app/testpmd: add ability to split outgoing packets
[dpdk.git] / app / test-pmd / csumonly.c
index b023f12..7e4f662 100644 (file)
@@ -51,7 +51,6 @@
 #include <rte_memcpy.h>
 #include <rte_memzone.h>
 #include <rte_launch.h>
-#include <rte_tailq.h>
 #include <rte_eal.h>
 #include <rte_per_lcore.h>
 #include <rte_lcore.h>
 #define IP_HDRLEN  0x05 /* default IP header length == five 32-bits words. */
 #define IP_VHL_DEF (IP_VERSION | IP_HDRLEN)
 
+#define GRE_KEY_PRESENT 0x2000
+#define GRE_KEY_LEN     4
+#define GRE_SUPPORTED_FIELDS GRE_KEY_PRESENT
+
 /* We cannot use rte_cpu_to_be_16() on a constant in a switch/case */
 #if RTE_BYTE_ORDER == RTE_LITTLE_ENDIAN
 #define _htons(x) ((uint16_t)((((x) & 0x00ffU) << 8) | (((x) & 0xff00U) >> 8)))
 #define _htons(x) (x)
 #endif
 
+/* structure that caches offload info for the current packet */
+struct testpmd_offload_info {
+       uint16_t ethertype;
+       uint16_t l2_len;
+       uint16_t l3_len;
+       uint16_t l4_len;
+       uint8_t l4_proto;
+       uint8_t is_tunnel;
+       uint16_t outer_ethertype;
+       uint16_t outer_l2_len;
+       uint16_t outer_l3_len;
+       uint8_t outer_l4_proto;
+       uint16_t tso_segsz;
+};
+
+/* simplified GRE header */
+struct simple_gre_hdr {
+       uint16_t flags;
+       uint16_t proto;
+} __attribute__((__packed__));
+
 static uint16_t
 get_psd_sum(void *l3_hdr, uint16_t ethertype, uint64_t ol_flags)
 {
@@ -106,38 +130,36 @@ get_udptcp_checksum(void *l3_hdr, void *l4_hdr, uint16_t ethertype)
 
 /* Parse an IPv4 header to fill l3_len, l4_len, and l4_proto */
 static void
-parse_ipv4(struct ipv4_hdr *ipv4_hdr, uint16_t *l3_len, uint8_t *l4_proto,
-       uint16_t *l4_len)
+parse_ipv4(struct ipv4_hdr *ipv4_hdr, struct testpmd_offload_info *info)
 {
        struct tcp_hdr *tcp_hdr;
 
-       *l3_len = (ipv4_hdr->version_ihl & 0x0f) * 4;
-       *l4_proto = ipv4_hdr->next_proto_id;
+       info->l3_len = (ipv4_hdr->version_ihl & 0x0f) * 4;
+       info->l4_proto = ipv4_hdr->next_proto_id;
 
        /* only fill l4_len for TCP, it's useful for TSO */
-       if (*l4_proto == IPPROTO_TCP) {
-               tcp_hdr = (struct tcp_hdr *)((char *)ipv4_hdr + *l3_len);
-               *l4_len = (tcp_hdr->data_off & 0xf0) >> 2;
+       if (info->l4_proto == IPPROTO_TCP) {
+               tcp_hdr = (struct tcp_hdr *)((char *)ipv4_hdr + info->l3_len);
+               info->l4_len = (tcp_hdr->data_off & 0xf0) >> 2;
        } else
-               *l4_len = 0;
+               info->l4_len = 0;
 }
 
 /* Parse an IPv6 header to fill l3_len, l4_len, and l4_proto */
 static void
-parse_ipv6(struct ipv6_hdr *ipv6_hdr, uint16_t *l3_len, uint8_t *l4_proto,
-       uint16_t *l4_len)
+parse_ipv6(struct ipv6_hdr *ipv6_hdr, struct testpmd_offload_info *info)
 {
        struct tcp_hdr *tcp_hdr;
 
-       *l3_len = sizeof(struct ipv6_hdr);
-       *l4_proto = ipv6_hdr->proto;
+       info->l3_len = sizeof(struct ipv6_hdr);
+       info->l4_proto = ipv6_hdr->proto;
 
        /* only fill l4_len for TCP, it's useful for TSO */
-       if (*l4_proto == IPPROTO_TCP) {
-               tcp_hdr = (struct tcp_hdr *)((char *)ipv6_hdr + *l3_len);
-               *l4_len = (tcp_hdr->data_off & 0xf0) >> 2;
+       if (info->l4_proto == IPPROTO_TCP) {
+               tcp_hdr = (struct tcp_hdr *)((char *)ipv6_hdr + info->l3_len);
+               info->l4_len = (tcp_hdr->data_off & 0xf0) >> 2;
        } else
-               *l4_len = 0;
+               info->l4_len = 0;
 }
 
 /*
@@ -146,39 +168,156 @@ parse_ipv6(struct ipv6_hdr *ipv6_hdr, uint16_t *l3_len, uint8_t *l4_proto,
  * header. The l4_len argument is only set in case of TCP (useful for TSO).
  */
 static void
-parse_ethernet(struct ether_hdr *eth_hdr, uint16_t *ethertype, uint16_t *l2_len,
-       uint16_t *l3_len, uint8_t *l4_proto, uint16_t *l4_len)
+parse_ethernet(struct ether_hdr *eth_hdr, struct testpmd_offload_info *info)
 {
        struct ipv4_hdr *ipv4_hdr;
        struct ipv6_hdr *ipv6_hdr;
 
-       *l2_len = sizeof(struct ether_hdr);
-       *ethertype = eth_hdr->ether_type;
+       info->l2_len = sizeof(struct ether_hdr);
+       info->ethertype = eth_hdr->ether_type;
 
-       if (*ethertype == _htons(ETHER_TYPE_VLAN)) {
+       if (info->ethertype == _htons(ETHER_TYPE_VLAN)) {
                struct vlan_hdr *vlan_hdr = (struct vlan_hdr *)(eth_hdr + 1);
 
-               *l2_len  += sizeof(struct vlan_hdr);
-               *ethertype = vlan_hdr->eth_proto;
+               info->l2_len  += sizeof(struct vlan_hdr);
+               info->ethertype = vlan_hdr->eth_proto;
        }
 
-       switch (*ethertype) {
+       switch (info->ethertype) {
        case _htons(ETHER_TYPE_IPv4):
-               ipv4_hdr = (struct ipv4_hdr *) ((char *)eth_hdr + *l2_len);
-               parse_ipv4(ipv4_hdr, l3_len, l4_proto, l4_len);
+               ipv4_hdr = (struct ipv4_hdr *) ((char *)eth_hdr + info->l2_len);
+               parse_ipv4(ipv4_hdr, info);
                break;
        case _htons(ETHER_TYPE_IPv6):
-               ipv6_hdr = (struct ipv6_hdr *) ((char *)eth_hdr + *l2_len);
-               parse_ipv6(ipv6_hdr, l3_len, l4_proto, l4_len);
+               ipv6_hdr = (struct ipv6_hdr *) ((char *)eth_hdr + info->l2_len);
+               parse_ipv6(ipv6_hdr, info);
                break;
        default:
-               *l4_len = 0;
-               *l3_len = 0;
-               *l4_proto = 0;
+               info->l4_len = 0;
+               info->l3_len = 0;
+               info->l4_proto = 0;
                break;
        }
 }
 
+/* Parse a vxlan header */
+static void
+parse_vxlan(struct udp_hdr *udp_hdr,
+           struct testpmd_offload_info *info,
+           uint32_t pkt_type)
+{
+       struct 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;
+
+       eth_hdr = (struct ether_hdr *)((char *)udp_hdr +
+               sizeof(struct udp_hdr) +
+               sizeof(struct vxlan_hdr));
+
+       parse_ethernet(eth_hdr, info);
+       info->l2_len += ETHER_VXLAN_HLEN; /* add udp + vxlan */
+}
+
+/* Parse a gre header */
+static void
+parse_gre(struct simple_gre_hdr *gre_hdr, struct testpmd_offload_info *info)
+{
+       struct ether_hdr *eth_hdr;
+       struct ipv4_hdr *ipv4_hdr;
+       struct ipv6_hdr *ipv6_hdr;
+       uint8_t gre_len = 0;
+
+       /* check which fields are supported */
+       if ((gre_hdr->flags & _htons(~GRE_SUPPORTED_FIELDS)) != 0)
+               return;
+
+       gre_len += sizeof(struct simple_gre_hdr);
+
+       if (gre_hdr->flags & _htons(GRE_KEY_PRESENT))
+               gre_len += GRE_KEY_LEN;
+
+       if (gre_hdr->proto == _htons(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;
+
+               ipv4_hdr = (struct ipv4_hdr *)((char *)gre_hdr + gre_len);
+
+               parse_ipv4(ipv4_hdr, info);
+               info->ethertype = _htons(ETHER_TYPE_IPv4);
+               info->l2_len = 0;
+
+       } else if (gre_hdr->proto == _htons(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;
+
+               ipv6_hdr = (struct ipv6_hdr *)((char *)gre_hdr + gre_len);
+
+               info->ethertype = _htons(ETHER_TYPE_IPv6);
+               parse_ipv6(ipv6_hdr, info);
+               info->l2_len = 0;
+
+       } else if (gre_hdr->proto == _htons(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;
+
+               eth_hdr = (struct ether_hdr *)((char *)gre_hdr + gre_len);
+
+               parse_ethernet(eth_hdr, info);
+       } else
+               return;
+
+       info->l2_len += gre_len;
+}
+
+
+/* Parse an encapsulated ip or ipv6 header */
+static void
+parse_encap_ip(void *encap_ip, struct testpmd_offload_info *info)
+{
+       struct ipv4_hdr *ipv4_hdr = encap_ip;
+       struct ipv6_hdr *ipv6_hdr = encap_ip;
+       uint8_t ip_version;
+
+       ip_version = (ipv4_hdr->version_ihl & 0xf0) >> 4;
+
+       if (ip_version != 4 && ip_version != 6)
+               return;
+
+       info->is_tunnel = 1;
+       info->outer_ethertype = info->ethertype;
+       info->outer_l2_len = info->l2_len;
+       info->outer_l3_len = info->l3_len;
+
+       if (ip_version == 4) {
+               parse_ipv4(ipv4_hdr, info);
+               info->ethertype = _htons(ETHER_TYPE_IPv4);
+       } else {
+               parse_ipv6(ipv6_hdr, info);
+               info->ethertype = _htons(ETHER_TYPE_IPv6);
+       }
+       info->l2_len = 0;
+}
+
 /* modify the IPv4 or IPv4 source address of a packet */
 static void
 change_ip_addresses(void *l3_hdr, uint16_t ethertype)
@@ -197,8 +336,8 @@ change_ip_addresses(void *l3_hdr, uint16_t ethertype)
 /* if possible, calculate the checksum of a packet in hw or sw,
  * depending on the testpmd command line configuration */
 static uint64_t
-process_inner_cksums(void *l3_hdr, uint16_t ethertype, uint16_t l3_len,
-       uint8_t l4_proto, uint16_t tso_segsz, uint16_t testpmd_ol_flags)
+process_inner_cksums(void *l3_hdr, const struct testpmd_offload_info *info,
+       uint16_t testpmd_ol_flags)
 {
        struct ipv4_hdr *ipv4_hdr = l3_hdr;
        struct udp_hdr *udp_hdr;
@@ -206,12 +345,12 @@ process_inner_cksums(void *l3_hdr, uint16_t ethertype, uint16_t l3_len,
        struct sctp_hdr *sctp_hdr;
        uint64_t ol_flags = 0;
 
-       if (ethertype == _htons(ETHER_TYPE_IPv4)) {
+       if (info->ethertype == _htons(ETHER_TYPE_IPv4)) {
                ipv4_hdr = l3_hdr;
                ipv4_hdr->hdr_checksum = 0;
 
                ol_flags |= PKT_TX_IPV4;
-               if (tso_segsz != 0 && l4_proto == IPPROTO_TCP) {
+               if (info->tso_segsz != 0 && info->l4_proto == IPPROTO_TCP) {
                        ol_flags |= PKT_TX_IP_CKSUM;
                } else {
                        if (testpmd_ol_flags & TESTPMD_TX_OFFLOAD_IP_CKSUM)
@@ -220,41 +359,44 @@ process_inner_cksums(void *l3_hdr, uint16_t ethertype, uint16_t l3_len,
                                ipv4_hdr->hdr_checksum =
                                        rte_ipv4_cksum(ipv4_hdr);
                }
-       } else if (ethertype == _htons(ETHER_TYPE_IPv6))
+       } else if (info->ethertype == _htons(ETHER_TYPE_IPv6))
                ol_flags |= PKT_TX_IPV6;
        else
                return 0; /* packet type not supported, nothing to do */
 
-       if (l4_proto == IPPROTO_UDP) {
-               udp_hdr = (struct udp_hdr *)((char *)l3_hdr + l3_len);
+       if (info->l4_proto == IPPROTO_UDP) {
+               udp_hdr = (struct 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 (testpmd_ol_flags & TESTPMD_TX_OFFLOAD_UDP_CKSUM) {
                                ol_flags |= PKT_TX_UDP_CKSUM;
                                udp_hdr->dgram_cksum = get_psd_sum(l3_hdr,
-                                       ethertype, ol_flags);
+                                       info->ethertype, ol_flags);
                        } else {
                                udp_hdr->dgram_cksum =
                                        get_udptcp_checksum(l3_hdr, udp_hdr,
-                                               ethertype);
+                                               info->ethertype);
                        }
                }
-       } else if (l4_proto == IPPROTO_TCP) {
-               tcp_hdr = (struct tcp_hdr *)((char *)l3_hdr + l3_len);
+       } else if (info->l4_proto == IPPROTO_TCP) {
+               tcp_hdr = (struct tcp_hdr *)((char *)l3_hdr + info->l3_len);
                tcp_hdr->cksum = 0;
-               if (tso_segsz != 0) {
+               if (info->tso_segsz != 0) {
                        ol_flags |= PKT_TX_TCP_SEG;
-                       tcp_hdr->cksum = get_psd_sum(l3_hdr, ethertype, ol_flags);
+                       tcp_hdr->cksum = get_psd_sum(l3_hdr, info->ethertype,
+                               ol_flags);
                } else if (testpmd_ol_flags & TESTPMD_TX_OFFLOAD_TCP_CKSUM) {
                        ol_flags |= PKT_TX_TCP_CKSUM;
-                       tcp_hdr->cksum = get_psd_sum(l3_hdr, ethertype, ol_flags);
+                       tcp_hdr->cksum = get_psd_sum(l3_hdr, info->ethertype,
+                               ol_flags);
                } else {
                        tcp_hdr->cksum =
-                               get_udptcp_checksum(l3_hdr, tcp_hdr, ethertype);
+                               get_udptcp_checksum(l3_hdr, tcp_hdr,
+                                       info->ethertype);
                }
-       } else if (l4_proto == IPPROTO_SCTP) {
-               sctp_hdr = (struct sctp_hdr *)((char *)l3_hdr + l3_len);
+       } else if (info->l4_proto == IPPROTO_SCTP) {
+               sctp_hdr = (struct sctp_hdr *)((char *)l3_hdr + info->l3_len);
                sctp_hdr->cksum = 0;
                /* sctp payload must be a multiple of 4 to be
                 * offloaded */
@@ -274,15 +416,15 @@ process_inner_cksums(void *l3_hdr, uint16_t ethertype, uint16_t l3_len,
  * meaning IP + UDP). The caller already checked that it's a vxlan
  * packet */
 static uint64_t
-process_outer_cksums(void *outer_l3_hdr, uint16_t outer_ethertype,
-       uint16_t outer_l3_len, uint16_t testpmd_ol_flags)
+process_outer_cksums(void *outer_l3_hdr, struct testpmd_offload_info *info,
+       uint16_t testpmd_ol_flags)
 {
        struct ipv4_hdr *ipv4_hdr = outer_l3_hdr;
        struct ipv6_hdr *ipv6_hdr = outer_l3_hdr;
        struct udp_hdr *udp_hdr;
        uint64_t ol_flags = 0;
 
-       if (outer_ethertype == _htons(ETHER_TYPE_IPv4)) {
+       if (info->outer_ethertype == _htons(ETHER_TYPE_IPv4)) {
                ipv4_hdr->hdr_checksum = 0;
                ol_flags |= PKT_TX_OUTER_IPV4;
 
@@ -293,14 +435,17 @@ process_outer_cksums(void *outer_l3_hdr, uint16_t outer_ethertype,
        } else if (testpmd_ol_flags & TESTPMD_TX_OFFLOAD_OUTER_IP_CKSUM)
                ol_flags |= PKT_TX_OUTER_IPV6;
 
+       if (info->outer_l4_proto != IPPROTO_UDP)
+               return ol_flags;
+
        /* outer UDP checksum is always done in software as we have no
         * hardware supporting it today, and no API for it. */
 
-       udp_hdr = (struct udp_hdr *)((char *)outer_l3_hdr + outer_l3_len);
+       udp_hdr = (struct udp_hdr *)((char *)outer_l3_hdr + info->outer_l3_len);
        /* do not recalculate udp cksum if it was 0 */
        if (udp_hdr->dgram_cksum != 0) {
                udp_hdr->dgram_cksum = 0;
-               if (outer_ethertype == _htons(ETHER_TYPE_IPv4))
+               if (info->outer_ethertype == _htons(ETHER_TYPE_IPv4))
                        udp_hdr->dgram_cksum =
                                rte_ipv4_udptcp_cksum(ipv4_hdr, udp_hdr);
                else
@@ -311,6 +456,155 @@ process_outer_cksums(void *outer_l3_hdr, uint16_t outer_ethertype,
        return ol_flags;
 }
 
+/*
+ * Helper function.
+ * Performs actual copying.
+ * Returns number of segments in the destination mbuf on success,
+ * or negative error code on failure.
+ */
+static int
+mbuf_copy_split(const struct rte_mbuf *ms, struct rte_mbuf *md[],
+       uint16_t seglen[], uint8_t nb_seg)
+{
+       uint32_t dlen, slen, tlen;
+       uint32_t i, len;
+       const struct rte_mbuf *m;
+       const uint8_t *src;
+       uint8_t *dst;
+
+       dlen = 0;
+       slen = 0;
+       tlen = 0;
+
+       dst = NULL;
+       src = NULL;
+
+       m = ms;
+       i = 0;
+       while (ms != NULL && i != nb_seg) {
+
+               if (slen == 0) {
+                       slen = rte_pktmbuf_data_len(ms);
+                       src = rte_pktmbuf_mtod(ms, const uint8_t *);
+               }
+
+               if (dlen == 0) {
+                       dlen = RTE_MIN(seglen[i], slen);
+                       md[i]->data_len = dlen;
+                       md[i]->next = (i + 1 == nb_seg) ? NULL : md[i + 1];
+                       dst = rte_pktmbuf_mtod(md[i], uint8_t *);
+               }
+
+               len = RTE_MIN(slen, dlen);
+               memcpy(dst, src, len);
+               tlen += len;
+               slen -= len;
+               dlen -= len;
+               src += len;
+               dst += len;
+
+               if (slen == 0)
+                       ms = ms->next;
+               if (dlen == 0)
+                       i++;
+       }
+
+       if (ms != NULL)
+               return -ENOBUFS;
+       else if (tlen != m->pkt_len)
+               return -EINVAL;
+
+       md[0]->nb_segs = nb_seg;
+       md[0]->pkt_len = tlen;
+       md[0]->vlan_tci = m->vlan_tci;
+       md[0]->vlan_tci_outer = m->vlan_tci_outer;
+       md[0]->ol_flags = m->ol_flags;
+       md[0]->tx_offload = m->tx_offload;
+
+       return nb_seg;
+}
+
+/*
+ * Allocate a new mbuf with up to tx_pkt_nb_segs segments.
+ * Copy packet contents and offload information into then new segmented mbuf.
+ */
+static struct rte_mbuf *
+pkt_copy_split(const struct rte_mbuf *pkt)
+{
+       int32_t n, rc;
+       uint32_t i, len, nb_seg;
+       struct rte_mempool *mp;
+       uint16_t seglen[RTE_MAX_SEGS_PER_PKT];
+       struct rte_mbuf *p, *md[RTE_MAX_SEGS_PER_PKT];
+
+       mp = current_fwd_lcore()->mbp;
+
+       if (tx_pkt_split == TX_PKT_SPLIT_RND)
+               nb_seg = random() % tx_pkt_nb_segs + 1;
+       else
+               nb_seg = tx_pkt_nb_segs;
+
+       memcpy(seglen, tx_pkt_seg_lengths, nb_seg * sizeof(seglen[0]));
+
+       /* calculate number of segments to use and their length. */
+       len = 0;
+       for (i = 0; i != nb_seg && len < pkt->pkt_len; i++) {
+               len += seglen[i];
+               md[i] = NULL;
+       }
+
+       n = pkt->pkt_len - len;
+
+       /* update size of the last segment to fit rest of the packet */
+       if (n >= 0) {
+               seglen[i - 1] += n;
+               len += n;
+       }
+
+       nb_seg = i;
+       while (i != 0) {
+               p = rte_pktmbuf_alloc(mp);
+               if (p == NULL) {
+                       RTE_LOG(ERR, USER1,
+                               "failed to allocate %u-th of %u mbuf "
+                               "from mempool: %s\n",
+                               nb_seg - i, nb_seg, mp->name);
+                       break;
+               }
+
+               md[--i] = p;
+               if (rte_pktmbuf_tailroom(md[i]) < seglen[i]) {
+                       RTE_LOG(ERR, USER1, "mempool %s, %u-th segment: "
+                               "expected seglen: %u, "
+                               "actual mbuf tailroom: %u\n",
+                               mp->name, i, seglen[i],
+                               rte_pktmbuf_tailroom(md[i]));
+                       break;
+               }
+       }
+
+       /* all mbufs successfully allocated, do copy */
+       if (i == 0) {
+               rc = mbuf_copy_split(pkt, md, seglen, nb_seg);
+               if (rc < 0)
+                       RTE_LOG(ERR, USER1,
+                               "mbuf_copy_split for %p(len=%u, nb_seg=%hhu) "
+                               "into %u segments failed with error code: %d\n",
+                               pkt, pkt->pkt_len, pkt->nb_segs, nb_seg, rc);
+
+               /* figure out how many mbufs to free. */
+               i = RTE_MAX(rc, 0);
+       }
+
+       /* free unused mbufs */
+       for (; i != nb_seg; i++) {
+               rte_pktmbuf_free_seg(md[i]);
+               md[i] = NULL;
+       }
+
+       return md[0];
+}
+
 /*
  * Receive a burst of packets, and for each packet:
  *  - parse packet, and try to recognize a supported packet type (1)
@@ -326,6 +620,9 @@ process_outer_cksums(void *outer_l3_hdr, uint16_t outer_ethertype,
  *   Ether / (vlan) / IP|IP6 / UDP|TCP|SCTP .
  *   Ether / (vlan) / outer IP|IP6 / outer UDP / VxLAN / Ether / 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
@@ -338,23 +635,17 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
 {
        struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
        struct rte_port *txp;
-       struct rte_mbuf *m;
+       struct rte_mbuf *m, *p;
        struct ether_hdr *eth_hdr;
        void *l3_hdr = NULL, *outer_l3_hdr = NULL; /* can be IPv4 or IPv6 */
-       struct udp_hdr *udp_hdr;
        uint16_t nb_rx;
        uint16_t nb_tx;
        uint16_t i;
        uint64_t ol_flags;
        uint16_t testpmd_ol_flags;
-       uint8_t l4_proto, l4_tun_len = 0;
-       uint16_t ethertype = 0, outer_ethertype = 0;
-       uint16_t l2_len = 0, l3_len = 0, l4_len = 0;
-       uint16_t outer_l2_len = 0, outer_l3_len = 0;
-       uint16_t tso_segsz;
-       int tunnel = 0;
        uint32_t rx_bad_ip_csum;
        uint32_t rx_bad_l4_csum;
+       struct testpmd_offload_info info;
 
 #ifdef RTE_TEST_PMD_RECORD_CORE_CYCLES
        uint64_t start_tsc;
@@ -381,13 +672,13 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
 
        txp = &ports[fs->tx_port];
        testpmd_ol_flags = txp->tx_ol_flags;
-       tso_segsz = txp->tso_segsz;
+       memset(&info, 0, sizeof(info));
+       info.tso_segsz = txp->tso_segsz;
 
        for (i = 0; i < nb_rx; i++) {
 
                ol_flags = 0;
-               tunnel = 0;
-               l4_tun_len = 0;
+               info.is_tunnel = 0;
                m = pkts_burst[i];
 
                /* Update the L3/L4 checksum error packet statistics */
@@ -398,49 +689,44 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
                 * and inner headers */
 
                eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
-               parse_ethernet(eth_hdr, &ethertype, &l2_len, &l3_len,
-                       &l4_proto, &l4_len);
-               l3_hdr = (char *)eth_hdr + l2_len;
-
-               /* check if it's a supported tunnel (only vxlan for now) */
-               if ((testpmd_ol_flags & TESTPMD_TX_OFFLOAD_PARSE_TUNNEL) &&
-                       l4_proto == IPPROTO_UDP) {
-                       udp_hdr = (struct udp_hdr *)((char *)l3_hdr + l3_len);
-
-                       /* check udp destination port, 4789 is the default
-                        * vxlan port (rfc7348) */
-                       if (udp_hdr->dst_port == _htons(4789)) {
-                               l4_tun_len = ETHER_VXLAN_HLEN;
-                               tunnel = 1;
-
-                       /* currently, this flag is set by i40e only if the
-                        * packet is vxlan */
-                       } else if (m->ol_flags & (PKT_RX_TUNNEL_IPV4_HDR |
-                                       PKT_RX_TUNNEL_IPV6_HDR))
-                               tunnel = 1;
-
-                       if (tunnel == 1) {
-                               outer_ethertype = ethertype;
-                               outer_l2_len = l2_len;
-                               outer_l3_len = l3_len;
-                               outer_l3_hdr = l3_hdr;
-
-                               eth_hdr = (struct ether_hdr *)((char *)udp_hdr +
-                                       sizeof(struct udp_hdr) +
-                                       sizeof(struct vxlan_hdr));
-
-                               parse_ethernet(eth_hdr, &ethertype, &l2_len,
-                                       &l3_len, &l4_proto, &l4_len);
-                               l3_hdr = (char *)eth_hdr + l2_len;
+               ether_addr_copy(&peer_eth_addrs[fs->peer_addr],
+                               &eth_hdr->d_addr);
+               ether_addr_copy(&ports[fs->tx_port].eth_addr,
+                               &eth_hdr->s_addr);
+               parse_ethernet(eth_hdr, &info);
+               l3_hdr = (char *)eth_hdr + info.l2_len;
+
+               /* check if it's a supported tunnel */
+               if (testpmd_ol_flags & TESTPMD_TX_OFFLOAD_PARSE_TUNNEL) {
+                       if (info.l4_proto == IPPROTO_UDP) {
+                               struct udp_hdr *udp_hdr;
+                               udp_hdr = (struct udp_hdr *)((char *)l3_hdr +
+                                       info.l3_len);
+                               parse_vxlan(udp_hdr, &info, m->packet_type);
+                       } else if (info.l4_proto == IPPROTO_GRE) {
+                               struct simple_gre_hdr *gre_hdr;
+                               gre_hdr = (struct simple_gre_hdr *)
+                                       ((char *)l3_hdr + info.l3_len);
+                               parse_gre(gre_hdr, &info);
+                       } 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);
                        }
                }
 
+               /* update l3_hdr and outer_l3_hdr if a tunnel was parsed */
+               if (info.is_tunnel) {
+                       outer_l3_hdr = l3_hdr;
+                       l3_hdr = (char *)l3_hdr + info.outer_l3_len + info.l2_len;
+               }
+
                /* step 2: change all source IPs (v4 or v6) so we need
                 * to recompute the chksums even if they were correct */
 
-               change_ip_addresses(l3_hdr, ethertype);
-               if (tunnel == 1)
-                       change_ip_addresses(outer_l3_hdr, outer_ethertype);
+               change_ip_addresses(l3_hdr, info.ethertype);
+               if (info.is_tunnel == 1)
+                       change_ip_addresses(outer_l3_hdr, info.outer_ethertype);
 
                /* step 3: depending on user command line configuration,
                 * recompute checksum either in software or flag the
@@ -448,25 +734,25 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
                 * is configured, prepare the mbuf for TCP segmentation. */
 
                /* process checksums of inner headers first */
-               ol_flags |= process_inner_cksums(l3_hdr, ethertype,
-                       l3_len, l4_proto, tso_segsz, testpmd_ol_flags);
+               ol_flags |= process_inner_cksums(l3_hdr, &info, testpmd_ol_flags);
 
                /* Then process outer headers if any. Note that the software
                 * checksum will be wrong if one of the inner checksums is
                 * processed in hardware. */
-               if (tunnel == 1) {
-                       ol_flags |= process_outer_cksums(outer_l3_hdr,
-                               outer_ethertype, outer_l3_len, testpmd_ol_flags);
+               if (info.is_tunnel == 1) {
+                       ol_flags |= process_outer_cksums(outer_l3_hdr, &info,
+                               testpmd_ol_flags);
                }
 
                /* step 4: fill the mbuf meta data (flags and header lengths) */
 
-               if (tunnel == 1) {
+               if (info.is_tunnel == 1) {
                        if (testpmd_ol_flags & TESTPMD_TX_OFFLOAD_OUTER_IP_CKSUM) {
-                               m->outer_l2_len = outer_l2_len;
-                               m->outer_l3_len = outer_l3_len;
-                               m->l2_len = l4_tun_len + l2_len;
-                               m->l3_len = l3_len;
+                               m->outer_l2_len = info.outer_l2_len;
+                               m->outer_l3_len = info.outer_l3_len;
+                               m->l2_len = info.l2_len;
+                               m->l3_len = info.l3_len;
+                               m->l4_len = info.l4_len;
                        }
                        else {
                                /* if there is a outer UDP cksum
@@ -474,23 +760,32 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
                                   the outer checksum will be wrong as
                                   the payload will be modified by the
                                   hardware */
-                               m->l2_len = outer_l2_len + outer_l3_len +
-                                       sizeof(struct udp_hdr) +
-                                       sizeof(struct vxlan_hdr) + l2_len;
-                               m->l3_len = l3_len;
-                               m->l4_len = l4_len;
+                               m->l2_len = info.outer_l2_len +
+                                       info.outer_l3_len + info.l2_len;
+                               m->l3_len = info.l3_len;
+                               m->l4_len = info.l4_len;
                        }
                } else {
                        /* this is only useful if an offload flag is
                         * set, but it does not hurt to fill it in any
                         * case */
-                       m->l2_len = l2_len;
-                       m->l3_len = l3_len;
-                       m->l4_len = l4_len;
+                       m->l2_len = info.l2_len;
+                       m->l3_len = info.l3_len;
+                       m->l4_len = info.l4_len;
                }
-               m->tso_segsz = tso_segsz;
+               m->tso_segsz = info.tso_segsz;
                m->ol_flags = ol_flags;
 
+               /* Do split & copy for the packet. */
+               if (tx_pkt_split != TX_PKT_SPLIT_OFF) {
+                       p = pkt_copy_split(m);
+                       if (p != NULL) {
+                               rte_pktmbuf_free(m);
+                               m = p;
+                               pkts_burst[i] = m;
+                       }
+               }
+
                /* if verbose mode is enabled, dump debug info */
                if (verbose_level > 0) {
                        struct {
@@ -512,30 +807,32 @@ pkt_burst_checksum_forward(struct fwd_stream *fs)
                        const char *name;
 
                        printf("-----------------\n");
+                       printf("mbuf=%p, pkt_len=%u, nb_segs=%hhu:\n",
+                               m, m->pkt_len, m->nb_segs);
                        /* dump rx parsed packet info */
                        printf("rx: l2_len=%d ethertype=%x l3_len=%d "
                                "l4_proto=%d l4_len=%d\n",
-                               l2_len, rte_be_to_cpu_16(ethertype),
-                               l3_len, l4_proto, l4_len);
-                       if (tunnel == 1)
+                               info.l2_len, rte_be_to_cpu_16(info.ethertype),
+                               info.l3_len, info.l4_proto, info.l4_len);
+                       if (info.is_tunnel == 1)
                                printf("rx: outer_l2_len=%d outer_ethertype=%x "
-                                       "outer_l3_len=%d\n", outer_l2_len,
-                                       rte_be_to_cpu_16(outer_ethertype),
-                                       outer_l3_len);
+                                       "outer_l3_len=%d\n", info.outer_l2_len,
+                                       rte_be_to_cpu_16(info.outer_ethertype),
+                                       info.outer_l3_len);
                        /* dump tx packet info */
                        if ((testpmd_ol_flags & (TESTPMD_TX_OFFLOAD_IP_CKSUM |
                                                TESTPMD_TX_OFFLOAD_UDP_CKSUM |
                                                TESTPMD_TX_OFFLOAD_TCP_CKSUM |
                                                TESTPMD_TX_OFFLOAD_SCTP_CKSUM)) ||
-                               tso_segsz != 0)
+                               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 ((tunnel == 1) &&
+                       if ((info.is_tunnel == 1) &&
                                (testpmd_ol_flags & TESTPMD_TX_OFFLOAD_OUTER_IP_CKSUM))
                                printf("tx: m->outer_l2_len=%d m->outer_l3_len=%d\n",
                                        m->outer_l2_len, m->outer_l3_len);
-                       if (tso_segsz != 0)
+                       if (info.tso_segsz != 0)
                                printf("tx: m->tso_segsz=%d\n", m->tso_segsz);
                        printf("tx: flags=");
                        for (j = 0; j < sizeof(tx_flags)/sizeof(*tx_flags); j++) {
@@ -574,4 +871,3 @@ struct fwd_engine csum_fwd_engine = {
        .port_fwd_end   = NULL,
        .packet_fwd     = pkt_burst_checksum_forward,
 };
-