net/i40e: move testpmd commands
[dpdk.git] / app / test-pmd / csumonly.c
index 2133dfc..7df201e 100644 (file)
@@ -24,7 +24,6 @@
 #include <rte_eal.h>
 #include <rte_per_lcore.h>
 #include <rte_lcore.h>
-#include <rte_atomic.h>
 #include <rte_branch_prediction.h>
 #include <rte_mempool.h>
 #include <rte_mbuf.h>
 #include <rte_ip.h>
 #include <rte_tcp.h>
 #include <rte_udp.h>
+#include <rte_vxlan.h>
 #include <rte_sctp.h>
+#include <rte_gtp.h>
 #include <rte_prefetch.h>
 #include <rte_string_fns.h>
 #include <rte_flow.h>
+#ifdef RTE_LIB_GRO
 #include <rte_gro.h>
+#endif
+#ifdef RTE_LIB_GSO
 #include <rte_gso.h>
+#endif
+#include <rte_geneve.h>
 
 #include "testpmd.h"
 
 #define IP_DEFTTL  64   /* from RFC 1340. */
-#define IP_VERSION 0x40
-#define IP_HDRLEN  0x05 /* default IP header length == five 32-bits words. */
-#define IP_VHL_DEF (IP_VERSION | IP_HDRLEN)
 
 #define GRE_CHECKSUM_PRESENT   0x8000
 #define GRE_KEY_PRESENT                0x2000
 #define _htons(x) (x)
 #endif
 
-uint16_t vxlan_gpe_udp_port = 4790;
+uint16_t vxlan_gpe_udp_port = RTE_VXLAN_GPE_DEFAULT_PORT;
+uint16_t geneve_udp_port = RTE_GENEVE_DEFAULT_PORT;
 
 /* structure that caches offload info for the current packet */
 struct testpmd_offload_info {
        uint16_t ethertype;
+#ifdef RTE_LIB_GSO
        uint8_t gso_enable;
+#endif
        uint16_t l2_len;
        uint16_t l3_len;
        uint16_t l4_len;
@@ -87,15 +93,16 @@ struct testpmd_offload_info {
 struct simple_gre_hdr {
        uint16_t flags;
        uint16_t proto;
-} __attribute__((__packed__));
+} __rte_packed;
 
 static uint16_t
-get_udptcp_checksum(void *l3_hdr, void *l4_hdr, uint16_t ethertype)
+get_udptcp_checksum(struct rte_mbuf *m, void *l3_hdr, uint16_t l4_off,
+                   uint16_t ethertype)
 {
-       if (ethertype == _htons(RTE_ETHER_TYPE_IPv4))
-               return rte_ipv4_udptcp_cksum(l3_hdr, l4_hdr);
-       else /* assume ethertype == RTE_ETHER_TYPE_IPv6 */
-               return rte_ipv6_udptcp_cksum(l3_hdr, l4_hdr);
+       if (ethertype == _htons(RTE_ETHER_TYPE_IPV4))
+               return rte_ipv4_udptcp_cksum_mbuf(m, l3_hdr, l4_off);
+       else /* assume ethertype == RTE_ETHER_TYPE_IPV6 */
+               return rte_ipv6_udptcp_cksum_mbuf(m, l3_hdr, l4_off);
 }
 
 /* Parse an IPv4 header to fill l3_len, l4_len, and l4_proto */
@@ -104,7 +111,7 @@ parse_ipv4(struct rte_ipv4_hdr *ipv4_hdr, struct testpmd_offload_info *info)
 {
        struct rte_tcp_hdr *tcp_hdr;
 
-       info->l3_len = (ipv4_hdr->version_ihl & 0x0f) * 4;
+       info->l3_len = rte_ipv4_hdr_len(ipv4_hdr);
        info->l4_proto = ipv4_hdr->next_proto_id;
 
        /* only fill l4_len for TCP, it's useful for TSO */
@@ -113,7 +120,7 @@ parse_ipv4(struct rte_ipv4_hdr *ipv4_hdr, struct testpmd_offload_info *info)
                        ((char *)ipv4_hdr + info->l3_len);
                info->l4_len = (tcp_hdr->data_off & 0xf0) >> 2;
        } else if (info->l4_proto == IPPROTO_UDP)
-               info->l4_len = sizeof(struct udp_hdr);
+               info->l4_len = sizeof(struct rte_udp_hdr);
        else
                info->l4_len = 0;
 }
@@ -133,40 +140,41 @@ parse_ipv6(struct rte_ipv6_hdr *ipv6_hdr, struct testpmd_offload_info *info)
                        ((char *)ipv6_hdr + info->l3_len);
                info->l4_len = (tcp_hdr->data_off & 0xf0) >> 2;
        } else if (info->l4_proto == IPPROTO_UDP)
-               info->l4_len = sizeof(struct udp_hdr);
+               info->l4_len = sizeof(struct rte_udp_hdr);
        else
                info->l4_len = 0;
 }
 
 /*
  * Parse an ethernet header to fill the ethertype, l2_len, l3_len and
- * ipproto. This function is able to recognize IPv4/IPv6 with one optional vlan
- * header. The l4_len argument is only set in case of TCP (useful for TSO).
+ * ipproto. This function is able to recognize IPv4/IPv6 with optional VLAN
+ * headers. The l4_len argument is only set in case of TCP (useful for TSO).
  */
 static void
 parse_ethernet(struct rte_ether_hdr *eth_hdr, struct testpmd_offload_info *info)
 {
        struct rte_ipv4_hdr *ipv4_hdr;
        struct rte_ipv6_hdr *ipv6_hdr;
+       struct rte_vlan_hdr *vlan_hdr;
 
        info->l2_len = sizeof(struct rte_ether_hdr);
        info->ethertype = eth_hdr->ether_type;
 
-       if (info->ethertype == _htons(RTE_ETHER_TYPE_VLAN)) {
-               struct rte_vlan_hdr *vlan_hdr = (
-                       struct rte_vlan_hdr *)(eth_hdr + 1);
-
+       while (info->ethertype == _htons(RTE_ETHER_TYPE_VLAN) ||
+              info->ethertype == _htons(RTE_ETHER_TYPE_QINQ)) {
+               vlan_hdr = (struct rte_vlan_hdr *)
+                       ((char *)eth_hdr + info->l2_len);
                info->l2_len  += sizeof(struct rte_vlan_hdr);
                info->ethertype = vlan_hdr->eth_proto;
        }
 
        switch (info->ethertype) {
-       case _htons(RTE_ETHER_TYPE_IPv4):
+       case _htons(RTE_ETHER_TYPE_IPV4):
                ipv4_hdr = (struct rte_ipv4_hdr *)
                        ((char *)eth_hdr + info->l2_len);
                parse_ipv4(ipv4_hdr, info);
                break;
-       case _htons(RTE_ETHER_TYPE_IPv6):
+       case _htons(RTE_ETHER_TYPE_IPV6):
                ipv6_hdr = (struct rte_ipv6_hdr *)
                        ((char *)eth_hdr + info->l2_len);
                parse_ipv6(ipv6_hdr, info);
@@ -179,29 +187,91 @@ parse_ethernet(struct rte_ether_hdr *eth_hdr, struct testpmd_offload_info *info)
        }
 }
 
-/* Parse a vxlan header */
+/* Fill in outer layers length */
 static void
-parse_vxlan(struct udp_hdr *udp_hdr,
-           struct testpmd_offload_info *info,
-           uint32_t pkt_type)
+update_tunnel_outer(struct testpmd_offload_info *info)
 {
-       struct rte_ether_hdr *eth_hdr;
-
-       /* check udp destination port, 4789 is the default vxlan port
-        * (rfc7348) or that the rx offload flag is set (i40e only
-        * currently) */
-       if (udp_hdr->dst_port != _htons(4789) &&
-               RTE_ETH_IS_TUNNEL_PKT(pkt_type) == 0)
-               return;
-
        info->is_tunnel = 1;
        info->outer_ethertype = info->ethertype;
        info->outer_l2_len = info->l2_len;
        info->outer_l3_len = info->l3_len;
        info->outer_l4_proto = info->l4_proto;
+}
+
+/*
+ * Parse a GTP protocol header.
+ * No optional fields and next extension header type.
+ */
+static void
+parse_gtp(struct rte_udp_hdr *udp_hdr,
+         struct testpmd_offload_info *info)
+{
+       struct rte_ipv4_hdr *ipv4_hdr;
+       struct rte_ipv6_hdr *ipv6_hdr;
+       struct rte_gtp_hdr *gtp_hdr;
+       uint8_t gtp_len = sizeof(*gtp_hdr);
+       uint8_t ip_ver;
+
+       /* Check udp destination port. */
+       if (udp_hdr->dst_port != _htons(RTE_GTPC_UDP_PORT) &&
+           udp_hdr->src_port != _htons(RTE_GTPC_UDP_PORT) &&
+           udp_hdr->dst_port != _htons(RTE_GTPU_UDP_PORT))
+               return;
+
+       update_tunnel_outer(info);
+       info->l2_len = 0;
+
+       gtp_hdr = (struct rte_gtp_hdr *)((char *)udp_hdr +
+                 sizeof(struct rte_udp_hdr));
+       if (gtp_hdr->e || gtp_hdr->s || gtp_hdr->pn)
+               gtp_len += sizeof(struct rte_gtp_hdr_ext_word);
+       /*
+        * Check message type. If message type is 0xff, it is
+        * a GTP data packet. If not, it is a GTP control packet
+        */
+       if (gtp_hdr->msg_type == 0xff) {
+               ip_ver = *(uint8_t *)((char *)gtp_hdr + gtp_len);
+               ip_ver = (ip_ver) & 0xf0;
+
+               if (ip_ver == RTE_GTP_TYPE_IPV4) {
+                       ipv4_hdr = (struct rte_ipv4_hdr *)((char *)gtp_hdr +
+                                  gtp_len);
+                       info->ethertype = _htons(RTE_ETHER_TYPE_IPV4);
+                       parse_ipv4(ipv4_hdr, info);
+               } else if (ip_ver == RTE_GTP_TYPE_IPV6) {
+                       ipv6_hdr = (struct rte_ipv6_hdr *)((char *)gtp_hdr +
+                                  gtp_len);
+                       info->ethertype = _htons(RTE_ETHER_TYPE_IPV6);
+                       parse_ipv6(ipv6_hdr, info);
+               }
+       } else {
+               info->ethertype = 0;
+               info->l4_len = 0;
+               info->l3_len = 0;
+               info->l4_proto = 0;
+       }
+
+       info->l2_len += RTE_ETHER_GTP_HLEN;
+}
+
+/* Parse a vxlan header */
+static void
+parse_vxlan(struct rte_udp_hdr *udp_hdr,
+           struct testpmd_offload_info *info)
+{
+       struct rte_ether_hdr *eth_hdr;
+
+       /* check udp destination port, RTE_VXLAN_DEFAULT_PORT (4789) is the
+        * default vxlan port (rfc7348) or that the rx offload flag is set
+        * (i40e only currently)
+        */
+       if (udp_hdr->dst_port != _htons(RTE_VXLAN_DEFAULT_PORT))
+               return;
+
+       update_tunnel_outer(info);
 
        eth_hdr = (struct rte_ether_hdr *)((char *)udp_hdr +
-               sizeof(struct udp_hdr) +
+               sizeof(struct rte_udp_hdr) +
                sizeof(struct rte_vxlan_hdr));
 
        parse_ethernet(eth_hdr, info);
@@ -210,7 +280,7 @@ parse_vxlan(struct udp_hdr *udp_hdr,
 
 /* Parse a vxlan-gpe header */
 static void
-parse_vxlan_gpe(struct udp_hdr *udp_hdr,
+parse_vxlan_gpe(struct rte_udp_hdr *udp_hdr,
            struct testpmd_offload_info *info)
 {
        struct rte_ether_hdr *eth_hdr;
@@ -224,43 +294,31 @@ parse_vxlan_gpe(struct udp_hdr *udp_hdr,
                return;
 
        vxlan_gpe_hdr = (struct rte_vxlan_gpe_hdr *)((char *)udp_hdr +
-                               sizeof(struct udp_hdr));
+                               sizeof(struct rte_udp_hdr));
 
        if (!vxlan_gpe_hdr->proto || vxlan_gpe_hdr->proto ==
            RTE_VXLAN_GPE_TYPE_IPV4) {
-               info->is_tunnel = 1;
-               info->outer_ethertype = info->ethertype;
-               info->outer_l2_len = info->l2_len;
-               info->outer_l3_len = info->l3_len;
-               info->outer_l4_proto = info->l4_proto;
+               update_tunnel_outer(info);
 
                ipv4_hdr = (struct rte_ipv4_hdr *)((char *)vxlan_gpe_hdr +
                           vxlan_gpe_len);
 
                parse_ipv4(ipv4_hdr, info);
-               info->ethertype = _htons(RTE_ETHER_TYPE_IPv4);
+               info->ethertype = _htons(RTE_ETHER_TYPE_IPV4);
                info->l2_len = 0;
 
        } else if (vxlan_gpe_hdr->proto == RTE_VXLAN_GPE_TYPE_IPV6) {
-               info->is_tunnel = 1;
-               info->outer_ethertype = info->ethertype;
-               info->outer_l2_len = info->l2_len;
-               info->outer_l3_len = info->l3_len;
-               info->outer_l4_proto = info->l4_proto;
+               update_tunnel_outer(info);
 
                ipv6_hdr = (struct rte_ipv6_hdr *)((char *)vxlan_gpe_hdr +
                           vxlan_gpe_len);
 
-               info->ethertype = _htons(RTE_ETHER_TYPE_IPv6);
+               info->ethertype = _htons(RTE_ETHER_TYPE_IPV6);
                parse_ipv6(ipv6_hdr, info);
                info->l2_len = 0;
 
        } else if (vxlan_gpe_hdr->proto == RTE_VXLAN_GPE_TYPE_ETH) {
-               info->is_tunnel = 1;
-               info->outer_ethertype = info->ethertype;
-               info->outer_l2_len = info->l2_len;
-               info->outer_l3_len = info->l3_len;
-               info->outer_l4_proto = info->l4_proto;
+               update_tunnel_outer(info);
 
                eth_hdr = (struct rte_ether_hdr *)((char *)vxlan_gpe_hdr +
                          vxlan_gpe_len);
@@ -272,6 +330,53 @@ parse_vxlan_gpe(struct udp_hdr *udp_hdr,
        info->l2_len += RTE_ETHER_VXLAN_GPE_HLEN;
 }
 
+/* Parse a geneve header */
+static void
+parse_geneve(struct rte_udp_hdr *udp_hdr,
+           struct testpmd_offload_info *info)
+{
+       struct rte_ether_hdr *eth_hdr;
+       struct rte_ipv4_hdr *ipv4_hdr;
+       struct rte_ipv6_hdr *ipv6_hdr;
+       struct rte_geneve_hdr *geneve_hdr;
+       uint16_t geneve_len;
+
+       /* Check udp destination port. */
+       if (udp_hdr->dst_port != _htons(geneve_udp_port))
+               return;
+
+       geneve_hdr = (struct rte_geneve_hdr *)((char *)udp_hdr +
+                               sizeof(struct rte_udp_hdr));
+       geneve_len = sizeof(struct rte_geneve_hdr) + geneve_hdr->opt_len * 4;
+       if (!geneve_hdr->proto || geneve_hdr->proto ==
+           _htons(RTE_ETHER_TYPE_IPV4)) {
+               update_tunnel_outer(info);
+               ipv4_hdr = (struct rte_ipv4_hdr *)((char *)geneve_hdr +
+                          geneve_len);
+               parse_ipv4(ipv4_hdr, info);
+               info->ethertype = _htons(RTE_ETHER_TYPE_IPV4);
+               info->l2_len = 0;
+       } else if (geneve_hdr->proto == _htons(RTE_ETHER_TYPE_IPV6)) {
+               update_tunnel_outer(info);
+               ipv6_hdr = (struct rte_ipv6_hdr *)((char *)geneve_hdr +
+                          geneve_len);
+               info->ethertype = _htons(RTE_ETHER_TYPE_IPV6);
+               parse_ipv6(ipv6_hdr, info);
+               info->l2_len = 0;
+
+       } else if (geneve_hdr->proto == _htons(RTE_GENEVE_TYPE_ETH)) {
+               update_tunnel_outer(info);
+               eth_hdr = (struct rte_ether_hdr *)((char *)geneve_hdr +
+                         geneve_len);
+               parse_ethernet(eth_hdr, info);
+       } else
+               return;
+
+       info->l2_len +=
+               (sizeof(struct rte_udp_hdr) + sizeof(struct rte_geneve_hdr) +
+               ((struct rte_geneve_hdr *)geneve_hdr)->opt_len * 4);
+}
+
 /* Parse a gre header */
 static void
 parse_gre(struct simple_gre_hdr *gre_hdr, struct testpmd_offload_info *info)
@@ -290,38 +395,26 @@ parse_gre(struct simple_gre_hdr *gre_hdr, struct testpmd_offload_info *info)
        if (gre_hdr->flags & _htons(GRE_CHECKSUM_PRESENT))
                gre_len += GRE_EXT_LEN;
 
-       if (gre_hdr->proto == _htons(RTE_ETHER_TYPE_IPv4)) {
-               info->is_tunnel = 1;
-               info->outer_ethertype = info->ethertype;
-               info->outer_l2_len = info->l2_len;
-               info->outer_l3_len = info->l3_len;
-               info->outer_l4_proto = info->l4_proto;
+       if (gre_hdr->proto == _htons(RTE_ETHER_TYPE_IPV4)) {
+               update_tunnel_outer(info);
 
                ipv4_hdr = (struct rte_ipv4_hdr *)((char *)gre_hdr + gre_len);
 
                parse_ipv4(ipv4_hdr, info);
-               info->ethertype = _htons(RTE_ETHER_TYPE_IPv4);
+               info->ethertype = _htons(RTE_ETHER_TYPE_IPV4);
                info->l2_len = 0;
 
-       } else if (gre_hdr->proto == _htons(RTE_ETHER_TYPE_IPv6)) {
-               info->is_tunnel = 1;
-               info->outer_ethertype = info->ethertype;
-               info->outer_l2_len = info->l2_len;
-               info->outer_l3_len = info->l3_len;
-               info->outer_l4_proto = info->l4_proto;
+       } else if (gre_hdr->proto == _htons(RTE_ETHER_TYPE_IPV6)) {
+               update_tunnel_outer(info);
 
                ipv6_hdr = (struct rte_ipv6_hdr *)((char *)gre_hdr + gre_len);
 
-               info->ethertype = _htons(RTE_ETHER_TYPE_IPv6);
+               info->ethertype = _htons(RTE_ETHER_TYPE_IPV6);
                parse_ipv6(ipv6_hdr, info);
                info->l2_len = 0;
 
        } else if (gre_hdr->proto == _htons(RTE_ETHER_TYPE_TEB)) {
-               info->is_tunnel = 1;
-               info->outer_ethertype = info->ethertype;
-               info->outer_l2_len = info->l2_len;
-               info->outer_l3_len = info->l3_len;
-               info->outer_l4_proto = info->l4_proto;
+               update_tunnel_outer(info);
 
                eth_hdr = (struct rte_ether_hdr *)((char *)gre_hdr + gre_len);
 
@@ -353,10 +446,10 @@ parse_encap_ip(void *encap_ip, struct testpmd_offload_info *info)
 
        if (ip_version == 4) {
                parse_ipv4(ipv4_hdr, info);
-               info->ethertype = _htons(RTE_ETHER_TYPE_IPv4);
+               info->ethertype = _htons(RTE_ETHER_TYPE_IPV4);
        } else {
                parse_ipv6(ipv6_hdr, info);
-               info->ethertype = _htons(RTE_ETHER_TYPE_IPv6);
+               info->ethertype = _htons(RTE_ETHER_TYPE_IPV6);
        }
        info->l2_len = 0;
 }
@@ -365,14 +458,15 @@ parse_encap_ip(void *encap_ip, struct testpmd_offload_info *info)
  * depending on the testpmd command line configuration */
 static uint64_t
 process_inner_cksums(void *l3_hdr, const struct testpmd_offload_info *info,
-       uint64_t tx_offloads)
+       uint64_t tx_offloads, struct rte_mbuf *m)
 {
        struct rte_ipv4_hdr *ipv4_hdr = l3_hdr;
-       struct udp_hdr *udp_hdr;
+       struct rte_udp_hdr *udp_hdr;
        struct rte_tcp_hdr *tcp_hdr;
        struct rte_sctp_hdr *sctp_hdr;
        uint64_t ol_flags = 0;
        uint32_t max_pkt_len, tso_segsz = 0;
+       uint16_t l4_off;
 
        /* ensure packet is large enough to require tso */
        if (!info->is_tunnel) {
@@ -388,64 +482,80 @@ process_inner_cksums(void *l3_hdr, const struct testpmd_offload_info *info,
                        tso_segsz = info->tunnel_tso_segsz;
        }
 
-       if (info->ethertype == _htons(RTE_ETHER_TYPE_IPv4)) {
+       if (info->ethertype == _htons(RTE_ETHER_TYPE_IPV4)) {
                ipv4_hdr = l3_hdr;
-               ipv4_hdr->hdr_checksum = 0;
 
-               ol_flags |= PKT_TX_IPV4;
+               ol_flags |= RTE_MBUF_F_TX_IPV4;
                if (info->l4_proto == IPPROTO_TCP && tso_segsz) {
-                       ol_flags |= PKT_TX_IP_CKSUM;
+                       ol_flags |= RTE_MBUF_F_TX_IP_CKSUM;
                } else {
-                       if (tx_offloads & DEV_TX_OFFLOAD_IPV4_CKSUM)
-                               ol_flags |= PKT_TX_IP_CKSUM;
-                       else
+                       if (tx_offloads & RTE_ETH_TX_OFFLOAD_IPV4_CKSUM) {
+                               ol_flags |= RTE_MBUF_F_TX_IP_CKSUM;
+                       } else {
+                               ipv4_hdr->hdr_checksum = 0;
                                ipv4_hdr->hdr_checksum =
                                        rte_ipv4_cksum(ipv4_hdr);
+                       }
                }
-       } else if (info->ethertype == _htons(RTE_ETHER_TYPE_IPv6))
-               ol_flags |= PKT_TX_IPV6;
+       } else if (info->ethertype == _htons(RTE_ETHER_TYPE_IPV6))
+               ol_flags |= RTE_MBUF_F_TX_IPV6;
        else
                return 0; /* packet type not supported, nothing to do */
 
        if (info->l4_proto == IPPROTO_UDP) {
-               udp_hdr = (struct udp_hdr *)((char *)l3_hdr + info->l3_len);
+               udp_hdr = (struct rte_udp_hdr *)((char *)l3_hdr + info->l3_len);
                /* do not recalculate udp cksum if it was 0 */
                if (udp_hdr->dgram_cksum != 0) {
-                       udp_hdr->dgram_cksum = 0;
-                       if (tx_offloads & DEV_TX_OFFLOAD_UDP_CKSUM)
-                               ol_flags |= PKT_TX_UDP_CKSUM;
-                       else {
+                       if (tx_offloads & RTE_ETH_TX_OFFLOAD_UDP_CKSUM) {
+                               ol_flags |= RTE_MBUF_F_TX_UDP_CKSUM;
+                       } else {
+                               if (info->is_tunnel)
+                                       l4_off = info->outer_l2_len +
+                                                info->outer_l3_len +
+                                                info->l2_len + info->l3_len;
+                               else
+                                       l4_off = info->l2_len + info->l3_len;
+                               udp_hdr->dgram_cksum = 0;
                                udp_hdr->dgram_cksum =
-                                       get_udptcp_checksum(l3_hdr, udp_hdr,
+                                       get_udptcp_checksum(m, l3_hdr, l4_off,
                                                info->ethertype);
                        }
                }
+#ifdef RTE_LIB_GSO
                if (info->gso_enable)
-                       ol_flags |= PKT_TX_UDP_SEG;
+                       ol_flags |= RTE_MBUF_F_TX_UDP_SEG;
+#endif
        } else if (info->l4_proto == IPPROTO_TCP) {
                tcp_hdr = (struct rte_tcp_hdr *)((char *)l3_hdr + info->l3_len);
-               tcp_hdr->cksum = 0;
                if (tso_segsz)
-                       ol_flags |= PKT_TX_TCP_SEG;
-               else if (tx_offloads & DEV_TX_OFFLOAD_TCP_CKSUM)
-                       ol_flags |= PKT_TX_TCP_CKSUM;
-               else {
+                       ol_flags |= RTE_MBUF_F_TX_TCP_SEG;
+               else if (tx_offloads & RTE_ETH_TX_OFFLOAD_TCP_CKSUM) {
+                       ol_flags |= RTE_MBUF_F_TX_TCP_CKSUM;
+               } else {
+                       if (info->is_tunnel)
+                               l4_off = info->outer_l2_len + info->outer_l3_len +
+                                        info->l2_len + info->l3_len;
+                       else
+                               l4_off = info->l2_len + info->l3_len;
+                       tcp_hdr->cksum = 0;
                        tcp_hdr->cksum =
-                               get_udptcp_checksum(l3_hdr, tcp_hdr,
+                               get_udptcp_checksum(m, l3_hdr, l4_off,
                                        info->ethertype);
                }
+#ifdef RTE_LIB_GSO
                if (info->gso_enable)
-                       ol_flags |= PKT_TX_TCP_SEG;
+                       ol_flags |= RTE_MBUF_F_TX_TCP_SEG;
+#endif
        } else if (info->l4_proto == IPPROTO_SCTP) {
                sctp_hdr = (struct rte_sctp_hdr *)
                        ((char *)l3_hdr + info->l3_len);
-               sctp_hdr->cksum = 0;
                /* sctp payload must be a multiple of 4 to be
                 * offloaded */
-               if ((tx_offloads & DEV_TX_OFFLOAD_SCTP_CKSUM) &&
+               if ((tx_offloads & RTE_ETH_TX_OFFLOAD_SCTP_CKSUM) &&
                        ((ipv4_hdr->total_length & 0x3) == 0)) {
-                       ol_flags |= PKT_TX_SCTP_CKSUM;
+                       ol_flags |= RTE_MBUF_F_TX_SCTP_CKSUM;
                } else {
+                       sctp_hdr->cksum = 0;
                        /* XXX implement CRC32c, example available in
                         * RFC3309 */
                }
@@ -457,35 +567,46 @@ process_inner_cksums(void *l3_hdr, const struct testpmd_offload_info *info,
 /* Calculate the checksum of outer header */
 static uint64_t
 process_outer_cksums(void *outer_l3_hdr, struct testpmd_offload_info *info,
-       uint64_t tx_offloads, int tso_enabled)
+       uint64_t tx_offloads, int tso_enabled, struct rte_mbuf *m)
 {
        struct rte_ipv4_hdr *ipv4_hdr = outer_l3_hdr;
        struct rte_ipv6_hdr *ipv6_hdr = outer_l3_hdr;
-       struct udp_hdr *udp_hdr;
+       struct rte_udp_hdr *udp_hdr;
        uint64_t ol_flags = 0;
 
-       if (info->outer_ethertype == _htons(RTE_ETHER_TYPE_IPv4)) {
+       if (info->outer_ethertype == _htons(RTE_ETHER_TYPE_IPV4)) {
                ipv4_hdr->hdr_checksum = 0;
-               ol_flags |= PKT_TX_OUTER_IPV4;
+               ol_flags |= RTE_MBUF_F_TX_OUTER_IPV4;
 
-               if (tx_offloads & DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM)
-                       ol_flags |= PKT_TX_OUTER_IP_CKSUM;
+               if (tx_offloads & RTE_ETH_TX_OFFLOAD_OUTER_IPV4_CKSUM)
+                       ol_flags |= RTE_MBUF_F_TX_OUTER_IP_CKSUM;
                else
                        ipv4_hdr->hdr_checksum = rte_ipv4_cksum(ipv4_hdr);
        } else
-               ol_flags |= PKT_TX_OUTER_IPV6;
+               ol_flags |= RTE_MBUF_F_TX_OUTER_IPV6;
 
        if (info->outer_l4_proto != IPPROTO_UDP)
                return ol_flags;
 
+       udp_hdr = (struct rte_udp_hdr *)
+               ((char *)outer_l3_hdr + info->outer_l3_len);
+
+       if (tso_enabled)
+               ol_flags |= RTE_MBUF_F_TX_TCP_SEG;
+
        /* Skip SW outer UDP checksum generation if HW supports it */
-       if (tx_offloads & DEV_TX_OFFLOAD_OUTER_UDP_CKSUM) {
-               ol_flags |= PKT_TX_OUTER_UDP_CKSUM;
+       if (tx_offloads & RTE_ETH_TX_OFFLOAD_OUTER_UDP_CKSUM) {
+               if (info->outer_ethertype == _htons(RTE_ETHER_TYPE_IPV4))
+                       udp_hdr->dgram_cksum
+                               = rte_ipv4_phdr_cksum(ipv4_hdr, ol_flags);
+               else
+                       udp_hdr->dgram_cksum
+                               = rte_ipv6_phdr_cksum(ipv6_hdr, ol_flags);
+
+               ol_flags |= RTE_MBUF_F_TX_OUTER_UDP_CKSUM;
                return ol_flags;
        }
 
-       udp_hdr = (struct udp_hdr *)((char *)outer_l3_hdr + info->outer_l3_len);
-
        /* outer UDP checksum is done in software. In the other side, for
         * UDP tunneling, like VXLAN or Geneve, outer UDP checksum can be
         * set to zero.
@@ -500,12 +621,9 @@ process_outer_cksums(void *outer_l3_hdr, struct testpmd_offload_info *info,
        /* do not recalculate udp cksum if it was 0 */
        if (udp_hdr->dgram_cksum != 0) {
                udp_hdr->dgram_cksum = 0;
-               if (info->outer_ethertype == _htons(RTE_ETHER_TYPE_IPv4))
-                       udp_hdr->dgram_cksum =
-                               rte_ipv4_udptcp_cksum(ipv4_hdr, udp_hdr);
-               else
-                       udp_hdr->dgram_cksum =
-                               rte_ipv6_udptcp_cksum(ipv6_hdr, udp_hdr);
+               udp_hdr->dgram_cksum = get_udptcp_checksum(m, outer_l3_hdr,
+                                       info->outer_l2_len + info->outer_l3_len,
+                                       info->outer_ethertype);
        }
 
        return ol_flags;
@@ -595,7 +713,7 @@ pkt_copy_split(const struct rte_mbuf *pkt)
        mp = current_fwd_lcore()->mbp;
 
        if (tx_pkt_split == TX_PKT_SPLIT_RND)
-               nb_seg = random() % tx_pkt_nb_segs + 1;
+               nb_seg = rte_rand() % tx_pkt_nb_segs + 1;
        else
                nb_seg = tx_pkt_nb_segs;
 
@@ -660,6 +778,28 @@ pkt_copy_split(const struct rte_mbuf *pkt)
        return md[0];
 }
 
+#if defined(RTE_LIB_GRO) || defined(RTE_LIB_GSO)
+/*
+ * Re-calculate IP checksum for merged/fragmented packets.
+ */
+static void
+pkts_ip_csum_recalc(struct rte_mbuf **pkts_burst, const uint16_t nb_pkts, uint64_t tx_offloads)
+{
+       int i;
+       struct rte_ipv4_hdr *ipv4_hdr;
+       for (i = 0; i < nb_pkts; i++) {
+               if ((pkts_burst[i]->ol_flags & RTE_MBUF_F_TX_IPV4) &&
+                       (tx_offloads & RTE_ETH_TX_OFFLOAD_IPV4_CKSUM) == 0) {
+                       ipv4_hdr = rte_pktmbuf_mtod_offset(pkts_burst[i],
+                                               struct rte_ipv4_hdr *,
+                                               pkts_burst[i]->l2_len);
+                       ipv4_hdr->hdr_checksum = 0;
+                       ipv4_hdr->hdr_checksum = rte_ipv4_cksum(ipv4_hdr);
+               }
+       }
+}
+#endif
+
 /*
  * Receive a burst of packets, and for each packet:
  *  - parse packet, and try to recognize a supported packet type (1)
@@ -678,13 +818,14 @@ pkt_copy_split(const struct rte_mbuf *pkt)
  *           UDP|TCP|SCTP
  *   Ether / (vlan) / outer IP|IP6 / outer UDP / VXLAN-GPE / IP|IP6 /
  *           UDP|TCP|SCTP
+ *   Ether / (vlan) / outer IP / outer UDP / GTP / IP|IP6 / UDP|TCP|SCTP
  *   Ether / (vlan) / outer IP|IP6 / GRE / Ether / IP|IP6 / UDP|TCP|SCTP
  *   Ether / (vlan) / outer IP|IP6 / GRE / IP|IP6 / UDP|TCP|SCTP
  *   Ether / (vlan) / outer IP|IP6 / IP|IP6 / UDP|TCP|SCTP
  *
  * The testpmd command line for this forward engine sets the flags
  * TESTPMD_TX_OFFLOAD_* in ports[tx_port].tx_ol_flags. They control
- * wether a checksum must be calculated in software or in hardware. The
+ * whether a checksum must be calculated in software or in hardware. The
  * IP, UDP, TCP and SCTP flags always concern the inner layer. The
  * OUTER_IP is only useful for tunnel packets.
  */
@@ -692,16 +833,20 @@ static void
 pkt_burst_checksum_forward(struct fwd_stream *fs)
 {
        struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
+#ifdef RTE_LIB_GSO
        struct rte_mbuf *gso_segments[GSO_MAX_PKT_BURST];
        struct rte_gso_ctx *gso_ctx;
+#endif
        struct rte_mbuf **tx_pkts_burst;
        struct rte_port *txp;
        struct rte_mbuf *m, *p;
        struct rte_ether_hdr *eth_hdr;
        void *l3_hdr = NULL, *outer_l3_hdr = NULL; /* can be IPv4 or IPv6 */
+#ifdef RTE_LIB_GRO
        void **gro_ctx;
        uint16_t gro_pkts_num;
        uint8_t gro_enable;
+#endif
        uint16_t nb_rx;
        uint16_t nb_tx;
        uint16_t nb_prep;
@@ -712,41 +857,38 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
        uint32_t rx_bad_ip_csum;
        uint32_t rx_bad_l4_csum;
        uint32_t rx_bad_outer_l4_csum;
+       uint32_t rx_bad_outer_ip_csum;
        struct testpmd_offload_info info;
-       uint16_t nb_segments = 0;
-       int ret;
 
-#ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
-       uint64_t start_tsc;
-       uint64_t end_tsc;
-       uint64_t core_cycles;
-#endif
+       uint64_t start_tsc = 0;
 
-#ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
-       start_tsc = rte_rdtsc();
-#endif
+       get_start_cycles(&start_tsc);
 
        /* receive a burst of packet */
        nb_rx = rte_eth_rx_burst(fs->rx_port, fs->rx_queue, pkts_burst,
                                 nb_pkt_per_burst);
+       inc_rx_burst_stats(fs, nb_rx);
        if (unlikely(nb_rx == 0))
                return;
-#ifdef RTE_TEST_PMD_RECORD_BURST_STATS
-       fs->rx_burst_stats.pkt_burst_spread[nb_rx]++;
-#endif
+
        fs->rx_packets += nb_rx;
        rx_bad_ip_csum = 0;
        rx_bad_l4_csum = 0;
        rx_bad_outer_l4_csum = 0;
+       rx_bad_outer_ip_csum = 0;
+#ifdef RTE_LIB_GRO
        gro_enable = gro_ports[fs->rx_port].enable;
+#endif
 
        txp = &ports[fs->tx_port];
        tx_offloads = txp->dev_conf.txmode.offloads;
        memset(&info, 0, sizeof(info));
        info.tso_segsz = txp->tso_segsz;
        info.tunnel_tso_segsz = txp->tunnel_tso_segsz;
+#ifdef RTE_LIB_GSO
        if (gso_ports[fs->tx_port].enable)
                info.gso_enable = 1;
+#endif
 
        for (i = 0; i < nb_rx; i++) {
                if (likely(i < nb_rx - 1))
@@ -757,44 +899,65 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
                info.is_tunnel = 0;
                info.pkt_len = rte_pktmbuf_pkt_len(m);
                tx_ol_flags = m->ol_flags &
-                             (IND_ATTACHED_MBUF | EXT_ATTACHED_MBUF);
+                             (RTE_MBUF_F_INDIRECT | RTE_MBUF_F_EXTERNAL);
                rx_ol_flags = m->ol_flags;
 
                /* Update the L3/L4 checksum error packet statistics */
-               if ((rx_ol_flags & PKT_RX_IP_CKSUM_MASK) == PKT_RX_IP_CKSUM_BAD)
+               if ((rx_ol_flags & RTE_MBUF_F_RX_IP_CKSUM_MASK) == RTE_MBUF_F_RX_IP_CKSUM_BAD)
                        rx_bad_ip_csum += 1;
-               if ((rx_ol_flags & PKT_RX_L4_CKSUM_MASK) == PKT_RX_L4_CKSUM_BAD)
+               if ((rx_ol_flags & RTE_MBUF_F_RX_L4_CKSUM_MASK) == RTE_MBUF_F_RX_L4_CKSUM_BAD)
                        rx_bad_l4_csum += 1;
-               if (rx_ol_flags & PKT_RX_OUTER_L4_CKSUM_BAD)
+               if (rx_ol_flags & RTE_MBUF_F_RX_OUTER_L4_CKSUM_BAD)
                        rx_bad_outer_l4_csum += 1;
+               if (rx_ol_flags & RTE_MBUF_F_RX_OUTER_IP_CKSUM_BAD)
+                       rx_bad_outer_ip_csum += 1;
 
                /* step 1: dissect packet, parsing optional vlan, ip4/ip6, vxlan
                 * and inner headers */
 
                eth_hdr = rte_pktmbuf_mtod(m, struct rte_ether_hdr *);
                rte_ether_addr_copy(&peer_eth_addrs[fs->peer_addr],
-                               &eth_hdr->d_addr);
+                               &eth_hdr->dst_addr);
                rte_ether_addr_copy(&ports[fs->tx_port].eth_addr,
-                               &eth_hdr->s_addr);
+                               &eth_hdr->src_addr);
                parse_ethernet(eth_hdr, &info);
                l3_hdr = (char *)eth_hdr + info.l2_len;
 
                /* check if it's a supported tunnel */
                if (txp->parse_tunnel) {
                        if (info.l4_proto == IPPROTO_UDP) {
-                               struct udp_hdr *udp_hdr;
+                               struct rte_udp_hdr *udp_hdr;
 
-                               udp_hdr = (struct udp_hdr *)((char *)l3_hdr +
-                                       info.l3_len);
+                               udp_hdr = (struct rte_udp_hdr *)
+                                       ((char *)l3_hdr + info.l3_len);
+                               parse_gtp(udp_hdr, &info);
+                               if (info.is_tunnel) {
+                                       tx_ol_flags |= RTE_MBUF_F_TX_TUNNEL_GTP;
+                                       goto tunnel_update;
+                               }
                                parse_vxlan_gpe(udp_hdr, &info);
                                if (info.is_tunnel) {
-                                       tx_ol_flags |= PKT_TX_TUNNEL_VXLAN_GPE;
-                               } else {
-                                       parse_vxlan(udp_hdr, &info,
-                                                   m->packet_type);
-                                       if (info.is_tunnel)
-                                               tx_ol_flags |=
-                                                       PKT_TX_TUNNEL_VXLAN;
+                                       tx_ol_flags |=
+                                               RTE_MBUF_F_TX_TUNNEL_VXLAN_GPE;
+                                       goto tunnel_update;
+                               }
+                               parse_vxlan(udp_hdr, &info);
+                               if (info.is_tunnel) {
+                                       tx_ol_flags |=
+                                               RTE_MBUF_F_TX_TUNNEL_VXLAN;
+                                       goto tunnel_update;
+                               }
+                               parse_geneve(udp_hdr, &info);
+                               if (info.is_tunnel) {
+                                       tx_ol_flags |=
+                                               RTE_MBUF_F_TX_TUNNEL_GENEVE;
+                                       goto tunnel_update;
+                               }
+                               /* Always keep last. */
+                               if (unlikely(RTE_ETH_IS_TUNNEL_PKT(
+                                                       m->packet_type) != 0)) {
+                                       TESTPMD_LOG(DEBUG, "Unknown tunnel packet. UDP dst port: %hu",
+                                               udp_hdr->dst_port);
                                }
                        } else if (info.l4_proto == IPPROTO_GRE) {
                                struct simple_gre_hdr *gre_hdr;
@@ -803,17 +966,18 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
                                        ((char *)l3_hdr + info.l3_len);
                                parse_gre(gre_hdr, &info);
                                if (info.is_tunnel)
-                                       tx_ol_flags |= PKT_TX_TUNNEL_GRE;
+                                       tx_ol_flags |= RTE_MBUF_F_TX_TUNNEL_GRE;
                        } else if (info.l4_proto == IPPROTO_IPIP) {
                                void *encap_ip_hdr;
 
                                encap_ip_hdr = (char *)l3_hdr + info.l3_len;
                                parse_encap_ip(encap_ip_hdr, &info);
                                if (info.is_tunnel)
-                                       tx_ol_flags |= PKT_TX_TUNNEL_IPIP;
+                                       tx_ol_flags |= RTE_MBUF_F_TX_TUNNEL_IPIP;
                        }
                }
 
+tunnel_update:
                /* update l3_hdr and outer_l3_hdr if a tunnel was parsed */
                if (info.is_tunnel) {
                        outer_l3_hdr = l3_hdr;
@@ -827,7 +991,7 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
 
                /* process checksums of inner headers first */
                tx_ol_flags |= process_inner_cksums(l3_hdr, &info,
-                       tx_offloads);
+                       tx_offloads, m);
 
                /* Then process outer headers if any. Note that the software
                 * checksum will be wrong if one of the inner checksums is
@@ -835,7 +999,8 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
                if (info.is_tunnel == 1) {
                        tx_ol_flags |= process_outer_cksums(outer_l3_hdr, &info,
                                        tx_offloads,
-                                       !!(tx_ol_flags & PKT_TX_TCP_SEG));
+                                       !!(tx_ol_flags & RTE_MBUF_F_TX_TCP_SEG),
+                                       m);
                }
 
                /* step 3: fill the mbuf meta data (flags and header lengths) */
@@ -844,10 +1009,9 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
                if (info.is_tunnel == 1) {
                        if (info.tunnel_tso_segsz ||
                            (tx_offloads &
-                            DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM) ||
+                            RTE_ETH_TX_OFFLOAD_OUTER_IPV4_CKSUM) ||
                            (tx_offloads &
-                            DEV_TX_OFFLOAD_OUTER_UDP_CKSUM) ||
-                           (tx_ol_flags & PKT_TX_OUTER_IPV6)) {
+                            RTE_ETH_TX_OFFLOAD_OUTER_UDP_CKSUM)) {
                                m->outer_l2_len = info.outer_l2_len;
                                m->outer_l3_len = info.outer_l3_len;
                                m->l2_len = info.l2_len;
@@ -900,7 +1064,7 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
                                "l4_proto=%d l4_len=%d flags=%s\n",
                                info.l2_len, rte_be_to_cpu_16(info.ethertype),
                                info.l3_len, info.l4_proto, info.l4_len, buf);
-                       if (rx_ol_flags & PKT_RX_LRO)
+                       if (rx_ol_flags & RTE_MBUF_F_RX_LRO)
                                printf("rx: m->lro_segsz=%u\n", m->tso_segsz);
                        if (info.is_tunnel == 1)
                                printf("rx: outer_l2_len=%d outer_ethertype=%x "
@@ -908,30 +1072,30 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
                                        rte_be_to_cpu_16(info.outer_ethertype),
                                        info.outer_l3_len);
                        /* dump tx packet info */
-                       if ((tx_offloads & (DEV_TX_OFFLOAD_IPV4_CKSUM |
-                                           DEV_TX_OFFLOAD_UDP_CKSUM |
-                                           DEV_TX_OFFLOAD_TCP_CKSUM |
-                                           DEV_TX_OFFLOAD_SCTP_CKSUM)) ||
+                       if ((tx_offloads & (RTE_ETH_TX_OFFLOAD_IPV4_CKSUM |
+                                           RTE_ETH_TX_OFFLOAD_UDP_CKSUM |
+                                           RTE_ETH_TX_OFFLOAD_TCP_CKSUM |
+                                           RTE_ETH_TX_OFFLOAD_SCTP_CKSUM)) ||
                                info.tso_segsz != 0)
                                printf("tx: m->l2_len=%d m->l3_len=%d "
                                        "m->l4_len=%d\n",
                                        m->l2_len, m->l3_len, m->l4_len);
                        if (info.is_tunnel == 1) {
                                if ((tx_offloads &
-                                   DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM) ||
+                                   RTE_ETH_TX_OFFLOAD_OUTER_IPV4_CKSUM) ||
                                    (tx_offloads &
-                                   DEV_TX_OFFLOAD_OUTER_UDP_CKSUM) ||
-                                   (tx_ol_flags & PKT_TX_OUTER_IPV6))
+                                   RTE_ETH_TX_OFFLOAD_OUTER_UDP_CKSUM) ||
+                                   (tx_ol_flags & RTE_MBUF_F_TX_OUTER_IPV6))
                                        printf("tx: m->outer_l2_len=%d "
                                                "m->outer_l3_len=%d\n",
                                                m->outer_l2_len,
                                                m->outer_l3_len);
                                if (info.tunnel_tso_segsz != 0 &&
-                                               (m->ol_flags & PKT_TX_TCP_SEG))
+                                               (m->ol_flags & RTE_MBUF_F_TX_TCP_SEG))
                                        printf("tx: m->tso_segsz=%d\n",
                                                m->tso_segsz);
                        } else if (info.tso_segsz != 0 &&
-                                       (m->ol_flags & PKT_TX_TCP_SEG))
+                                       (m->ol_flags & RTE_MBUF_F_TX_TCP_SEG))
                                printf("tx: m->tso_segsz=%d\n", m->tso_segsz);
                        rte_get_tx_ol_flag_list(m->ol_flags, buf, sizeof(buf));
                        printf("tx: flags=%s", buf);
@@ -939,6 +1103,7 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
                }
        }
 
+#ifdef RTE_LIB_GRO
        if (unlikely(gro_enable)) {
                if (gro_flush_cycles == GRO_DEFAULT_FLUSH_CYCLES) {
                        nb_rx = rte_gro_reassemble_burst(pkts_burst, nb_rx,
@@ -959,20 +1124,34 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
                                fs->gro_times = 0;
                        }
                }
+
+               pkts_ip_csum_recalc(pkts_burst, nb_rx, tx_offloads);
        }
+#endif
+
+#ifdef RTE_LIB_GSO
+       if (gso_ports[fs->tx_port].enable != 0) {
+               uint16_t nb_segments = 0;
 
-       if (gso_ports[fs->tx_port].enable == 0)
-               tx_pkts_burst = pkts_burst;
-       else {
                gso_ctx = &(current_fwd_lcore()->gso_ctx);
                gso_ctx->gso_size = gso_max_segment_size;
                for (i = 0; i < nb_rx; i++) {
+                       int ret;
+
                        ret = rte_gso_segment(pkts_burst[i], gso_ctx,
                                        &gso_segments[nb_segments],
                                        GSO_MAX_PKT_BURST - nb_segments);
-                       if (ret >= 0)
+                       if (ret >= 1) {
+                               /* pkts_burst[i] can be freed safely here. */
+                               rte_pktmbuf_free(pkts_burst[i]);
                                nb_segments += ret;
-                       else {
+                       } else if (ret == 0) {
+                               /* 0 means it can be transmitted directly
+                                * without gso.
+                                */
+                               gso_segments[nb_segments] = pkts_burst[i];
+                               nb_segments += 1;
+                       } else {
                                TESTPMD_LOG(DEBUG, "Unable to segment packet");
                                rte_pktmbuf_free(pkts_burst[i]);
                        }
@@ -980,13 +1159,18 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
 
                tx_pkts_burst = gso_segments;
                nb_rx = nb_segments;
-       }
+
+               pkts_ip_csum_recalc(tx_pkts_burst, nb_rx, tx_offloads);
+       } else
+#endif
+               tx_pkts_burst = pkts_burst;
 
        nb_prep = rte_eth_tx_prepare(fs->tx_port, fs->tx_queue,
                        tx_pkts_burst, nb_rx);
        if (nb_prep != nb_rx)
-               printf("Preparing packet burst to transmit failed: %s\n",
-                               rte_strerror(rte_errno));
+               fprintf(stderr,
+                       "Preparing packet burst to transmit failed: %s\n",
+                       rte_strerror(rte_errno));
 
        nb_tx = rte_eth_tx_burst(fs->tx_port, fs->tx_queue, tx_pkts_burst,
                        nb_prep);
@@ -1006,10 +1190,9 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
        fs->rx_bad_ip_csum += rx_bad_ip_csum;
        fs->rx_bad_l4_csum += rx_bad_l4_csum;
        fs->rx_bad_outer_l4_csum += rx_bad_outer_l4_csum;
+       fs->rx_bad_outer_ip_csum += rx_bad_outer_ip_csum;
 
-#ifdef RTE_TEST_PMD_RECORD_BURST_STATS
-       fs->tx_burst_stats.pkt_burst_spread[nb_tx]++;
-#endif
+       inc_tx_burst_stats(fs, nb_tx);
        if (unlikely(nb_tx < nb_rx)) {
                fs->fwd_dropped += (nb_rx - nb_tx);
                do {
@@ -1017,16 +1200,25 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
                } while (++nb_tx < nb_rx);
        }
 
-#ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
-       end_tsc = rte_rdtsc();
-       core_cycles = (end_tsc - start_tsc);
-       fs->core_cycles = (uint64_t) (fs->core_cycles + core_cycles);
-#endif
+       get_end_cycles(fs, start_tsc);
+}
+
+static void
+stream_init_checksum_forward(struct fwd_stream *fs)
+{
+       bool rx_stopped, tx_stopped;
+
+       rx_stopped = ports[fs->rx_port].rxq[fs->rx_queue].state ==
+                                               RTE_ETH_QUEUE_STATE_STOPPED;
+       tx_stopped = ports[fs->tx_port].txq[fs->tx_queue].state ==
+                                               RTE_ETH_QUEUE_STATE_STOPPED;
+       fs->disabled = rx_stopped || tx_stopped;
 }
 
 struct fwd_engine csum_fwd_engine = {
        .fwd_mode_name  = "csum",
        .port_fwd_begin = NULL,
        .port_fwd_end   = NULL,
+       .stream_init    = stream_init_checksum_forward,
        .packet_fwd     = pkt_burst_checksum_forward,
 };