dump_data,
+	.get_module_info	= nfp_port_get_module_info,
+	.get_module_eeprom	= nfp_port_get_module_eeprom,
 	.get_link_ksettings	= nfp_net_get_link_ksettings,
 	.set_link_ksettings	= nfp_net_set_link_ksettings,
 	.get_fecparam		= nfp_port_get_fecparam,
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_main.c b/drivers/net/ethernet/netronome/nfp/nfp_net_main.c
index 08f5fdbd8e41..986464d4a206 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_net_main.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_net_main.c
@@ -150,34 +150,39 @@ nfp_net_pf_init_vnic(struct nfp_pf *pf, struct nfp_net *nn, unsigned int id)
 
 	nn->id = id;
 
+	if (nn->port) {
+		err = nfp_devlink_port_register(pf->app, nn->port);
+		if (err)
+			return err;
+	}
+
 	err = nfp_net_init(nn);
 	if (err)
-		return err;
+		goto err_devlink_port_clean;
 
 	nfp_net_debugfs_vnic_add(nn, pf->ddir);
 
-	if (nn->port) {
-		err = nfp_devlink_port_register(pf->app, nn->port);
-		if (err)
-			goto err_dfs_clean;
-	}
+	if (nn->port)
+		nfp_devlink_port_type_eth_set(nn->port);
 
 	nfp_net_info(nn);
 
 	if (nfp_net_is_data_vnic(nn)) {
 		err = nfp_app_vnic_init(pf->app, nn);
 		if (err)
-			goto err_devlink_port_clean;
+			goto err_devlink_port_type_clean;
 	}
 
 	return 0;
 
-err_devlink_port_clean:
+err_devlink_port_type_clean:
 	if (nn->port)
-		nfp_devlink_port_unregister(nn->port);
-err_dfs_clean:
+		nfp_devlink_port_type_clear(nn->port);
 	nfp_net_debugfs_dir_clean(&nn->debugfs_dir);
 	nfp_net_clean(nn);
+err_devlink_port_clean:
+	if (nn->port)
+		nfp_devlink_port_unregister(nn->port);
 	return err;
 }
 
@@ -221,9 +226,11 @@ static void nfp_net_pf_clean_vnic(struct nfp_pf *pf, struct nfp_net *nn)
 	if (nfp_net_is_data_vnic(nn))
 		nfp_app_vnic_clean(pf->app, nn);
 	if (nn->port)
-		nfp_devlink_port_unregister(nn->port);
+		nfp_devlink_port_type_clear(nn->port);
 	nfp_net_debugfs_dir_clean(&nn->debugfs_dir);
 	nfp_net_clean(nn);
+	if (nn->port)
+		nfp_devlink_port_unregister(nn->port);
 }
 
 static int nfp_net_pf_alloc_irqs(struct nfp_pf *pf)
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_repr.c b/drivers/net/ethernet/netronome/nfp/nfp_net_repr.c
index 94d228c04496..036edcc1fa18 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_net_repr.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_net_repr.c
@@ -267,13 +267,13 @@ const struct net_device_ops nfp_repr_netdev_ops = {
 	.ndo_set_vf_mac		= nfp_app_set_vf_mac,
 	.ndo_set_vf_vlan	= nfp_app_set_vf_vlan,
 	.ndo_set_vf_spoofchk	= nfp_app_set_vf_spoofchk,
+	.ndo_set_vf_trust	= nfp_app_set_vf_trust,
 	.ndo_get_vf_config	= nfp_app_get_vf_config,
 	.ndo_set_vf_link_state	= nfp_app_set_vf_link_state,
 	.ndo_fix_features	= nfp_repr_fix_features,
 	.ndo_set_features	= nfp_port_set_features,
 	.ndo_set_mac_address    = eth_mac_addr,
-	.ndo_get_port_parent_id	= nfp_port_get_port_parent_id,
-	.ndo_get_devlink	= nfp_devlink_get_devlink,
+	.ndo_get_devlink_port	= nfp_devlink_get_devlink_port,
 };
 
 void
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_sriov.c b/drivers/net/ethernet/netronome/nfp/nfp_net_sriov.c
index b6ec46ed0540..3fdaaf8ed2ba 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_net_sriov.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_net_sriov.c
@@ -1,5 +1,5 @@
 // SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
-/* Copyright (C) 2017 Netronome Systems, Inc. */
+/* Copyright (C) 2017-2019 Netronome Systems, Inc. */
 
 #include <linux/bitfield.h>
 #include <linux/errno.h>
@@ -146,6 +146,30 @@ int nfp_app_set_vf_spoofchk(struct net_device *netdev, int vf, bool enable)
 				    "spoofchk");
 }
 
+int nfp_app_set_vf_trust(struct net_device *netdev, int vf, bool enable)
+{
+	struct nfp_app *app = nfp_app_from_netdev(netdev);
+	unsigned int vf_offset;
+	u8 vf_ctrl;
+	int err;
+
+	err = nfp_net_sriov_check(app, vf, NFP_NET_VF_CFG_MB_CAP_TRUST,
+				  "trust");
+	if (err)
+		return err;
+
+	/* Write trust control bit to VF entry in VF config symbol */
+	vf_offset = NFP_NET_VF_CFG_MB_SZ + vf * NFP_NET_VF_CFG_SZ +
+		NFP_NET_VF_CFG_CTRL;
+	vf_ctrl = readb(app->pf->vfcfg_tbl2 + vf_offset);
+	vf_ctrl &= ~NFP_NET_VF_CFG_CTRL_TRUST;
+	vf_ctrl |= FIELD_PREP(NFP_NET_VF_CFG_CTRL_TRUST, enable);
+	writeb(vf_ctrl, app->pf->vfcfg_tbl2 + vf_offset);
+
+	return nfp_net_sriov_update(app, vf, NFP_NET_VF_CFG_MB_UPD_TRUST,
+				    "trust");
+}
+
 int nfp_app_set_vf_link_state(struct net_device *netdev, int vf,
 			      int link_state)
 {
@@ -213,6 +237,7 @@ int nfp_app_get_vf_config(struct net_device *netdev, int vf,
 	ivi->qos = FIELD_GET(NFP_NET_VF_CFG_VLAN_QOS, vlan_tci);
 
 	ivi->spoofchk = FIELD_GET(NFP_NET_VF_CFG_CTRL_SPOOF, flags);
+	ivi->trusted = FIELD_GET(NFP_NET_VF_CFG_CTRL_TRUST, flags);
 	ivi->linkstate = FIELD_GET(NFP_NET_VF_CFG_CTRL_LINK_STATE, flags);
 
 	return 0;
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_sriov.h b/drivers/net/ethernet/netronome/nfp/nfp_net_sriov.h
index c9f09c5bb5ee..a3db0cbf6425 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_net_sriov.h
+++ b/drivers/net/ethernet/netronome/nfp/nfp_net_sriov.h
@@ -1,5 +1,5 @@
 /* SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause) */
-/* Copyright (C) 2017 Netronome Systems, Inc. */
+/* Copyright (C) 2017-2019 Netronome Systems, Inc. */
 
 #ifndef _NFP_NET_SRIOV_H_
 #define _NFP_NET_SRIOV_H_
@@ -19,12 +19,14 @@
 #define   NFP_NET_VF_CFG_MB_CAP_VLAN			  (0x1 << 1)
 #define   NFP_NET_VF_CFG_MB_CAP_SPOOF			  (0x1 << 2)
 #define   NFP_NET_VF_CFG_MB_CAP_LINK_STATE		  (0x1 << 3)
+#define   NFP_NET_VF_CFG_MB_CAP_TRUST			  (0x1 << 4)
 #define NFP_NET_VF_CFG_MB_RET				0x2
 #define NFP_NET_VF_CFG_MB_UPD				0x4
 #define   NFP_NET_VF_CFG_MB_UPD_MAC			  (0x1 << 0)
 #define   NFP_NET_VF_CFG_MB_UPD_VLAN			  (0x1 << 1)
 #define   NFP_NET_VF_CFG_MB_UPD_SPOOF			  (0x1 << 2)
 #define   NFP_NET_VF_CFG_MB_UPD_LINK_STATE		  (0x1 << 3)
+#define   NFP_NET_VF_CFG_MB_UPD_TRUST			  (0x1 << 4)
 #define NFP_NET_VF_CFG_MB_VF_NUM			0x7
 
 /* VF config entry
@@ -35,6 +37,7 @@
 #define   NFP_NET_VF_CFG_MAC_HI				  0x0
 #define   NFP_NET_VF_CFG_MAC_LO				  0x6
 #define NFP_NET_VF_CFG_CTRL				0x4
+#define   NFP_NET_VF_CFG_CTRL_TRUST			  0x8
 #define   NFP_NET_VF_CFG_CTRL_SPOOF			  0x4
 #define   NFP_NET_VF_CFG_CTRL_LINK_STATE		  0x3
 #define     NFP_NET_VF_CFG_LS_MODE_AUTO			    0
@@ -48,6 +51,7 @@ int nfp_app_set_vf_mac(struct net_device *netdev, int vf, u8 *mac);
 int nfp_app_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos,
 			__be16 vlan_proto);
 int nfp_app_set_vf_spoofchk(struct net_device *netdev, int vf, bool setting);
+int nfp_app_set_vf_trust(struct net_device *netdev, int vf, bool setting);
 int nfp_app_set_vf_link_state(struct net_device *netdev, int vf,
 			      int link_state);
 int nfp_app_get_vf_config(struct net_device *netdev, int vf,
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_netvf_main.c b/drivers/net/ethernet/netronome/nfp/nfp_netvf_main.c
index 1145849ca7ba..e4977cdf7678 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_netvf_main.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_netvf_main.c
@@ -282,8 +282,14 @@ err_free_vf:
 
 static void nfp_netvf_pci_remove(struct pci_dev *pdev)
 {
-	struct nfp_net_vf *vf = pci_get_drvdata(pdev);
-	struct nfp_net *nn = vf->nn;
+	struct nfp_net_vf *vf;
+	struct nfp_net *nn;
+
+	vf = pci_get_drvdata(pdev);
+	if (!vf)
+		return;
+
+	nn = vf->nn;
 
 	/* Note, the order is slightly different from above as we need
 	 * to keep the nn pointer around till we have freed everything.
@@ -317,4 +323,5 @@ struct pci_driver nfp_netvf_pci_driver = {
 	.id_table    = nfp_netvf_pci_device_ids,
 	.probe       = nfp_netvf_pci_probe,
 	.remove      = nfp_netvf_pci_remove,
+	.shutdown    = nfp_netvf_pci_remove,
 };
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_port.c b/drivers/net/ethernet/netronome/nfp/nfp_port.c
index 93c5bfc0510b..fcd16877e6e0 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_port.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_port.c
@@ -30,22 +30,6 @@ struct nfp_port *nfp_port_from_netdev(struct net_device *netdev)
 	return NULL;
 }
 
-int nfp_port_get_port_parent_id(struct net_device *netdev,
-				struct netdev_phys_item_id *ppid)
-{
-	struct nfp_port *port;
-	const u8 *serial;
-
-	port = nfp_port_from_netdev(netdev);
-	if (!port)
-		return -EOPNOTSUPP;
-
-	ppid->id_len = nfp_cpp_serial(port->app->cpp, &serial);
-	memcpy(&ppid->id, serial, ppid->id_len);
-
-	return 0;
-}
-
 int nfp_port_setup_tc(struct net_device *netdev, enum tc_setup_type type,
 		      void *type_data)
 {
diff --git a/drivers/net/ethernet/netronome/nfp/nfp_port.h b/drivers/net/ethernet/netronome/nfp/nfp_port.h
index 90ae053f5c07..d7fd203bb180 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_port.h
+++ b/drivers/net/ethernet/netronome/nfp/nfp_port.h
@@ -131,6 +131,8 @@ int nfp_net_refresh_port_table_sync(struct nfp_pf *pf);
 
 int nfp_devlink_port_register(struct nfp_app *app, struct nfp_port *port);
 void nfp_devlink_port_unregister(struct nfp_port *port);
+void nfp_devlink_port_type_eth_set(struct nfp_port *port);
+void nfp_devlink_port_type_clear(struct nfp_port *port);
 
 /**
  * Mac stats (0x0000 - 0x0200)
diff --git a/drivers/net/ethernet/netronome/nfp/nfpcore/nfp_nsp.c b/drivers/net/ethernet/netronome/nfp/nfpcore/nfp_nsp.c
index 3a4e224a64b7..42cf4fd875ea 100644
--- a/drivers/net/ethernet/netronome/nfp/nfpcore/nfp_nsp.c
+++ b/drivers/net/ethernet/netronome/nfp/nfpcore/nfp_nsp.c
@@ -79,6 +79,8 @@
 #define NFP_VERSIONS_NCSI_OFF	22
 #define NFP_VERSIONS_CFGR_OFF	26
 
+#define NSP_SFF_EEPROM_BLOCK_LEN	8
+
 enum nfp_nsp_cmd {
 	SPCODE_NOOP		= 0, /* No operation */
 	SPCODE_SOFT_RESET	= 1, /* Soft reset the NFP */
@@ -95,6 +97,7 @@ enum nfp_nsp_cmd {
 	SPCODE_FW_STORED	= 16, /* If no FW loaded, load flash app FW */
 	SPCODE_HWINFO_LOOKUP	= 17, /* Lookup HWinfo with overwrites etc. */
 	SPCODE_VERSIONS		= 21, /* Report FW versions */
+	SPCODE_READ_SFF_EEPROM	= 22, /* Read module EEPROM */
 };
 
 struct nfp_nsp_dma_buf {
@@ -965,3 +968,62 @@ const char *nfp_nsp_versions_get(enum nfp_nsp_versions id, bool flash,
 
 	return (const char *)&buf[buf_off];
 }
+
+static int
+__nfp_nsp_module_eeprom(struct nfp_nsp *state, void *buf, unsigned int size)
+{
+	struct nfp_nsp_command_buf_arg module_eeprom = {
+		{
+			.code		= SPCODE_READ_SFF_EEPROM,
+			.option		= size,
+		},
+		.in_buf		= buf,
+		.in_size	= size,
+		.out_buf	= buf,
+		.out_size	= size,
+	};
+
+	return nfp_nsp_command_buf(state, &module_eeprom);
+}
+
+int nfp_nsp_read_module_eeprom(struct nfp_nsp *state, int eth_index,
+			       unsigned int offset, void *data,
+			       unsigned int len, unsigned int *read_len)
+{
+	struct eeprom_buf {
+		u8 metalen;
+		__le16 length;
+		__le16 offset;
+		__le16 readlen;
+		u8 eth_index;
+		u8 data[0];
+	} __packed *buf;
+	int bufsz, ret;
+
+	BUILD_BUG_ON(offsetof(struct eeprom_buf, data) % 8);
+
+	/* Buffer must be large enough and rounded to the next block size. */
+	bufsz = struct_size(buf, data, round_up(len, NSP_SFF_EEPROM_BLOCK_LEN));
+	buf = kzalloc(bufsz, GFP_KERNEL);
+	if (!buf)
+		return -ENOMEM;
+
+	buf->metalen =
+		offsetof(struct eeprom_buf, data) / NSP_SFF_EEPROM_BLOCK_LEN;
+	buf->length = cpu_to_le16(len);
+	buf->offset = cpu_to_le16(offset);
+	buf->eth_index = eth_index;
+
+	ret = __nfp_nsp_module_eeprom(state, buf, bufsz);
+
+	*read_len = min_t(unsigned int, len, le16_to_cpu(buf->readlen));
+	if (*read_len)
+		memcpy(data, buf->data, *read_len);
+
+	if (!ret && *read_len < len)
+		ret = -EIO;
+
+	kfree(buf);
+
+	return ret;
+}
diff --git a/drivers/net/ethernet/netronome/nfp/nfpcore/nfp_nsp.h b/drivers/net/ethernet/netronome/nfp/nfpcore/nfp_nsp.h
index bd9c358c646f..22ee6985ee1c 100644
--- a/drivers/net/ethernet/netronome/nfp/nfpcore/nfp_nsp.h
+++ b/drivers/net/ethernet/netronome/nfp/nfpcore/nfp_nsp.h
@@ -22,6 +22,9 @@ int nfp_nsp_write_flash(struct nfp_nsp *state, const struct firmware *fw);
 int nfp_nsp_mac_reinit(struct nfp_nsp *state);
 int nfp_nsp_load_stored_fw(struct nfp_nsp *state);
 int nfp_nsp_hwinfo_lookup(struct nfp_nsp *state, void *buf, unsigned int size);
+int nfp_nsp_read_module_eeprom(struct nfp_nsp *state, int eth_index,
+			       unsigned int offset, void *data,
+			       unsigned int len, unsigned int *read_len);
 
 static inline bool nfp_nsp_has_mac_reinit(struct nfp_nsp *state)
 {
@@ -43,6 +46,11 @@ static inline bool nfp_nsp_has_versions(struct nfp_nsp *state)
 	return nfp_nsp_get_abi_ver_minor(state) > 27;
 }
 
+static inline bool nfp_nsp_has_read_module_eeprom(struct nfp_nsp *state)
+{
+	return nfp_nsp_get_abi_ver_minor(state) > 28;
+}
+
 enum nfp_eth_interface {
 	NFP_INTERFACE_NONE	= 0,
 	NFP_INTERFACE_SFP	= 1,