[ovs-dev] [PATCH] Extends the existing mirror configuration parameters

Timothy Miskell timothy.miskell at intel.com
Mon May 10 16:00:45 UTC 2021


From: Liang-min Wang <liang-min.wang at intel.com>

The following parameters are added:
 - mirror-offload: to turn on/off mirror offloading.
 - output-port-name: specify a port, using name string, that is on a different
   bridge
 - output-src-vlan: output port vlan for each select-src-port.
 - output-dst-vlan: output port vlan for each select-dst-port.
 - flow-src-mac: use src mac address of each select-dst-port for the header
   scan.
 - flow-dst-mac: use dst mac address of each select-src-port for the header
   scan.
 - mirror-tunnel-addr: BDF string of the tunnel device.

ovs-vsctl test change because new mirroring parameters are introduced in this patch

Create a defer procedure call thread to handle all mirror offload requests.
This is a light-weight thread which remains in sleep-state when there is no new request.
This is created between ovs-vsctl and mirror offloading back end

Implementing DPDK tx-burst (VIRTIO ingress traffic
mirror) and rx-burst (VIRTIO egress traffic mirror) callbacks.
Each callback  functions implement the following tasks:
 1. Enable per-packet VLAN insertion
   - for port mirroring, all packets are enabled per-packet VLAN insertion.
   - for flow mirroring, only packet header matches the required mac address
     are enabled.
 2. Sending the packets to the specified transport port (output-port in
    mirror offload configuration)
   - for port mirroring, all packets are sent to the transport port.
   - for flow mirroring, only matched packets are sent.
 3. Restore each packet attributes (remove DPDK per-packet offload flag)

Signed-off-by: Liang-min Wang <liang-min.wang at intel.com>
Tested-by: Timothy Miskell <Timothy.Miskell at intel.com>
Suggested-by: Munish Mehan <mm6021 at att.com>
---
 lib/automake.mk            |   2 +
 lib/netdev-dpdk-mirror.c   | 516 +++++++++++++++++++++++++++++++++++++
 lib/netdev-dpdk-mirror.h   |  83 ++++++
 lib/netdev-dpdk.c          | 397 ++++++++++++++++++++++++++++
 lib/netdev-provider.h      |  16 ++
 lib/netdev.c               | 386 +++++++++++++++++++++++++++
 lib/netdev.h               |  16 ++
 tests/ovs-vsctl.at         |   2 +
 vswitchd/bridge.c          | 271 ++++++++++++++++++-
 vswitchd/vswitch.ovsschema |  24 +-
 vswitchd/vswitch.xml       |  50 ++++
 11 files changed, 1759 insertions(+), 4 deletions(-)
 create mode 100644 lib/netdev-dpdk-mirror.c
 create mode 100644 lib/netdev-dpdk-mirror.h

diff --git a/lib/automake.mk b/lib/automake.mk
index 39901bd6d..dcafbfaca 100644
--- a/lib/automake.mk
+++ b/lib/automake.mk
@@ -170,6 +170,7 @@ lib_libopenvswitch_la_SOURCES = \
 	lib/multipath.h \
 	lib/namemap.c \
 	lib/netdev-dpdk.h \
+	lib/netdev-dpdk-mirror.h \
 	lib/netdev-dummy.c \
 	lib/netdev-offload.c \
 	lib/netdev-offload.h \
@@ -460,6 +461,7 @@ if DPDK_NETDEV
 lib_libopenvswitch_la_SOURCES += \
 	lib/dpdk.c \
 	lib/netdev-dpdk.c \
+	lib/netdev-dpdk-mirror.c \
 	lib/netdev-offload-dpdk.c
 else
 lib_libopenvswitch_la_SOURCES += \
diff --git a/lib/netdev-dpdk-mirror.c b/lib/netdev-dpdk-mirror.c
new file mode 100644
index 000000000..ff2701660
--- /dev/null
+++ b/lib/netdev-dpdk-mirror.c
@@ -0,0 +1,516 @@
+/*
+ * Copyright (c) 2014, 2015, 2016, 2017 Nicira, Inc.
+ * Copyright (c) 2019 Mellanox Technologies, Ltd.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+#include <config.h>
+#include <rte_ethdev.h>
+
+#include "netdev-dpdk-mirror.h"
+#include "openvswitch/vlog.h"
+#include "openvswitch/dynamic-string.h"
+#include "util.h"
+
+#define MAC_ADDR_MAP           0x0000FFFFFFFFFFFFULL
+#define is_mac_addr_match(a,b) (((a^b)&MAC_ADDR_MAP) == 0)
+#define INIT_MIRROR_DB_SIZE    8
+#define INVALID_DEVICE_ID      0xFFFFFFFF
+
+VLOG_DEFINE_THIS_MODULE(netdev_dpdk_mirror);
+
+/* port/flow mirror database management routines */
+/*
+ * The below API is for port/flow mirror offloading which uses a different DPDK
+ * interface as rte-flow.
+ */
+static int mirror_port_db_size = 0;
+static int mirror_port_used = 0;
+static struct mirror_offload_port *mirror_port_db = NULL;
+
+static void
+netdev_mirror_db_init(struct mirror_offload_port *db, int size)
+{
+    int i;
+
+    for (i = 0; i < size; i++) {
+        db[i].dev_id = INVALID_DEVICE_ID;
+        memset(&db[i].rx, 0, sizeof(struct mirror_param));
+        memset(&db[i].tx, 0, sizeof(struct mirror_param));
+    }
+}
+
+/* Double the db size when it runs out of space */
+static int
+netdev_mirror_db_resize(void)
+{
+    int new_size = mirror_port_db_size << 1;
+    struct mirror_offload_port *new_db = xmalloc(
+        sizeof(struct mirror_offload_port)*new_size);
+
+    memcpy(new_db, mirror_port_db, sizeof(struct mirror_offload_port)
+        *mirror_port_db_size);
+    netdev_mirror_db_init(&new_db[mirror_port_db_size], mirror_port_db_size);
+    mirror_port_db_size = new_size;
+    mirror_port_db = new_db;
+
+    return 0;
+}
+
+
+static struct mirror_offload_port*
+netdev_mirror_data_find(uint32_t dev_id)
+{
+    int i;
+
+    if (mirror_port_db == NULL) {
+        return NULL;
+    }
+
+    for (i = 0; i < mirror_port_db_size; i++) {
+        if (dev_id == mirror_port_db[i].dev_id) {
+            return &mirror_port_db[i];
+        }
+    }
+    return NULL;
+}
+
+static struct mirror_offload_port*
+netdev_mirror_data_add(uint32_t dev_id, int tx,
+    struct mirror_param *new_param)
+{
+    struct mirror_offload_port *target = NULL;
+    int i;
+
+    if (!mirror_port_db) {
+        mirror_port_db_size = INIT_MIRROR_DB_SIZE;
+        mirror_port_db = xmalloc(sizeof(struct mirror_offload_port)*
+            mirror_port_db_size);
+        netdev_mirror_db_init(mirror_port_db, mirror_port_db_size);
+    }
+    target = netdev_mirror_data_find(dev_id);
+    if (target) {
+        if (tx) {
+            if (target->tx.mirror_cb) {
+                VLOG_ERR("Attempt to add ingress mirror offloading"
+                    " on port, %d, while one is outstanding\n", dev_id);
+                return target;
+            }
+
+            memcpy(&target->tx, new_param, sizeof(*new_param));
+        } else {
+            if (target->rx.mirror_cb) {
+                VLOG_ERR("Attempt to add egress mirror offloading"
+                    " on port, %d, while one is outstanding\n", dev_id);
+                return target;
+            }
+
+            memcpy(&target->rx, new_param, sizeof(struct mirror_param));
+        }
+    } else {
+        struct mirror_param *param;
+        /* find an unused spot on db */
+        for (i = 0; i < mirror_port_db_size; i++) {
+            if (mirror_port_db[i].dev_id == INVALID_DEVICE_ID) {
+                break;
+            }
+        }
+        if (i == mirror_port_db_size && netdev_mirror_db_resize()) {
+                return NULL;
+        }
+
+        param = tx ? &mirror_port_db[i].tx : &mirror_port_db[i].rx;
+        memcpy(param, new_param, sizeof(struct mirror_param));
+
+        target = &mirror_port_db[i];
+        target->dev_id = dev_id;
+        mirror_port_used ++;
+    }
+    return target;
+}
+
+static void
+netdev_mirror_data_remove(uint32_t dev_id, int tx) {
+    struct mirror_offload_port *target = netdev_mirror_data_find(dev_id);
+
+    if (!target) {
+        VLOG_ERR("Attempt to remove unsaved port, %d, %s callback\n",
+        dev_id, tx?"tx": "rx");
+    }
+
+    if (tx) {
+        memset(&target->tx, 0, sizeof(struct mirror_param));
+    } else {
+        memset(&target->rx, 0, sizeof(struct mirror_param));
+    }
+
+    if ((target->rx.mirror_cb == NULL) &&
+        (target->tx.mirror_cb == NULL)) {
+        target->dev_id = INVALID_DEVICE_ID;
+        mirror_port_used --;
+        /* release port mirror db memory when there
+         * is no outstanding port mirror offloading
+         * configuration
+         */
+        if (mirror_port_used == 0) {
+            free(mirror_port_db);
+            mirror_port_db = NULL;
+            mirror_port_db_size = 0;
+        }
+    }
+}
+
+void
+netdev_mirror_data_proc(uint32_t dev_id, mirror_data_op op,
+    int tx, struct mirror_param *in_param,
+    struct mirror_offload_port **out_param)
+{
+    switch (op) {
+    case mirror_data_find:
+        *out_param = netdev_mirror_data_find(dev_id);
+        break;
+    case mirror_data_add:
+        *out_param = netdev_mirror_data_add(dev_id, tx, in_param);
+        break;
+    case mirror_data_rem:
+        netdev_mirror_data_remove(dev_id, tx);
+        break;
+    }
+}
+
+/* port/flow mirror traffic processors */
+static inline uint16_t
+netdev_custom_mirror_offload_cb(uint16_t qidx, struct rte_mbuf **pkts,
+    uint16_t nb_pkts, void *user_params)
+{
+    struct mirror_param *data = user_params;
+    uint16_t i, dst_qidx, match_count = 0;
+    uint16_t pkt_trans;
+    uint16_t dst_port_id = data->dst_port_id;
+    uint16_t dst_vlan_id = data->dst_vlan_id;
+    struct rte_mbuf **pkt_buf = &data->pkt_buf[qidx * data->max_burst_size];
+
+    if (nb_pkts == 0) {
+        return 0;
+    }
+
+    if (nb_pkts > data->max_burst_size) {
+        VLOG_ERR("Per-flow batch size, %d, exceeds maximum limit\n", nb_pkts);
+        return 0;
+    }
+
+    for (i = 0; i < nb_pkts; i++) {
+        if (data->custom_scan(pkts[i], user_params)) {
+            pkt_buf[match_count] = pkts[i];
+            pkt_buf[match_count]->ol_flags |= PKT_TX_VLAN_PKT;
+            pkt_buf[match_count]->vlan_tci = dst_vlan_id;
+            rte_mbuf_refcnt_update(pkt_buf[match_count], 1);
+            match_count++;
+        }
+    }
+
+    dst_qidx = (data->n_dst_queue > qidx)?qidx:(data->n_dst_queue -1);
+
+    rte_spinlock_lock(&data->locks[dst_qidx]);
+    pkt_trans = rte_eth_tx_burst(dst_port_id, dst_qidx, pkt_buf, match_count);
+    rte_spinlock_unlock(&data->locks[dst_qidx]);
+
+    for (i = 0; i < match_count; i++) {
+        pkt_buf[i]->ol_flags &= ~PKT_TX_VLAN_PKT;
+    }
+
+    while (unlikely (pkt_trans < match_count)) {
+        rte_pktmbuf_free(pkt_buf[pkt_trans]);
+        pkt_trans++;
+    }
+
+    return nb_pkts;
+}
+
+static inline uint16_t
+netdev_flow_mirror_offload_cb(uint16_t qidx, struct rte_mbuf **pkts,
+    uint16_t nb_pkts, void *user_params, uint32_t offset)
+{
+    struct mirror_param *data = user_params;
+    uint16_t i, dst_qidx, match_count = 0;
+    uint16_t pkt_trans;
+    uint16_t dst_port_id = data->dst_port_id;
+    uint16_t dst_vlan_id = data->dst_vlan_id;
+    uint64_t target_addr = *(uint64_t *) data->extra_data;
+    struct rte_mbuf **pkt_buf = &data->pkt_buf[qidx * data->max_burst_size];
+
+    if (nb_pkts == 0) {
+        return 0;
+    }
+
+    if (nb_pkts > data->max_burst_size) {
+        VLOG_ERR("Per-flow batch size, %d, exceeds maximum limit\n", nb_pkts);
+        return 0;
+    }
+
+    for (i = 0; i < nb_pkts; i++) {
+        uint64_t *dst_mac_addr =
+            rte_pktmbuf_mtod_offset(pkts[i], void *, offset);
+        if (is_mac_addr_match(target_addr, (*dst_mac_addr))) {
+            pkt_buf[match_count] = pkts[i];
+            pkt_buf[match_count]->ol_flags |= PKT_TX_VLAN_PKT;
+            pkt_buf[match_count]->vlan_tci = dst_vlan_id;
+            rte_mbuf_refcnt_update(pkt_buf[match_count], 1);
+            match_count ++;
+        }
+    }
+
+    dst_qidx = (data->n_dst_queue > qidx) ? qidx : (data->n_dst_queue -1);
+
+    rte_spinlock_lock(&data->locks[dst_qidx]);
+    pkt_trans = rte_eth_tx_burst(dst_port_id, dst_qidx, pkt_buf, match_count);
+    rte_spinlock_unlock(&data->locks[dst_qidx]);
+
+    for (i = 0; i < match_count; i++) {
+        pkt_buf[i]->ol_flags &= ~PKT_TX_VLAN_PKT;
+    }
+
+    while (unlikely (pkt_trans < match_count)) {
+        rte_pktmbuf_free(pkt_buf[pkt_trans]);
+        pkt_trans++;
+    }
+
+    return nb_pkts;
+}
+
+static inline uint16_t
+netdev_port_mirror_offload_cb(uint16_t qidx, struct rte_mbuf **pkts,
+    uint16_t nb_pkts, void *user_params)
+{
+    struct mirror_param *data = user_params;
+    uint16_t i, dst_qidx;
+    uint16_t pkt_trans;
+    uint16_t dst_port_id = data->dst_port_id;
+    uint16_t dst_vlan_id = data->dst_vlan_id;
+
+    if (nb_pkts == 0) {
+        return 0;
+    }
+
+    for (i = 0; i < nb_pkts; i++) {
+        pkts[i]->ol_flags |= PKT_TX_VLAN_PKT;
+        pkts[i]->vlan_tci = dst_vlan_id;
+        rte_mbuf_refcnt_update(pkts[i], 1);
+    }
+
+    dst_qidx = (data->n_dst_queue > qidx) ? qidx : (data->n_dst_queue -1);
+
+    rte_spinlock_lock(&data->locks[dst_qidx]);
+    pkt_trans = rte_eth_tx_burst(dst_port_id, dst_qidx, pkts, nb_pkts);
+    rte_spinlock_unlock(&data->locks[dst_qidx]);
+
+    for (i = 0; i < nb_pkts; i++) {
+        pkts[i]->ol_flags &= ~PKT_TX_VLAN_PKT;
+    }
+
+    while (unlikely (pkt_trans < nb_pkts)) {
+        rte_pktmbuf_free(pkts[pkt_trans]);
+        pkt_trans++;
+    }
+
+    return nb_pkts;
+}
+
+static inline uint16_t
+netdev_rx_custom_mirror_offload_cb(uint16_t port_id OVS_UNUSED,
+    uint16_t qidx, struct rte_mbuf **pkts, uint16_t nb_pkts,
+    uint16_t maxi_pkts OVS_UNUSED, void *user_params)
+{
+    return netdev_custom_mirror_offload_cb(qidx, pkts, nb_pkts, user_params);
+}
+
+static inline uint16_t
+netdev_tx_custom_mirror_offload_cb(uint16_t port_id OVS_UNUSED,
+    uint16_t qidx, struct rte_mbuf **pkts, uint16_t nb_pkts,
+    void *user_params)
+{
+    return netdev_custom_mirror_offload_cb(qidx, pkts, nb_pkts, user_params);
+}
+
+static inline uint16_t
+netdev_rx_flow_mirror_offload_cb(uint16_t port_id OVS_UNUSED,
+    uint16_t qidx, struct rte_mbuf **pkts, uint16_t nb_pkts,
+    uint16_t maxi_pkts OVS_UNUSED, void *user_params)
+{
+    return netdev_flow_mirror_offload_cb(qidx, pkts, nb_pkts, user_params, 0);
+}
+
+static inline uint16_t
+netdev_tx_flow_mirror_offload_cb(uint16_t port_id OVS_UNUSED,
+    uint16_t qidx, struct rte_mbuf **pkts, uint16_t nb_pkts,
+    void *user_params)
+{
+    return netdev_flow_mirror_offload_cb(qidx, pkts, nb_pkts, user_params, 6);
+}
+
+static inline uint16_t
+netdev_rx_port_mirror_offload_cb(uint16_t port_id OVS_UNUSED,
+    uint16_t qidx, struct rte_mbuf **pkts, uint16_t nb_pkts,
+    uint16_t max_pkts OVS_UNUSED, void *user_params)
+{
+    return netdev_port_mirror_offload_cb(qidx, pkts, nb_pkts, user_params);
+}
+
+static inline uint16_t
+netdev_tx_port_mirror_offload_cb(uint16_t port_id OVS_UNUSED,
+    uint16_t qidx, struct rte_mbuf **pkts, uint16_t nb_pkts,
+    void *user_params)
+{
+    return netdev_port_mirror_offload_cb(qidx, pkts, nb_pkts, user_params);
+}
+
+static rte_rx_callback_fn
+netdev_mirror_rx_cb(rte_mirror_type mirror_type)
+{
+    switch (mirror_type) {
+    case mirror_port:
+        return netdev_rx_port_mirror_offload_cb;
+    case mirror_flow_mac:
+        return netdev_rx_flow_mirror_offload_cb;
+    case mirror_flow_custom:
+        return netdev_rx_custom_mirror_offload_cb;
+    case mirror_invalid:
+        return NULL;
+    }
+    VLOG_ERR("Un-supported mirror type\n");
+    return NULL;
+}
+
+static rte_tx_callback_fn
+netdev_mirror_tx_cb(rte_mirror_type mirror_type)
+{
+    switch (mirror_type) {
+    case mirror_port:
+        return netdev_tx_port_mirror_offload_cb;
+    case mirror_flow_mac:
+        return netdev_tx_flow_mirror_offload_cb;
+        break;
+    case mirror_flow_custom:
+        return netdev_tx_custom_mirror_offload_cb;
+    case mirror_invalid:
+        return NULL;
+    }
+    VLOG_ERR("Un-supported mirror type\n");
+    return NULL;
+}
+
+void
+netdev_mirror_cb_set(struct mirror_param *data, uint16_t port_id,
+    int pmd_cb, int tx)
+{
+    unsigned int qid;
+
+    data->pkt_buf = NULL;
+    if (data->extra_data_size) {
+        data->pkt_buf = xmalloc(sizeof(mirror_fn_cb)*data->max_burst_size *
+            data->n_src_queue);
+    }
+
+    data->mirror_cb = xmalloc(sizeof(struct rte_eth_rxtx_callback *)
+        * data->n_src_queue);
+    for (qid = 0; qid < data->n_src_queue; qid++) {
+        if (pmd_cb) {
+            if (tx) {
+                data->mirror_cb[qid].pmd = rte_eth_add_tx_callback(port_id,
+                    qid, netdev_mirror_tx_cb(data->mirror_type), data);
+            } else {
+                data->mirror_cb[qid].pmd = rte_eth_add_rx_callback(port_id,
+                    qid, netdev_mirror_rx_cb(data->mirror_type), data);
+            }
+        } else {
+            struct rte_eth_rxtx_callback *rxtx_cb =
+                xmalloc(sizeof(struct rte_eth_rxtx_callback));
+
+            data->mirror_cb[qid].direct = rxtx_cb;
+            rxtx_cb->next = NULL;
+            rxtx_cb->param = data;
+
+            if (tx) {
+                rxtx_cb->fn.tx = netdev_mirror_tx_cb(data->mirror_type);
+            } else {
+                rxtx_cb->fn.rx = netdev_mirror_rx_cb(data->mirror_type);
+            }
+        }
+    }
+}
+
+/* port/flow mirroring device (port) register/un-registe routines */
+int
+netdev_eth_register_mirror(uint16_t src_port, struct mirror_param *param,
+    int tx_cb)
+{
+    struct mirror_offload_port *port_info = NULL;
+    struct mirror_param *data;
+
+    netdev_mirror_data_proc(src_port, mirror_data_add, tx_cb, param,
+        &port_info);
+    if (!port_info) {
+        return -1;
+    }
+
+    data = tx_cb ? &port_info->tx : &port_info->rx;
+    netdev_mirror_cb_set(data, src_port, 1, tx_cb);
+
+    return 0;
+}
+
+int
+netdev_eth_unregister_mirror(uint16_t src_port, int tx_cb)
+{
+    /* release both cb and pkt_buf */
+    unsigned int i;
+    struct mirror_offload_port *port_info = NULL;
+    struct mirror_param *data;
+
+    netdev_mirror_data_proc(src_port, mirror_data_find, tx_cb, NULL,
+        &port_info);
+    if (port_info == NULL) {
+        VLOG_ERR("Source port %d is not on outstanding port mirror db\n",
+            src_port);
+        return -1;
+    }
+    data = tx_cb ? &port_info->tx : &port_info->rx;
+
+    for (i = 0; i < data->n_src_queue; i++) {
+        if (data->mirror_cb[i].pmd) {
+            if (tx_cb) {
+                rte_eth_remove_tx_callback(src_port, i,
+                    data->mirror_cb[i].pmd);
+            } else {
+                rte_eth_remove_rx_callback(src_port, i,
+                    data->mirror_cb[i].pmd);
+            }
+        }
+        data->mirror_cb[i].pmd = NULL;
+    }
+    free(data->mirror_cb);
+
+    if (data->pkt_buf) {
+        free(data->pkt_buf);
+        data->pkt_buf = NULL;
+    }
+
+    if (data->extra_data) {
+        free(data->extra_data);
+        data->extra_data = NULL;
+        data->extra_data_size = 0;
+    }
+
+    netdev_mirror_data_proc(src_port, mirror_data_rem, tx_cb, NULL, NULL);
+    return 0;
+}
diff --git a/lib/netdev-dpdk-mirror.h b/lib/netdev-dpdk-mirror.h
new file mode 100644
index 000000000..ee4b933ba
--- /dev/null
+++ b/lib/netdev-dpdk-mirror.h
@@ -0,0 +1,83 @@
+/*
+ * Copyright (c) 2014, 2015, 2016 Nicira, Inc.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef NETDEV_DPDK_MIRROR_H
+#define NETDEV_DPDK_MIRROR_H
+
+#include "openvswitch/types.h"
+
+#ifdef  __cplusplus
+extern "C" {
+#endif
+
+typedef enum {
+    mirror_data_find, /* find the mirror-data allocated */
+    mirror_data_add, /* add a new mirror_param data int DB */
+    mirror_data_rem, /* remove a mirror_param from the DB */
+} mirror_data_op;
+
+typedef int (*rte_mirror_scan_fn)(struct rte_mbuf *pkt, void *user_param);
+typedef enum {
+    mirror_port, /* port mirror */
+    mirror_flow_mac, /* flow mirror according to source mac */
+    mirror_flow_custom,  /* flow mirror according to a callback scn */
+    mirror_invalid,      /* invalid mirror_type */
+} rte_mirror_type;
+
+typedef union {
+    const struct rte_eth_rxtx_callback *pmd;
+    struct rte_eth_rxtx_callback *direct;
+} mirror_fn_cb;
+
+struct mirror_param {
+    uint16_t dst_port_id;
+    uint16_t dst_vlan_id;
+    rte_spinlock_t *locks;
+    int n_src_queue;
+    int n_dst_queue;
+    struct rte_mbuf **pkt_buf;
+    mirror_fn_cb *mirror_cb;
+    unsigned int max_burst_size;
+    rte_mirror_scan_fn custom_scan;
+    rte_mirror_type mirror_type;
+    unsigned int extra_data_size;
+    void *extra_data; /* extra mirror parameter */
+};
+
+struct mirror_offload_port {
+    uint32_t dev_id;
+    struct mirror_param rx;
+    struct mirror_param tx;
+};
+
+bool netdev_port_started(uint16_t port_id, uint32_t *num_tx_queue);
+int netdev_get_portid_from_addr(const char *pci_addr_str, uint16_t *port_id);
+int netdev_tunnel_port_setup(uint16_t portid, uint32_t *num_queue);
+
+void netdev_mirror_data_proc(uint32_t dev_id, mirror_data_op op,
+    int tx, struct mirror_param *in_param,
+    struct mirror_offload_port **out_param);
+void netdev_mirror_cb_set(struct mirror_param *data, uint16_t port_id,
+    int pmd, int tx);
+int netdev_eth_register_mirror(uint16_t src_port,
+    struct mirror_param *param, int tx_cb);
+int netdev_eth_unregister_mirror(uint16_t src_port, int tx_cb);
+
+#ifdef  __cplusplus
+}
+#endif
+
+#endif /* netdev-dpdk-mirror.h */
diff --git a/lib/netdev-dpdk.c b/lib/netdev-dpdk.c
index 9d8096668..eb6644333 100644
--- a/lib/netdev-dpdk.c
+++ b/lib/netdev-dpdk.c
@@ -48,6 +48,7 @@
 #include "fatal-signal.h"
 #include "if-notifier.h"
 #include "netdev-provider.h"
+#include "netdev-dpdk-mirror.h"
 #include "netdev-vport.h"
 #include "odp-util.h"
 #include "openvswitch/dynamic-string.h"
@@ -171,6 +172,16 @@ static const struct rte_eth_conf port_conf = {
     },
 };
 
+struct mirror_tunnel_port_info {
+    uint16_t port_id;
+    rte_spinlock_t *locks;
+    uint32_t share_count;
+    uint32_t num_queue;
+    bool port_started;
+    struct mirror_tunnel_port_info *next;
+};
+static struct mirror_tunnel_port_info *mirror_tunnel_head = NULL;
+
 /*
  * These callbacks allow virtio-net devices to be added to vhost ports when
  * configuration has been fully completed.
@@ -443,6 +454,8 @@ struct netdev_dpdk {
         };
         struct dpdk_tx_queue *tx_q;
         struct rte_eth_link link;
+        mirror_fn_cb *rx_cb; /* shared pointer */
+        mirror_fn_cb *tx_cb;
     );
 
     PADDED_MEMBERS_CACHELINE_MARKER(CACHE_LINE_SIZE, cacheline1,
@@ -2417,6 +2430,13 @@ netdev_dpdk_vhost_rxq_recv(struct netdev_rxq *rxq,
     nb_rx = rte_vhost_dequeue_burst(vid, qid, dev->dpdk_mp->mp,
                                     (struct rte_mbuf **) batch->packets,
                                     NETDEV_MAX_BURST);
+
+    if (dev->rx_cb && dev->rx_cb[qid].direct->fn.rx) {
+        dev->rx_cb[qid].direct->fn.rx((uint16_t) vid, qid,
+        (struct rte_mbuf **) batch->packets, nb_rx,
+        NETDEV_MAX_BURST, dev->rx_cb[qid].direct->param);
+    }
+
     if (!nb_rx) {
         return EAGAIN;
     }
@@ -2634,6 +2654,10 @@ __netdev_dpdk_vhost_send(struct netdev *netdev, int qid,
         int vhost_qid = qid * VIRTIO_QNUM + VIRTIO_RXQ;
         unsigned int tx_pkts;
 
+        if (dev->tx_cb && dev->tx_cb[qid].direct->fn.tx) {
+            dev->tx_cb[qid].direct->fn.tx((uint16_t) vid, qid, cur_pkts, cnt,
+                dev->tx_cb[qid].direct->param);
+        }
         tx_pkts = rte_vhost_enqueue_burst(vid, vhost_qid, cur_pkts, cnt);
         if (OVS_LIKELY(tx_pkts)) {
             /* Packets have been sent.*/
@@ -5291,6 +5315,376 @@ netdev_dpdk_rte_flow_query_count(struct netdev *netdev,
     return ret;
 }
 
+/*
+ * mirror tunnel device management routines
+ * mirror tunnel devices are devices reserved solely for
+ * traffic mirroring
+ */
+static void
+netdev_dpdk_update_mt_list(struct mirror_tunnel_port_info *mt_port_info,
+                           bool add_port)
+{
+    struct mirror_tunnel_port_info *ptr = mirror_tunnel_head;
+
+    if (add_port) {
+        if (!ptr) {
+            mirror_tunnel_head = mt_port_info;
+            return;
+        }
+        while (ptr->next) {
+            ptr = ptr->next;
+        }
+        ptr->next = mt_port_info;
+    } else {
+        while (ptr->next &&
+            ptr->next->port_id != mt_port_info->port_id) {
+            ptr = ptr->next;
+        }
+
+        if (ptr->next) {
+            ptr->next = ptr->next->next;
+            free(mt_port_info);
+        } else {
+            if (ptr->port_id == mt_port_info->port_id) {
+                mirror_tunnel_head = NULL;
+                free(mt_port_info);
+            } else {
+                VLOG_ERR("Fail to find %s mirror port (%d) info\n",
+                 add_port?"add":"remove", mt_port_info->port_id);
+            }
+        }
+    }
+}
+
+static struct mirror_tunnel_port_info*
+netdev_dpdk_get_mt_port_info(uint16_t port_id)
+{
+    struct mirror_tunnel_port_info *mt_port_info;
+
+    if (mirror_tunnel_head) {
+        mt_port_info = mirror_tunnel_head;
+        while (mt_port_info) {
+            if (mt_port_info->port_id == port_id) {
+                return mt_port_info;
+            }
+            mt_port_info = mt_port_info->next;
+        }
+        VLOG_ERR("Could not tunnel port with port-id %d\n",
+            port_id);
+    }
+
+    mt_port_info = xmalloc(sizeof(struct mirror_tunnel_port_info));
+    memset(mt_port_info, 0, sizeof(*mt_port_info));
+    mt_port_info->port_id = port_id;
+    mt_port_info->next = NULL;
+
+    return mt_port_info;
+}
+
+static int
+netdev_dpdk_addr_to_portid(const char *pci_addr_str, uint16_t *port_id)
+{
+    struct rte_pci_device *pci_dev;
+    struct rte_pci_addr pci_addr;
+    int i;
+
+    if (rte_pci_addr_parse(pci_addr_str, &pci_addr)) {
+        VLOG_ERR("Incorrect pci address %s\n", pci_addr_str);
+        return -1;
+    }
+
+    for (i = 0; i < RTE_MAX_ETHPORTS; i++) {
+        struct rte_pci_addr *eth_pci_addr;
+
+        if (!rte_eth_devices[i].device) {
+            continue;
+        }
+
+        pci_dev = RTE_ETH_DEV_TO_PCI(&rte_eth_devices[i]);
+        if (!pci_dev) {
+            continue;
+        }
+
+        eth_pci_addr = &pci_dev->addr;
+
+        if (pci_addr.bus == eth_pci_addr->bus &&
+            pci_addr.devid == eth_pci_addr->devid &&
+            pci_addr.domain == eth_pci_addr->domain &&
+            pci_addr.function == eth_pci_addr->function) {
+            *port_id = i;
+
+            return 0;
+        }
+    }
+
+    return -1;
+}
+
+static int
+netdev_dpdk_mt_open(uint16_t port_id, struct mirror_param *param)
+{
+    struct rte_eth_dev_info dev_info;
+    struct rte_eth_txconf txq_conf;
+    struct rte_eth_rxconf rxq_conf;
+    struct rte_mempool *pktbuf;
+
+    struct mirror_tunnel_port_info *mt_info;
+
+    uint16_t nb_rxd = NIC_PORT_DEFAULT_RXQ_SIZE;
+    uint16_t nb_txd = NIC_PORT_DEFAULT_TXQ_SIZE;
+    unsigned int i, num_queue;
+
+    struct rte_eth_conf mt_port_conf = {
+        .rxmode = {
+            .split_hdr_size = 0,
+        },
+        .txmode = {
+            .mq_mode = ETH_MQ_TX_NONE,
+        },
+    };
+
+    mt_info = netdev_dpdk_get_mt_port_info(port_id);
+    if (!mt_info) {
+        return -1;
+    }
+
+    if (mt_info->port_started) {
+        param->n_dst_queue = mt_info->num_queue;
+        param->dst_port_id = port_id;
+        param->locks = mt_info->locks;
+        mt_info->share_count++;
+
+        return 0;
+    }
+
+    rte_eth_dev_info_get(port_id, &dev_info);
+    num_queue = param->n_src_queue;
+
+    /* A tunnel device doesn't require mbuf. It's used as
+     * hardware channel, transmit packets with
+     * mbuf provided by source. Need this mbuf creation
+     * to finish port initialization
+     */
+    pktbuf = rte_pktmbuf_pool_create(
+            "tunnel-port",
+            (dev_info.rx_desc_lim.nb_max + dev_info.tx_desc_lim.nb_max),
+            RTE_MEMPOOL_CACHE_MAX_SIZE, 0, RTE_MBUF_DEFAULT_BUF_SIZE,
+            rte_eth_dev_socket_id(port_id));
+
+    mt_port_conf.txmode.offloads |= DEV_TX_OFFLOAD_VLAN_INSERT;
+    if (dev_info.tx_offload_capa & DEV_TX_OFFLOAD_MBUF_FAST_FREE) {
+        mt_port_conf.txmode.offloads |= DEV_TX_OFFLOAD_MBUF_FAST_FREE;
+    }
+    rte_eth_dev_configure(port_id, 1, num_queue, &mt_port_conf);
+
+    /* init one Rx queue */
+    rxq_conf = dev_info.default_rxconf;
+    rxq_conf.offloads = mt_port_conf.rxmode.offloads;
+    if (rte_eth_rx_queue_setup(port_id, 0, nb_rxd,
+        rte_eth_dev_socket_id(port_id), &rxq_conf, pktbuf) < 0)
+        VLOG_ERR("fail to setup tunnel port (%d) rx-queue\n", port_id);
+
+    /* init # of Tx queue as part of mirror-tunnel setup */
+    txq_conf = dev_info.default_txconf;
+    txq_conf.offloads |= mt_port_conf.txmode.offloads;
+    for (i = 0; i < num_queue; i++) {
+        if (rte_eth_tx_queue_setup(port_id,
+            i, nb_txd,
+            rte_eth_dev_socket_id(port_id),
+            &txq_conf) < 0) {
+            VLOG_ERR("fail to setup tunnel port (%d) tx queue #%u\n",
+                port_id, i);
+            return -1;
+        }
+    }
+
+    if (rte_eth_dev_start(port_id) < 0) {
+        VLOG_ERR("fail to start tunnel port %d\n", port_id);
+        return -1;
+    }
+
+    mt_info->locks = xmalloc(num_queue * sizeof(rte_spinlock_t));
+    if (mt_info->locks) {
+        for (i = 0; i < mt_info->num_queue; i++) {
+            rte_spinlock_init(&mt_info->locks[i]);
+        }
+    } else {
+        return -1;
+    }
+    mt_info->share_count = 1;
+    mt_info->port_started = true;
+    mt_info->num_queue = num_queue;
+
+    param->n_dst_queue = mt_info->num_queue;
+    param->dst_port_id = port_id;
+    param->locks = mt_info->locks;
+
+    netdev_dpdk_update_mt_list(mt_info, true);
+    return 0;
+}
+
+static void
+netdev_dpdk_mt_close(uint16_t mirror_port_id)
+{
+    struct mirror_tunnel_port_info *mt_port_info =
+        netdev_dpdk_get_mt_port_info(mirror_port_id);
+
+    if (mt_port_info) {
+        mt_port_info->share_count--;
+        if (!mt_port_info->share_count) {
+            netdev_dpdk_update_mt_list(mt_port_info, false);
+            rte_eth_dev_stop(mirror_port_id);
+            rte_eth_dev_close(mirror_port_id);
+        }
+    }
+}
+
+/* vhost device mirror registration and un-registration routines */
+static int
+netdev_vhost_register_mirror(struct netdev_dpdk *dev,
+    struct mirror_param *param, int tx_cb)
+{
+    uint32_t vid = netdev_dpdk_get_vid(dev);
+    struct mirror_offload_port *port_info = NULL;
+    struct mirror_param *data;
+
+    netdev_mirror_data_proc(vid, mirror_data_add, tx_cb, param, &port_info);
+    if (!port_info) {
+        return -1;
+    }
+
+    data = tx_cb ? &port_info->tx : &port_info->rx;
+    netdev_mirror_cb_set(data, (uint16_t) vid, 0, tx_cb);
+
+    if (tx_cb) {
+        dev->tx_cb = data->mirror_cb;
+    } else {
+        dev->rx_cb = data->mirror_cb;
+    }
+
+    return 0;
+}
+
+static int
+netdev_vhost_unregister_mirror(struct netdev_dpdk *dev, int tx_cb)
+{
+    /* release both cb and pkt_buf */
+    unsigned int i;
+    uint32_t vid = netdev_dpdk_get_vid(dev);
+    struct mirror_offload_port *port_info = NULL;
+    struct mirror_param *data;
+
+    netdev_mirror_data_proc(vid, mirror_data_find, tx_cb, NULL, &port_info);
+    if (port_info == NULL) {
+        VLOG_ERR("Source port %d is not on outstanding port mirror db\n", vid);
+        return -1;
+    }
+    data = tx_cb ? &port_info->tx : &port_info->rx;
+
+    if (tx_cb) {
+        dev->tx_cb = NULL;
+    } else {
+        dev->rx_cb = NULL;
+    }
+
+    for (i = 0; i < data->n_src_queue; i++) {
+        free(data->mirror_cb[i].direct);
+    }
+
+    free(data->mirror_cb);
+
+    if (data->pkt_buf) {
+        free(data->pkt_buf);
+        data->pkt_buf = NULL;
+    }
+
+    if (data->extra_data) {
+        free(data->extra_data);
+        data->extra_data = NULL;
+        data->extra_data_size = 0;
+    }
+
+    netdev_mirror_data_proc(vid,  mirror_data_rem, tx_cb, NULL, NULL);
+    return 0;
+}
+
+static int
+netdev_dpdk_mirror_offload(struct netdev *src, struct eth_addr *flow_addr,
+                           uint16_t vlan_id, char *mirror_tunnel_addr,
+                           bool add_mirror, bool tx_cb) {
+    struct netdev_dpdk *src_dev = netdev_dpdk_cast(src);
+    bool eth_dev = src_dev->type == DPDK_DEV_ETH;
+    uint16_t mirror_port_id;
+    int status = 0;
+
+    if (netdev_dpdk_addr_to_portid(mirror_tunnel_addr, &mirror_port_id)) {
+        VLOG_ERR("Could not find tunnel port with BDF addr %s\n",
+            mirror_tunnel_addr);
+        return -1;
+    }
+
+    if (add_mirror) {
+        uint32_t i;
+        struct mirror_param data;
+        uint64_t mac_addr = 0;
+
+        memset(&data, 0, sizeof(struct mirror_param));
+        data.extra_data_size = 0;
+        data.extra_data = NULL;
+        data.mirror_type = mirror_port;
+        for (i = 0; i < 6; i++) {
+            mac_addr <<= 8;
+            mac_addr |= flow_addr->ea[6 - i - 1];
+        }
+        if (mac_addr) {
+            data.mirror_type = mirror_flow_mac;
+            data.extra_data_size = sizeof(uint64_t);
+            data.extra_data = xmalloc(sizeof(uint64_t));
+            memcpy(data.extra_data, &mac_addr, sizeof(uint64_t));
+        }
+        data.dst_vlan_id = vlan_id;
+        data.n_src_queue = tx_cb?src->n_txq:src->n_rxq;
+        data.max_burst_size = NETDEV_MAX_BURST;
+
+        if (netdev_dpdk_mt_open(mirror_port_id, &data)) {
+            VLOG_ERR("Fail to initialize mirror tunnel port %d\n",
+                mirror_port_id);
+            return -1;
+        }
+
+        VLOG_INFO("register %s device with %s mirror-offload with"
+            "src-port:%d (%s) and output-port:%d (%s) vlan-id=%d flow-mac="
+            "0x%" PRIx64 "\n",
+            eth_dev?"ethdev":"vhost",
+            tx_cb?"ingress":"egress", src_dev->port_id,
+            src->name, mirror_port_id, mirror_tunnel_addr, vlan_id,
+            (uint64_t)__builtin_bswap64(mac_addr));
+
+        if (eth_dev) {
+            status = netdev_eth_register_mirror(src_dev->port_id, &data,
+                tx_cb);
+        } else {
+            status = netdev_vhost_register_mirror(src_dev, &data, tx_cb);
+        }
+    } else {
+        VLOG_INFO("unregister %s device with %s mirror-offload with"
+            " src-port:%d(%s)\n",
+            eth_dev?"ethdev":"vhost",
+            tx_cb?"ingress":"egress", src_dev->port_id,
+            src->name);
+
+        if (eth_dev) {
+            status = netdev_eth_unregister_mirror(src_dev->port_id, tx_cb);
+        } else {
+            status = netdev_vhost_unregister_mirror(src_dev, tx_cb);
+        }
+
+        netdev_dpdk_mt_close(mirror_port_id);
+    }
+
+    return status;
+}
+
 #define NETDEV_DPDK_CLASS_COMMON                            \
     .is_pmd = true,                                         \
     .alloc = netdev_dpdk_alloc,                             \
@@ -5340,6 +5734,7 @@ static const struct netdev_class dpdk_class = {
     .construct = netdev_dpdk_construct,
     .set_config = netdev_dpdk_set_config,
     .send = netdev_dpdk_eth_send,
+    .mirror_offload = netdev_dpdk_mirror_offload,
 };
 
 static const struct netdev_class dpdk_vhost_class = {
@@ -5355,6 +5750,7 @@ static const struct netdev_class dpdk_vhost_class = {
     .reconfigure = netdev_dpdk_vhost_reconfigure,
     .rxq_recv = netdev_dpdk_vhost_rxq_recv,
     .rxq_enabled = netdev_dpdk_vhost_rxq_enabled,
+    .mirror_offload = netdev_dpdk_mirror_offload,
 };
 
 static const struct netdev_class dpdk_vhost_client_class = {
@@ -5371,6 +5767,7 @@ static const struct netdev_class dpdk_vhost_client_class = {
     .reconfigure = netdev_dpdk_vhost_client_reconfigure,
     .rxq_recv = netdev_dpdk_vhost_rxq_recv,
     .rxq_enabled = netdev_dpdk_vhost_rxq_enabled,
+    .mirror_offload = netdev_dpdk_mirror_offload,
 };
 
 void
diff --git a/lib/netdev-provider.h b/lib/netdev-provider.h
index 73dce2fca..dab278dcd 100644
--- a/lib/netdev-provider.h
+++ b/lib/netdev-provider.h
@@ -834,6 +834,22 @@ struct netdev_class {
     /* Get a block_id from the netdev.
      * Returns the block_id or 0 if none exists for netdev. */
     uint32_t (*get_block_id)(struct netdev *);
+
+    /* Configure a mirror offload setting on a netdev.
+     * 'src': netdev traffic to be mirrored
+     * 'flow_addr': the destination mac address is of source traffic for
+     *  inspection.
+     * 'dst': netdev where mirror traffic is transmitted.
+     * 'vlan_id': vlag to be added to the mirrored packets.
+     * 'mt_pci_addr': mirror tunnel pcie address.
+     * 'add_mirror': true: configure a mirror traffic; false: remove mirror
+     * 'ingress': true: mirror 'src' netdev Rx traffic; false: mirror
+     *  'src' netdev Tx traffic.
+     */
+    int (*mirror_offload)(struct netdev *src, struct eth_addr *flow_addr,
+                          uint16_t vlan_id, char *mt_pci_addr,
+                          bool add_mirror, bool ingress);
+
 };
 
 int netdev_register_provider(const struct netdev_class *);
diff --git a/lib/netdev.c b/lib/netdev.c
index 91e91955c..464c2f8fe 100644
--- a/lib/netdev.c
+++ b/lib/netdev.c
@@ -69,6 +69,8 @@ COVERAGE_DEFINE(netdev_get_stats);
 COVERAGE_DEFINE(netdev_send_prepare_drops);
 COVERAGE_DEFINE(netdev_push_header_drops);
 
+#define MIRROR_DB_INIT_SIZE 8
+
 struct netdev_saved_flags {
     struct netdev *netdev;
     struct ovs_list node;           /* In struct netdev's saved_flags_list. */
@@ -2297,3 +2299,387 @@ netdev_free_custom_stats_counters(struct netdev_custom_stats *custom_stats)
         }
     }
 }
+
+
+struct netdev_mirror_offload_item {
+    struct mirror_offload_info info;
+
+    struct ovs_list node;
+};
+
+struct netdev_mirror_offload {
+    struct ovs_mutex mutex;
+    struct ovs_list list;
+    pthread_cond_t cond;
+};
+
+static struct netdev_mirror_offload netdev_mirror_offload = {
+    .mutex = OVS_MUTEX_INITIALIZER,
+    .list  = OVS_LIST_INITIALIZER(&netdev_mirror_offload.list),
+};
+
+static struct ovsthread_once offload_thread_once
+    = OVSTHREAD_ONCE_INITIALIZER;
+
+static void *netdev_mirror_offload_main(void *data);
+
+/*
+ * Re-size mirror_db when it's out of space.
+ * Always double the buffer when it's needed
+ */
+static int
+netdev_mirror_db_resize(struct netdev_mirror_offload_item ***old_db,
+    int *old_db_size)
+{
+    struct netdev_mirror_offload_item **new_db;
+    int cur_size = *old_db_size;
+    int new_size;
+
+    if (!cur_size) {
+        new_size = MIRROR_DB_INIT_SIZE;
+    } else {
+        new_size = 2 * cur_size;
+    }
+
+    new_db = xzalloc(sizeof(struct netdev_mirror_offload_item *) * new_size);
+
+    if (!new_db) {
+        VLOG_ERR("Out of memory!!!");
+        return -1;
+    }
+    memset(new_db, 0, sizeof(struct netdev_mirror_offload_item *) * new_size);
+
+    if (cur_size) {
+        int i;
+
+        for (i = 0; i < cur_size; i++) {
+            new_db[i] = (*old_db)[i];
+        }
+        free(*old_db);
+    }
+
+    *old_db = new_db;
+    *old_db_size = new_size;
+
+    return 0;
+}
+
+static void
+netdev_free_mirror_offload(struct netdev_mirror_offload_item *offload)
+{
+    if (!offload) {
+        return;
+    }
+
+    if (offload->info.src) {
+        free(offload->info.src);
+    }
+    if (offload->info.dst) {
+        free(offload->info.dst);
+    }
+    if (offload->info.flow_dst_mac) {
+        free(offload->info.flow_dst_mac);
+    }
+    if (offload->info.flow_src_mac) {
+        free(offload->info.flow_src_mac);
+    }
+    if (offload->info.output_src_tags) {
+        free(offload->info.output_src_tags);
+    }
+    if (offload->info.output_dst_tags) {
+        free(offload->info.output_dst_tags);
+    }
+    if (offload->info.name) {
+        free(offload->info.name);
+    }
+    if (offload->info.mirror_tunnel_addr) {
+        free(offload->info.mirror_tunnel_addr);
+    }
+
+    free(offload);
+}
+
+static struct
+netdev_mirror_offload_item *
+netdev_alloc_mirror_offload(struct mirror_offload_info *info)
+{
+    struct netdev_mirror_offload_item *offload;
+    int i;
+
+    offload = xzalloc(sizeof(*offload));
+    memcpy(&offload->info, info, sizeof(struct mirror_offload_info));
+
+    if (info->name) {
+        offload->info.name = xzalloc(strlen(info->name) + 1);
+        if (offload->info.name) {
+            ovs_strzcpy(offload->info.name, info->name, strlen(info->name));
+        }
+    }
+
+    if (info->mirror_tunnel_addr) {
+        offload->info.mirror_tunnel_addr =
+            xzalloc(strlen(info->mirror_tunnel_addr) + 1);
+        if (offload->info.mirror_tunnel_addr) {
+            ovs_strzcpy(offload->info.mirror_tunnel_addr,
+                        info->mirror_tunnel_addr,
+                        strlen(info->mirror_tunnel_addr));
+        }
+    }
+
+    /* only add_mirror request include valid configuration */
+    if (info->n_src_port) {
+        offload->info.src = xzalloc(sizeof(struct netdev *)*info->n_src_port);
+        offload->info.flow_dst_mac = xzalloc(sizeof(struct eth_addr)*
+            info->n_src_port);
+        offload->info.output_src_tags = xzalloc(sizeof(uint16_t)*
+            info->n_src_port);
+        if (!offload->info.src || !offload->info.flow_dst_mac ||
+            !offload->info.output_src_tags) {
+            VLOG_ERR("Out of memory!!!");
+            netdev_free_mirror_offload(offload);
+            return NULL;
+        }
+
+        for (i = 0; i < info->n_src_port; i++) {
+            offload->info.src[i] = info->src[i];
+            offload->info.output_src_tags[i] = info->output_src_tags[i];
+            memcpy(&offload->info.flow_dst_mac[i], &info->flow_dst_mac[i],
+                sizeof(struct eth_addr));
+        }
+    }
+
+    if (info->n_dst_port) {
+        offload->info.dst = xzalloc(sizeof(struct netdev *)*info->n_dst_port);
+        offload->info.flow_src_mac = xzalloc(sizeof(struct eth_addr)*
+            info->n_dst_port);
+        offload->info.output_dst_tags = xzalloc(sizeof(uint16_t)*
+            info->n_dst_port);
+        if (!offload->info.dst || !offload->info.flow_src_mac ||
+            !offload->info.output_dst_tags) {
+            VLOG_ERR("Out of memory!!!");
+            netdev_free_mirror_offload(offload);
+            return NULL;
+        }
+
+        for (i = 0; i < info->n_dst_port; i++) {
+            offload->info.dst[i] = info->dst[i];
+            offload->info.output_dst_tags[i] = info->output_dst_tags[i];
+            memcpy(&offload->info.flow_src_mac[i], &info->flow_src_mac[i],
+                sizeof(struct eth_addr));
+        }
+    }
+
+    return offload;
+}
+
+static void
+netdev_append_mirror_offload(struct netdev_mirror_offload_item *offload)
+{
+    ovs_mutex_lock(&netdev_mirror_offload.mutex);
+    ovs_list_push_back(&netdev_mirror_offload.list, &offload->node);
+    xpthread_cond_signal(&netdev_mirror_offload.cond);
+    ovs_mutex_unlock(&netdev_mirror_offload.mutex);
+}
+
+void
+netdev_mirror_offload_put(struct mirror_offload_info *info)
+{
+    struct netdev_mirror_offload_item *offload;
+    /* only support tunnel port for traffic mirroring */
+    if (info->add_mirror && !info->mirror_tunnel_addr) {
+        return;
+    }
+
+    if (ovsthread_once_start(&offload_thread_once)) {
+        xpthread_cond_init(&netdev_mirror_offload.cond, NULL);
+        ovs_thread_create("netdev_mirror_offload",
+                          netdev_mirror_offload_main, NULL);
+        ovsthread_once_done(&offload_thread_once);
+    }
+
+    offload = netdev_alloc_mirror_offload(info);
+    netdev_append_mirror_offload(offload);
+}
+
+static int
+netdev_mirror_offload_configue(struct mirror_offload_info *info,
+    bool add_mirror)
+{
+    int un_support_count = 0;
+    int ret;
+
+    if (info->n_src_port) {
+        for (int i = 0; i < info->n_src_port; i++) {
+            const struct netdev_class *class =
+                info->src[i]->netdev_class;
+            if (!class) {
+                return -1;
+            }
+            if (class->mirror_offload) {
+                ret = class->mirror_offload(
+                    info->src[i],
+                    &info->flow_dst_mac[i],
+                    info->output_src_tags[i],
+                    info->mirror_tunnel_addr,
+                    add_mirror, false);
+                if (ret) {
+                    VLOG_ERR("Fail to %s mirror-offload"
+                        " configuration %s\n",
+                        add_mirror ? "add" : "remove",
+                        info->name);
+                    return ret;
+                }
+            } else {
+                un_support_count++;
+            }
+        }
+    }
+
+    if (info->n_dst_port) {
+        for (int i = 0; i < info->n_dst_port; i++) {
+            const struct netdev_class *class =
+                info->dst[i]->netdev_class;
+            if (!class) {
+                return -1;
+            }
+            if (class->mirror_offload) {
+                ret = class->mirror_offload(
+                    info->dst[i],
+                    &info->flow_src_mac[i],
+                    info->output_dst_tags[i],
+                    info->mirror_tunnel_addr,
+                    add_mirror, true);
+                if (ret) {
+                    VLOG_ERR("Fail to %s mirror-offload"
+                        " configuration %s\n",
+                        add_mirror ? "add" : "remove",
+                        info->name);
+                    return ret;
+                }
+            } else {
+                un_support_count++;
+            }
+        }
+    }
+
+    return un_support_count;
+}
+
+static void *
+netdev_mirror_offload_main(void *data OVS_UNUSED)
+{
+    struct netdev_mirror_offload_item *offload;
+    struct mirror_offload_info *info;
+    struct ovs_list *list;
+    struct netdev_mirror_offload_item **offload_db = NULL;
+    int offload_used_count = 0;
+    int offload_db_size = 0;
+    int ret, i, ind;
+
+    /* continue polling to check if there is an outstanding request */
+    for (;;) {
+        ovs_mutex_lock(&netdev_mirror_offload.mutex);
+        if (ovs_list_is_empty(&netdev_mirror_offload.list)) {
+            ovsrcu_quiesce_start();
+            ovs_mutex_cond_wait(&netdev_mirror_offload.cond,
+                                &netdev_mirror_offload.mutex);
+            ovsrcu_quiesce_end();
+        }
+        list = ovs_list_pop_front(&netdev_mirror_offload.list);
+        offload = CONTAINER_OF(list, struct netdev_mirror_offload_item,
+            node);
+        ovs_mutex_unlock(&netdev_mirror_offload.mutex);
+
+        if (!offload_db_size &&
+            netdev_mirror_db_resize(&offload_db, &offload_db_size)){
+            return NULL;
+        }
+
+        ind = offload_db_size;
+        for (i = 0; i < offload_db_size; i++) {
+            if (offload_db[i] &&
+                !strncmp(offload_db[i]->info.name, offload->info.name,
+                strlen(offload->info.name) + 1)) {
+                ind = i;
+                break;
+            }
+        }
+
+        if (!offload->info.add_mirror) {
+            /* remove mirror offload setup */
+            if (ind == offload_db_size) {
+                VLOG_WARN("Mirror offload remove configuration, %s, "
+                    "not found; clear mirror offload operation"
+                    " aborted\n", offload->info.name);
+                continue;
+            }
+        } else {
+            /* add mirror offload */
+            if (ind < offload_db_size) {
+                netdev_free_mirror_offload(offload);
+                VLOG_WARN("Attempt adding an existing mirror-offload "
+                    "configuration; request aborted\n");
+                continue;
+            }
+
+            if (offload_used_count == offload_db_size &&
+                netdev_mirror_db_resize(&offload_db, &offload_db_size)) {
+                return NULL;
+            }
+        }
+
+        info = offload->info.add_mirror ? &offload->info :
+            &offload_db[ind]->info;
+        ret = netdev_mirror_offload_configue(info, offload->info.add_mirror);
+
+        if (ret) {
+            VLOG_ERR("%s mirror configuration fails due to %s\n",
+                offload->info.add_mirror ? "Add" : "Remove",
+                ret > 0 ? "unsupport source traffic type" :
+                "device is not ready");
+            netdev_free_mirror_offload(offload);
+            continue;
+        } else {
+            VLOG_INFO("Succeed %s mirror-offload configuration: %s",
+                offload->info.add_mirror ? "adding" : "removing",
+                offload->info.name);
+        }
+
+        if (offload->info.add_mirror) {
+            for (i = 0; i < offload_db_size; i++) {
+                if (offload_db[i] == NULL) {
+                    offload_db[i] = offload;
+                    offload_used_count++;
+                    break;
+                }
+            }
+        } else {
+            /* remove the prior "add" request */
+            netdev_free_mirror_offload(offload_db[ind]);
+            offload_db[ind] = NULL;
+
+            /* remove the current("remove") request */
+            netdev_free_mirror_offload(offload);
+            offload_used_count--;
+        }
+
+        /* free db when the used count drop to 0 */
+        if (!offload_used_count) {
+            free(offload_db);
+            offload_db = NULL;
+            offload_db_size = 0;
+        }
+    }
+
+    /* clean up memory */
+    for (i = 0; i < offload_db_size; i++) {
+        if (offload_db[i]) {
+            netdev_free_mirror_offload(offload_db[i]);
+        }
+    }
+    if (offload_db) {
+        free(offload_db);
+    }
+
+    return NULL;
+}
diff --git a/lib/netdev.h b/lib/netdev.h
index b705a9e56..cce042fc7 100644
--- a/lib/netdev.h
+++ b/lib/netdev.h
@@ -201,6 +201,22 @@ int netdev_send(struct netdev *, int qid, struct dp_packet_batch *,
                 bool concurrent_txq);
 void netdev_send_wait(struct netdev *, int qid);
 
+/* Hardware assisted mirror offloading*/
+struct mirror_offload_info {
+    struct netdev **src;
+    struct netdev **dst;
+    int n_src_port;
+    int n_dst_port;
+    struct eth_addr *flow_src_mac;
+    struct eth_addr *flow_dst_mac;
+    uint16_t *output_src_tags;
+    uint16_t *output_dst_tags;
+    bool add_mirror;
+    char *mirror_tunnel_addr;
+    char *name;
+};
+void netdev_mirror_offload_put(struct mirror_offload_info *);
+
 /* native tunnel APIs */
 /* Structure to pass parameters required to build a tunnel header. */
 struct netdev_tnl_build_header_params {
diff --git a/tests/ovs-vsctl.at b/tests/ovs-vsctl.at
index dccb11741..ff6e9e625 100644
--- a/tests/ovs-vsctl.at
+++ b/tests/ovs-vsctl.at
@@ -1364,7 +1364,9 @@ _uuid               : <1>
 name                : eth1
 _uuid               : <2>
 name                : mymirror
+output_dst_vlan     : []
 output_port         : <1>
+output_src_vlan     : []
 output_vlan         : []
 select_all          : false
 select_dst_port     : [<0>]
diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c
index 5ed7e8234..7b7603513 100644
--- a/vswitchd/bridge.c
+++ b/vswitchd/bridge.c
@@ -38,6 +38,7 @@
 #include "mac-learning.h"
 #include "mcast-snooping.h"
 #include "netdev.h"
+#include "netdev-provider.h"
 #include "netdev-offload.h"
 #include "nx-match.h"
 #include "ofproto/bond.h"
@@ -330,6 +331,9 @@ static void mirror_destroy(struct mirror *);
 static bool mirror_configure(struct mirror *);
 static void mirror_refresh_stats(struct mirror *);
 
+static void mirror_offload_destroy(struct mirror *);
+static bool mirror_offload_configure(struct mirror *);
+
 static void iface_configure_lacp(struct iface *,
                                  struct lacp_member_settings *);
 static bool iface_create(struct bridge *, const struct ovsrec_interface *,
@@ -423,6 +427,35 @@ if_notifier_changed(struct if_notifier *notifier OVS_UNUSED)
     seq_wait(ifaces_changed, last_ifaces_changed);
     return changed;
 }
+
+static struct port *
+port_lookup_all(const char *port_name)
+{
+    struct bridge *br;
+    struct port *port = NULL;
+    int found = 0;
+
+    HMAP_FOR_EACH (br, node, &all_bridges) {
+        struct port *temp_port = NULL;
+        temp_port = port_lookup(br, port_name);
+        if (temp_port) {
+            if (!port) {
+                port = temp_port;
+            }
+            found++;
+        }
+    }
+
+    if (found) {
+        if (found > 1) {
+            VLOG_INFO("More than one bridge owns port with name:%s\n",
+                port_name);
+        }
+        return port;
+    }
+    return NULL;
+}
+
 
 /* Public functions. */
 
@@ -5055,14 +5088,228 @@ mirror_create(struct bridge *br, const struct ovsrec_mirror *cfg)
     return m;
 }
 
+static struct netdev *get_netdev_from_port(struct mirror *m,
+    struct port **port, const char *name)
+{
+    struct port *temp_port;
+    struct iface *iface;
+
+    *port = NULL;
+    temp_port = port_lookup(m->bridge, name);
+    if (temp_port) {
+        LIST_FOR_EACH (iface, port_elem, &temp_port->ifaces) {
+            if (iface) {
+                *port = temp_port;
+                return iface->netdev;
+            }
+        }
+    }
+    /* try different bridges */
+    temp_port = port_lookup_all(name);
+    if (temp_port) {
+        LIST_FOR_EACH (iface, port_elem, &temp_port->ifaces) {
+            if (iface) {
+                *port = temp_port;
+                return iface->netdev;
+            }
+        }
+    }
+    return NULL;
+}
+
+static void
+release_mirror_offload_info(struct mirror_offload_info *info)
+{
+    if (info->src) {
+        free(info->src);
+    }
+    if (info->dst) {
+        free(info->dst);
+    }
+    if (info->flow_dst_mac) {
+        free(info->flow_dst_mac);
+    }
+    if (info->flow_src_mac) {
+        free(info->flow_src_mac);
+    }
+    if (info->output_src_tags) {
+        free(info->output_src_tags);
+    }
+    if (info->output_dst_tags) {
+        free(info->output_dst_tags);
+    }
+    if (info->name) {
+        free(info->name);
+    }
+    if (info->mirror_tunnel_addr) {
+        free(info->mirror_tunnel_addr);
+    }
+}
+
+static int
+set_mirror_offload_info(struct mirror *m, struct mirror_offload_info *info)
+{
+    const struct ovsrec_mirror *cfg = m->cfg;
+    struct port *port = NULL;
+    int i;
+
+    if (m->name) {
+        info->name = xmalloc(strlen(m->name) + 1);
+        ovs_strzcpy(info->name, m->name, strlen(m->name));
+    }
+
+    if (cfg->mirror_tunnel_addr) {
+        info->mirror_tunnel_addr = xmalloc(strlen(cfg->mirror_tunnel_addr)
+            + 1);
+        ovs_strzcpy(info->mirror_tunnel_addr, cfg->mirror_tunnel_addr,
+                    strlen(cfg->mirror_tunnel_addr));
+    } else {
+        VLOG_ERR("mirror-offload configuration fails because"
+            " lack of tunnel device\n");
+        return -1;
+    }
+
+    /* source port */
+    info->n_src_port = cfg->n_select_src_port;
+    if (info->n_src_port) {
+        info->src = xmalloc(sizeof(struct netdev *)*info->n_src_port);
+        info->flow_dst_mac = xmalloc(sizeof(struct eth_addr)*
+            info->n_src_port);
+        if (info->n_src_port != cfg->n_output_src_vlan) {
+            VLOG_ERR("src port count:%d ouput src vlan count:%lu",
+                info->n_src_port, (unsigned long) cfg->n_output_src_vlan);
+            return -1;
+        }
+        info->output_src_tags = xmalloc(sizeof(uint16_t)*info->n_src_port);
+    }
+
+    if (info->n_src_port) {
+        /* find netdev instance for each port */
+        for (i = 0; i < info->n_src_port; i++) {
+            info->src[i] = get_netdev_from_port(m, &port,
+                cfg->select_src_port[i]->name);
+            if (!info->src[i]) {
+                VLOG_ERR("src-port: %s is not a netdev device\n",
+                    cfg->select_src_port[i]->name);
+                return -1;
+            }
+        }
+        memset(info->flow_dst_mac, 0, sizeof(struct eth_addr)*
+            info->n_src_port);
+
+        /*
+         * for source port, flow is separated by
+         * different dst mac addr
+         */
+        if (cfg->n_flow_dst_mac) {
+            int dst_count = (info->n_src_port > cfg->n_flow_dst_mac)?
+                cfg->n_flow_dst_mac:info->n_src_port;
+            for (i = 0; i < dst_count; i++) {
+                eth_addr_from_string(cfg->flow_dst_mac[i],
+                    &info->flow_dst_mac[i]);
+            }
+        }
+
+        if (cfg->n_output_src_vlan) {
+            int count = (cfg->n_output_src_vlan > info->n_src_port)?
+                info->n_src_port:cfg->n_output_src_vlan;
+            for (i = 0; i < count; i++) {
+                info->output_src_tags[i] = cfg->output_src_vlan[i] & 0xFFF;
+            }
+        }
+    }
+
+    /* dst ports */
+    info->n_dst_port = cfg->n_select_dst_port;
+    if (info->n_dst_port) {
+        info->dst = xmalloc(sizeof(struct netdev *)*info->n_dst_port);
+        info->flow_src_mac = xmalloc(sizeof(struct eth_addr)*
+            info->n_dst_port);
+        if (info->n_dst_port != cfg->n_output_dst_vlan) {
+            VLOG_ERR("dst port count:%d ouput dst vlan count:%lu\n",
+                info->n_dst_port, (unsigned long) cfg->n_output_dst_vlan);
+            return -1;
+        }
+        info->output_dst_tags = xmalloc(sizeof(uint16_t)*info->n_dst_port);
+    }
+
+    if (info->n_dst_port) {
+        for (i = 0; i < info->n_dst_port; i++) {
+            info->dst[i] = get_netdev_from_port(m, &port,
+                cfg->select_dst_port[i]->name);
+            if (!info->dst[i]) {
+                VLOG_ERR("dst-port: %s is not a netdev device\n",
+                    cfg->select_dst_port[i]->name);
+                return -1;
+            }
+        }
+        memset(info->flow_src_mac, 0, sizeof(struct eth_addr)*
+            info->n_dst_port);
+
+        /*
+         * for destination port, flow is separated by
+         * different src mac addr
+         */
+        if (cfg->n_flow_src_mac) {
+            int src_count = (info->n_dst_port > cfg->n_flow_src_mac)?
+                cfg->n_flow_src_mac:info->n_dst_port;
+            for (i = 0; i < src_count; i++) {
+                eth_addr_from_string(cfg->flow_src_mac[i],
+                    &info->flow_src_mac[i]);
+            }
+        }
+
+        if (cfg->n_output_dst_vlan) {
+            int count = (cfg->n_output_dst_vlan > info->n_dst_port)?
+                info->n_dst_port:cfg->n_output_dst_vlan;
+            for (i = 0; i < count; i++) {
+                info->output_dst_tags[i] = cfg->output_dst_vlan[i] & 0xFFF;
+            }
+        }
+    }
+
+    VLOG_INFO("sucess creating mirror-offload(%s): with %d src-port"
+        " streams %d dst-port streams to tunnel %s\n",
+        cfg->name, info->n_src_port, info->n_dst_port,
+        info->mirror_tunnel_addr?info->mirror_tunnel_addr:"none");
+    return 0;
+}
+
+static void
+mirror_offload_destroy(struct mirror *m)
+{
+    struct mirror_offload_info info;
+
+    memset(&info, 0, sizeof(struct mirror_offload_info));
+    info.add_mirror = false;
+    if (m->name) {
+        info.name = xmalloc(strlen(m->name) + 1);
+        if (info.name) {
+            ovs_strzcpy(info.name, m->name, strlen(m->name));
+        }
+    }
+
+    netdev_mirror_offload_put(&info);
+    if (info.name) {
+        free(info.name);
+    }
+    if (info.mirror_tunnel_addr) {
+        free(info.mirror_tunnel_addr);
+    }
+}
+
 static void
 mirror_destroy(struct mirror *m)
 {
     if (m) {
         struct bridge *br = m->bridge;
 
-        if (br->ofproto) {
-            ofproto_mirror_unregister(br->ofproto, m);
+        if (m->cfg && m->cfg->mirror_offload) {
+            mirror_offload_destroy(m);
+        } else {
+            if (br->ofproto) {
+                ofproto_mirror_unregister(br->ofproto, m);
+            }
         }
 
         hmap_remove(&br->mirrors, &m->hmap_node);
@@ -5094,12 +5341,32 @@ mirror_collect_ports(struct mirror *m,
     *n_out_portsp = n_out_ports;
 }
 
+static bool
+mirror_offload_configure(struct mirror *m)
+{
+    struct mirror_offload_info info;
+
+    memset(&info, 0, sizeof(struct mirror_offload_info));
+    info.add_mirror = true;
+    if (set_mirror_offload_info(m, &info)) {
+        release_mirror_offload_info(&info);
+        return false;
+    }
+
+    netdev_mirror_offload_put(&info);
+    release_mirror_offload_info(&info);
+    return true;
+}
+
 static bool
 mirror_configure(struct mirror *m)
 {
     const struct ovsrec_mirror *cfg = m->cfg;
     struct ofproto_mirror_settings s;
 
+    if (cfg->mirror_offload) {
+        return mirror_offload_configure(m);
+    }
     /* Set name. */
     if (strcmp(cfg->name, m->name)) {
         free(m->name);
diff --git a/vswitchd/vswitch.ovsschema b/vswitchd/vswitch.ovsschema
index 0666c8c76..4a1a34a1f 100644
--- a/vswitchd/vswitch.ovsschema
+++ b/vswitchd/vswitch.ovsschema
@@ -1,6 +1,6 @@
 {"name": "Open_vSwitch",
- "version": "8.2.0",
- "cksum": "1076640191 26427",
+ "version": "8.2.1",
+ "cksum": "4051567316 27206",
  "tables": {
    "Open_vSwitch": {
      "columns": {
@@ -418,8 +418,18 @@
      "columns": {
        "name": {
          "type": "string"},
+       "mirror_tunnel_addr": {
+         "type": "string"},
        "select_all": {
          "type": "boolean"},
+       "mirror_offload": {
+         "type": "boolean"},
+       "flow_src_mac": {
+         "type": {"key": {"type": "string"},
+                  "min": 0, "max": "unlimited"}},
+       "flow_dst_mac": {
+         "type": {"key": {"type": "string"},
+                  "min": 0, "max": "unlimited"}},
        "select_src_port": {
          "type": {"key": {"type": "uuid",
                           "refTable": "Port",
@@ -440,6 +450,16 @@
                           "refTable": "Port",
                           "refType": "weak"},
                   "min": 0, "max": 1}},
+       "output_src_vlan": {
+         "type": {"key": {"type": "integer",
+                          "minInteger": 0,
+                          "maxInteger": 4294967295},
+                  "min": 0, "max": 4096}},
+       "output_dst_vlan": {
+         "type": {"key": {"type": "integer",
+                          "minInteger": 0,
+                          "maxInteger": 4294967295},
+                  "min": 0, "max": 4096}},
        "output_vlan": {
          "type": {"key": {"type": "integer",
                           "minInteger": 1,
diff --git a/vswitchd/vswitch.xml b/vswitchd/vswitch.xml
index 4597a215d..fd2049a7f 100644
--- a/vswitchd/vswitch.xml
+++ b/vswitchd/vswitch.xml
@@ -4869,11 +4869,35 @@ ovs-vsctl add-port br0 p0 -- set Interface p0 type=patch options:peer=p1 \
         selected VLANs.
       </p>
 
+      <column name="mirror_tunnel_addr">
+        BDF string of the tunnel device on which mirrored traffic will be
+        transmitted.
+      </column>
+
       <column name="select_all">
         If true, every packet arriving or departing on any port is
         selected for mirroring.
       </column>
 
+      <column name="mirror_offload">
+        If true, a hw-assisted port mirroring is configured instead
+        default mirroring.
+      </column>
+
+      <column name="flow_src_mac">
+        The source MAC address(es) for per-flow mirroring. Each MAC
+        address is separate by ','. This parametr is paired with
+        select_dst_port. A '0' MAC address indicates the requested mirror
+        is a per-port mirroring, otherwise it's a per-flow mirroring
+      </column>
+
+      <column name="flow_dst_mac">
+        The destination MAC address(es) for per-flow mirroring. Each MAC
+        address is separate by ','. This parametr is paired with
+        select_src_port. A '0' MAC address indicates the requested mirror
+        is a per-port mirroring, otherwise it's a per-flow mirroring
+      </column>
+
       <column name="select_dst_port">
         Ports on which departing packets are selected for mirroring.
       </column>
@@ -4955,6 +4979,32 @@ ovs-vsctl add-port br0 p0 -- set Interface p0 type=patch options:peer=p1 \
         </p>
       </column>
 
+      <column name="output_src_vlan">
+        <p>Output VLAN for selected source port packets, if nonempty.</p>
+        <p>
+          <em>Please note:</em> This is different than
+          <ref column="output-vlan"/> This vlan is used to add an additional
+          vlan tag on the mirror traffic, regardless it contains vlan or not.
+          The receive end could choose to filter out this additional vlan.
+          This option is provided so the mirrored traffic could maintain its
+          original vlan informaiton, and this mirror can be used to filter
+          out un-wanted traffic such as in <ref column="mirror_offload"/>.
+        </p>
+      </column>
+
+      <column name="output_dst_vlan">
+        <p>Output VLAN for selected destination port packets, if nonempty.</p>
+        <p>
+          <em>Please note:</em> This is different than
+          <ref column="output-vlan"/> This vlan is used to add an additional
+          vlan tag on the mirror traffic, regardless it contains vlan or not.
+          The receive end could choose to filter out this additional vlan.
+          This option is provided so the mirrored traffic could maintain its
+          original vlan informaiton, and this mirror cab be used to filter
+          out un-wanted traffic such as in <ref column="mirror_offload"/>.
+        </p>
+      </column>
+
       <column name="snaplen">
         <p>Maximum per-packet number of bytes to mirror.</p>
         <p>A mirrored packet with size larger than <ref column="snaplen"/>
-- 
2.25.1



More information about the dev mailing list