aboutsummaryrefslogtreecommitdiffstats
path: root/src/plugins
diff options
context:
space:
mode:
authorFlorin Coras <fcoras@cisco.com>2019-08-29 12:03:17 -0700
committerDamjan Marion <dmarion@me.com>2019-08-30 13:15:42 +0000
commit57c88938f8a80e63e7307cacaca23b0ab3a51b6c (patch)
treeb62690ce6a9b9d07fee1b83e8b2cf3d8fb8bdece /src/plugins
parenta495a3ea146a8484dac9f6b594fb2b044437c7a4 (diff)
vcl: allow non-blocking connects
Type: feature Change-Id: I55349f482ce6781337f747b2f0d2c0a027c3a675 Signed-off-by: Florin Coras <fcoras@cisco.com>
Diffstat (limited to 'src/plugins')
-rw-r--r--src/plugins/hs_apps/vcl/sock_test_client.c21
-rw-r--r--src/plugins/hs_apps/vcl/vcl_test_client.c15
2 files changed, 22 insertions, 14 deletions
diff --git a/src/plugins/hs_apps/vcl/sock_test_client.c b/src/plugins/hs_apps/vcl/sock_test_client.c
index 13227f28265..ed43e7faaf1 100644
--- a/src/plugins/hs_apps/vcl/sock_test_client.c
+++ b/src/plugins/hs_apps/vcl/sock_test_client.c
@@ -651,14 +651,6 @@ sock_test_connect_test_sockets (uint32_t num_test_sockets)
errno_val);
return tsock->fd;
}
- if (fcntl (tsock->fd, F_SETFL, O_NONBLOCK) < 0)
- {
- errno_val = errno;
- perror ("ERROR in sock_test_connect_test_sockets()");
- fprintf (stderr, "CLIENT: ERROR: fcntl failed (errno = %d)!\n",
- errno_val);
- return -1;
- }
#ifdef VCL_TEST
rv = vppcom_session_connect (tsock->fd, &scm->server_endpt);
@@ -668,9 +660,8 @@ sock_test_connect_test_sockets (uint32_t num_test_sockets)
rv = -1;
}
#else
- rv =
- connect (tsock->fd, (struct sockaddr *) &scm->server_addr,
- scm->server_addr_size);
+ rv = connect (tsock->fd, (struct sockaddr *) &scm->server_addr,
+ scm->server_addr_size);
#endif
if (rv < 0)
{
@@ -680,6 +671,14 @@ sock_test_connect_test_sockets (uint32_t num_test_sockets)
"(errno = %d)!\n", errno_val);
return -1;
}
+ if (fcntl (tsock->fd, F_SETFL, O_NONBLOCK) < 0)
+ {
+ errno_val = errno;
+ perror ("ERROR in sock_test_connect_test_sockets()");
+ fprintf (stderr, "CLIENT: ERROR: fcntl failed (errno = %d)!\n",
+ errno_val);
+ return -1;
+ }
tsock->cfg = ctrl->cfg;
vcl_test_session_buf_alloc (tsock);
sock_test_cfg_sync (tsock);
diff --git a/src/plugins/hs_apps/vcl/vcl_test_client.c b/src/plugins/hs_apps/vcl/vcl_test_client.c
index b9bdd6eed55..55bc7889098 100644
--- a/src/plugins/hs_apps/vcl/vcl_test_client.c
+++ b/src/plugins/hs_apps/vcl/vcl_test_client.c
@@ -123,7 +123,7 @@ vtc_quic_connect_test_sessions (vcl_test_client_worker_t * wrk)
{
vcl_test_client_main_t *vcm = &vcl_client_main;
vcl_test_session_t *ts, *tq;
- uint32_t i;
+ uint32_t i, flags, flen;
int rv;
if (wrk->cfg.num_test_sessions < 1 || wrk->cfg.num_test_sessions_perq < 1)
@@ -155,7 +155,7 @@ vtc_quic_connect_test_sessions (vcl_test_client_worker_t * wrk)
for (i = 0; i < wrk->cfg.num_test_qsessions; i++)
{
tq = &wrk->qsessions[i];
- tq->fd = vppcom_session_create (vcm->proto, 1 /* is_nonblocking */ );
+ tq->fd = vppcom_session_create (vcm->proto, 0 /* is_nonblocking */ );
tq->session_index = i;
if (tq->fd < 0)
{
@@ -163,12 +163,16 @@ vtc_quic_connect_test_sessions (vcl_test_client_worker_t * wrk)
return tq->fd;
}
+ /* Connect is blocking */
rv = vppcom_session_connect (tq->fd, &vcm->server_endpt);
if (rv < 0)
{
vterr ("vppcom_session_connect()", rv);
return rv;
}
+ flags = O_NONBLOCK;
+ flen = sizeof (flags);
+ vppcom_session_attr (tq->fd, VPPCOM_ATTR_SET_FLAGS, &flags, &flen);
vtinf ("Test Qsession %d (fd %d) connected.", i, tq->fd);
}
wrk->n_qsessions = wrk->cfg.num_test_qsessions;
@@ -223,6 +227,7 @@ vtc_connect_test_sessions (vcl_test_client_worker_t * wrk)
vcl_test_client_main_t *vcm = &vcl_client_main;
vcl_test_session_t *ts;
uint32_t n_test_sessions;
+ uint32_t flags, flen;
int i, rv;
if (vcm->proto == VPPCOM_PROTO_QUIC)
@@ -253,19 +258,23 @@ vtc_connect_test_sessions (vcl_test_client_worker_t * wrk)
for (i = 0; i < n_test_sessions; i++)
{
ts = &wrk->sessions[i];
- ts->fd = vppcom_session_create (vcm->proto, 1 /* is_nonblocking */ );
+ ts->fd = vppcom_session_create (vcm->proto, 0 /* is_nonblocking */ );
if (ts->fd < 0)
{
vterr ("vppcom_session_create()", ts->fd);
return ts->fd;
}
+ /* Connect is blocking */
rv = vppcom_session_connect (ts->fd, &vcm->server_endpt);
if (rv < 0)
{
vterr ("vppcom_session_connect()", rv);
return rv;
}
+ flags = O_NONBLOCK;
+ flen = sizeof (flags);
+ vppcom_session_attr (ts->fd, VPPCOM_ATTR_SET_FLAGS, &flags, &flen);
vtinf ("Test session %d (fd %d) connected.", i, ts->fd);
}
wrk->n_sessions = n_test_sessions;
id='n374' href='#n374'>374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 
/*
 * Copyright (c) 2015 Cisco and/or its affiliates.
 * 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 <vppinfra/error.h>
#include <vppinfra/hash.h>
#include <vnet/vnet.h>
#include <vnet/ip/ip.h>
#include <vnet/ethernet/ethernet.h>
#include <vnet/vxlan/vxlan.h>

/* Statistics (not all errors) */
#define foreach_vxlan_encap_error    \
_(ENCAPSULATED, "good packets encapsulated")

static char * vxlan_encap_error_strings[] = {
#define _(sym,string) string,
  foreach_vxlan_encap_error
#undef _
};

typedef enum {
#define _(sym,str) VXLAN_ENCAP_ERROR_##sym,
    foreach_vxlan_encap_error
#undef _
    VXLAN_ENCAP_N_ERROR,
} vxlan_encap_error_t;

typedef enum {
    VXLAN_ENCAP_NEXT_DROP,
    VXLAN_ENCAP_N_NEXT,
} vxlan_encap_next_t;

typedef struct {
  u32 tunnel_index;
  u32 vni;
} vxlan_encap_trace_t;

u8 * format_vxlan_encap_trace (u8 * s, va_list * args)
{
  CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *);
  CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *);
  vxlan_encap_trace_t * t 
      = va_arg (*args, vxlan_encap_trace_t *);

  s = format (s, "VXLAN encap to vxlan_tunnel%d vni %d", 
	      t->tunnel_index, t->vni);
  return s;
}


#define foreach_fixed_header4_offset            \
    _(0) _(1) _(2) _(3)

#define foreach_fixed_header6_offset            \
    _(0) _(1) _(2) _(3) _(4) _(5) _(6)

always_inline uword
vxlan_encap_inline (vlib_main_t * vm,
		    vlib_node_runtime_t * node,
		    vlib_frame_t * from_frame,
		    u8 is_ip4, u8 csum_offload)
{
  u32 n_left_from, next_index, * from, * to_next;
  vxlan_main_t * vxm = &vxlan_main;
  vnet_main_t * vnm = vxm->vnet_main;
  vnet_interface_main_t * im = &vnm->interface_main;
  u32 pkts_encapsulated = 0;
  u16 old_l0 = 0, old_l1 = 0;
  u32 thread_index = vlib_get_thread_index();
  u32 stats_sw_if_index, stats_n_packets, stats_n_bytes;
  u32 sw_if_index0 = 0, sw_if_index1 = 0;
  u32 next0 = 0, next1 = 0;
  vnet_hw_interface_t * hi0, * hi1;
  vxlan_tunnel_t * t0 = NULL, * t1 = NULL;

  from = vlib_frame_vector_args (from_frame);
  n_left_from = from_frame->n_vectors;

  next_index = node->cached_next_index;
  stats_sw_if_index = node->runtime_data[0];
  stats_n_packets = stats_n_bytes = 0;

  while (n_left_from > 0)
    {
      u32 n_left_to_next;

      vlib_get_next_frame (vm, node, next_index,
			   to_next, n_left_to_next);

      while (n_left_from >= 4 && n_left_to_next >= 2)
	{
          u32 bi0, bi1;
	  vlib_buffer_t * b0, * b1;
          u32 flow_hash0, flow_hash1;
	  u32 len0, len1;
          ip4_header_t * ip4_0, * ip4_1;
          ip6_header_t * ip6_0, * ip6_1;
          udp_header_t * udp0, * udp1;
          u64 * copy_src0, * copy_dst0;
          u64 * copy_src1, * copy_dst1;
          u32 * copy_src_last0, * copy_dst_last0;
          u32 * copy_src_last1, * copy_dst_last1;
          u16 new_l0, new_l1;
          ip_csum_t sum0, sum1;

	  /* Prefetch next iteration. */
	  {
	    vlib_buffer_t * p2, * p3;

	    p2 = vlib_get_buffer (vm, from[2]);
	    p3 = vlib_get_buffer (vm, from[3]);

	    vlib_prefetch_buffer_header (p2, LOAD);
	    vlib_prefetch_buffer_header (p3, LOAD);

	    CLIB_PREFETCH (p2->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
	    CLIB_PREFETCH (p3->data, 2*CLIB_CACHE_LINE_BYTES, LOAD);
	  }

	  bi0 = from[0];
	  bi1 = from[1];
	  to_next[0] = bi0;
	  to_next[1] = bi1;
	  from += 2;
	  to_next += 2;
	  n_left_to_next -= 2;
	  n_left_from -= 2;

	  b0 = vlib_get_buffer (vm, bi0);
	  b1 = vlib_get_buffer (vm, bi1);

          flow_hash0 = vnet_l2_compute_flow_hash (b0);
          flow_hash1 = vnet_l2_compute_flow_hash (b1);

	  /* Get next node index and adj index from tunnel next_dpo */
	  if (sw_if_index0 != vnet_buffer(b0)->sw_if_index[VLIB_TX])
	    {
	      sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_TX];
	      hi0 = vnet_get_sup_hw_interface (vnm, sw_if_index0);
	      t0 = &vxm->tunnels[hi0->dev_instance];
	      /* Note: change to always set next0 if it may be set to drop */
	      next0 = t0->next_dpo.dpoi_next_node;
	    }
          vnet_buffer(b0)->ip.adj_index[VLIB_TX] = t0->next_dpo.dpoi_index;

	  /* Get next node index and adj index from tunnel next_dpo */
	  if (sw_if_index1 != vnet_buffer(b1)->sw_if_index[VLIB_TX])
	    {
	      sw_if_index1 = vnet_buffer(b1)->sw_if_index[VLIB_TX];
	      hi1 = vnet_get_sup_hw_interface (vnm, sw_if_index1); 
	      t1 = &vxm->tunnels[hi1->dev_instance];
	      /* Note: change to always set next1 if it may be set to drop */
	      next1 = t1->next_dpo.dpoi_next_node;
	    }
          vnet_buffer(b1)->ip.adj_index[VLIB_TX] = t1->next_dpo.dpoi_index;

          /* Apply the rewrite string. $$$$ vnet_rewrite? */
          vlib_buffer_advance (b0, -(word)_vec_len(t0->rewrite));
          vlib_buffer_advance (b1, -(word)_vec_len(t1->rewrite));

	  if (is_ip4)
	    {
	      /* IP4 VXLAN header should be 36 octects */
              ASSERT(sizeof(ip4_vxlan_header_t) == 36);
              ASSERT(vec_len(t0->rewrite) == sizeof(ip4_vxlan_header_t));
	      ASSERT(vec_len(t1->rewrite) == sizeof(ip4_vxlan_header_t));

	      ip4_0 = vlib_buffer_get_current(b0);
	      ip4_1 = vlib_buffer_get_current(b1);

	      /* Copy the fixed header */
	      copy_dst0 = (u64 *) ip4_0;
	      copy_src0 = (u64 *) t0->rewrite;
	      copy_dst1 = (u64 *) ip4_1;
	      copy_src1 = (u64 *) t1->rewrite;
	      /* Copy first 32 octets 8-bytes at a time */
#define _(offs) copy_dst0[offs] = copy_src0[offs];
	      foreach_fixed_header4_offset;
#undef _
#define _(offs) copy_dst1[offs] = copy_src1[offs];
	      foreach_fixed_header4_offset;
#undef _
	      /* Last 4 octets. Hopefully gcc will be our friend */
              copy_dst_last0 = (u32 *)(&copy_dst0[4]);
              copy_src_last0 = (u32 *)(&copy_src0[4]);
              copy_dst_last0[0] = copy_src_last0[0];
              copy_dst_last1 = (u32 *)(&copy_dst1[4]);
              copy_src_last1 = (u32 *)(&copy_src1[4]);
              copy_dst_last1[0] = copy_src_last1[0];

	      /* Fix the IP4 checksum and length */
	      if (csum_offload)
	        {
		  ip4_0->length = clib_host_to_net_u16 
		      (vlib_buffer_length_in_chain (vm, b0));
		  b0->flags |= 
		    VNET_BUFFER_F_OFFLOAD_IP_CKSUM | VNET_BUFFER_F_IS_IP4;
		  vnet_buffer (b0)->l3_hdr_offset = (u8 *) ip4_0 - b0->data;
		  ip4_1->length = clib_host_to_net_u16 
		      (vlib_buffer_length_in_chain (vm, b1));
		  b1->flags |= 
		    VNET_BUFFER_F_OFFLOAD_IP_CKSUM | VNET_BUFFER_F_IS_IP4;
		  vnet_buffer (b1)->l3_hdr_offset = (u8 *) ip4_1 - b1->data;
	        }
	      else
	        {
		  sum0 = ip4_0->checksum;
		  new_l0 = /* old_l0 always 0, see the rewrite setup */ 
		    clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
		  sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t,
					 length /* changed member */);
		  ip4_0->checksum = ip_csum_fold (sum0);
		  ip4_0->length = new_l0;
		  sum1 = ip4_1->checksum;
		  new_l1 = /* old_l1 always 0, see the rewrite setup */
		    clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1));
		  sum1 = ip_csum_update (sum1, old_l1, new_l1, ip4_header_t,
					 length /* changed member */);
		  ip4_1->checksum = ip_csum_fold (sum1);
		  ip4_1->length = new_l1;
	        }

	      /* Fix UDP length and set source port */
	      udp0 = (udp_header_t *)(ip4_0+1);
	      new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b0)
					     - sizeof (*ip4_0));
	      udp0->length = new_l0;
	      udp0->src_port = flow_hash0;
	      udp1 = (udp_header_t *)(ip4_1+1);
	      new_l1 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b1)
					     - sizeof (*ip4_1));
	      udp1->length = new_l1;
	      udp1->src_port = flow_hash1;

	      /* UDP checksum only if checksum offload is used */
	      if (csum_offload)
	        {
		  b0->flags |= VNET_BUFFER_F_OFFLOAD_UDP_CKSUM;
		  vnet_buffer (b0)->l4_hdr_offset = (u8 *) udp0 - b0->data;
		  b1->flags |=  VNET_BUFFER_F_OFFLOAD_UDP_CKSUM;
		  vnet_buffer (b1)->l4_hdr_offset = (u8 *) udp1 - b1->data;
		}
	    }
	  else /* ipv6 */
	    {
              int bogus = 0;

	      /* IP6 VXLAN header should be 56 octects */
              ASSERT(sizeof(ip6_vxlan_header_t) == 56);
              ASSERT(vec_len(t0->rewrite) == sizeof(ip6_vxlan_header_t));
	      ASSERT(vec_len(t1->rewrite) == sizeof(ip6_vxlan_header_t));
	      ip6_0 = vlib_buffer_get_current(b0);
	      ip6_1 = vlib_buffer_get_current(b1);

	      /* Copy the fixed header */
	      copy_dst0 = (u64 *) ip6_0;
	      copy_src0 = (u64 *) t0->rewrite;
	      copy_dst1 = (u64 *) ip6_1;
	      copy_src1 = (u64 *) t1->rewrite;
	      /* Copy first 56 (ip6) octets 8-bytes at a time */
#define _(offs) copy_dst0[offs] = copy_src0[offs];
	      foreach_fixed_header6_offset;
#undef _
#define _(offs) copy_dst1[offs] = copy_src1[offs];
	      foreach_fixed_header6_offset;
#undef _
	      /* Fix IP6 payload length */
	      new_l0 =
                clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)
				      - sizeof(*ip6_0));
	      ip6_0->payload_length = new_l0;
	      new_l1 =
                clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1)
				      - sizeof(*ip6_1));
	      ip6_1->payload_length = new_l1;

	      /* Fix UDP length  and set source port */
	      udp0 = (udp_header_t *)(ip6_0+1);
	      udp0->length = new_l0;
	      udp0->src_port = flow_hash0;
	      udp1 = (udp_header_t *)(ip6_1+1);
	      udp1->length = new_l1;
	      udp1->src_port = flow_hash1;

	      /* IPv6 UDP checksum is mandatory */
	      if (csum_offload)
	        {
		  b0->flags |= VNET_BUFFER_F_OFFLOAD_UDP_CKSUM;
		  vnet_buffer (b0)->l3_hdr_offset = (u8 *) ip6_0 - b0->data;
		  vnet_buffer (b0)->l4_hdr_offset = (u8 *) udp0 - b0->data;
		  b1->flags |=  VNET_BUFFER_F_OFFLOAD_UDP_CKSUM;
		  vnet_buffer (b1)->l3_hdr_offset = (u8 *) ip6_1 - b1->data;
		  vnet_buffer (b1)->l4_hdr_offset = (u8 *) udp1 - b1->data;
		}
	      else
	        {
		  udp0->checksum = ip6_tcp_udp_icmp_compute_checksum
		      (vm, b0, ip6_0, &bogus);
		  ASSERT(bogus == 0);
		  if (udp0->checksum == 0)
		      udp0->checksum = 0xffff;
		  udp1->checksum = ip6_tcp_udp_icmp_compute_checksum
		      (vm, b1, ip6_1, &bogus);
		  ASSERT(bogus == 0);
		  if (udp1->checksum == 0)
		      udp1->checksum = 0xffff;
	        }
	    }

          pkts_encapsulated += 2;
 	  len0 = vlib_buffer_length_in_chain (vm, b0);
 	  len1 = vlib_buffer_length_in_chain (vm, b1);
	  stats_n_packets += 2;
	  stats_n_bytes += len0 + len1;

	  /* Batch stats increment on the same vxlan tunnel so counter is not
	     incremented per packet. Note stats are still incremented for deleted
	     and admin-down tunnel where packets are dropped. It is not worthwhile
	     to check for this rare case and affect normal path performance. */
	  if (PREDICT_FALSE ((sw_if_index0 != stats_sw_if_index) ||
			     (sw_if_index1 != stats_sw_if_index))) 
	    {
	      stats_n_packets -= 2;
	      stats_n_bytes -= len0 + len1;
	      if (sw_if_index0 == sw_if_index1) 
	        {
		  if (stats_n_packets) 
		    vlib_increment_combined_counter 
		      (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
		       thread_index, stats_sw_if_index, 
		       stats_n_packets, stats_n_bytes);
		  stats_sw_if_index = sw_if_index0;
		  stats_n_packets = 2;
		  stats_n_bytes = len0 + len1;
	        }
	      else 
	        {
		  vlib_increment_combined_counter 
		      (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
		       thread_index, sw_if_index0, 1, len0);
		  vlib_increment_combined_counter 
		      (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
		       thread_index, sw_if_index1, 1, len1);
		}
	    }

	  if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) 
            {
              vxlan_encap_trace_t *tr = 
                vlib_add_trace (vm, node, b0, sizeof (*tr));
              tr->tunnel_index = t0 - vxm->tunnels;
              tr->vni = t0->vni;
           }

          if (PREDICT_FALSE(b1->flags & VLIB_BUFFER_IS_TRACED)) 
            {
              vxlan_encap_trace_t *tr = 
                vlib_add_trace (vm, node, b1, sizeof (*tr));
              tr->tunnel_index = t1 - vxm->tunnels;
              tr->vni = t1->vni;
            }

	  vlib_validate_buffer_enqueue_x2 (vm, node, next_index,
					   to_next, n_left_to_next,
					   bi0, bi1, next0, next1);
	}

      while (n_left_from > 0 && n_left_to_next > 0)
	{
	  u32 bi0;
	  vlib_buffer_t * b0;
          u32 flow_hash0;
	  u32 len0;
          ip4_header_t * ip4_0;
          ip6_header_t * ip6_0;
          udp_header_t * udp0;
          u64 * copy_src0, * copy_dst0;
          u32 * copy_src_last0, * copy_dst_last0;
          u16 new_l0;
          ip_csum_t sum0;

	  bi0 = from[0];
	  to_next[0] = bi0;
	  from += 1;
	  to_next += 1;
	  n_left_from -= 1;
	  n_left_to_next -= 1;

	  b0 = vlib_get_buffer (vm, bi0);

          flow_hash0 = vnet_l2_compute_flow_hash(b0);

	  /* Get next node index and adj index from tunnel next_dpo */
	  if (sw_if_index0 != vnet_buffer(b0)->sw_if_index[VLIB_TX])
	    {
	      sw_if_index0 = vnet_buffer(b0)->sw_if_index[VLIB_TX];
	      hi0 = vnet_get_sup_hw_interface (vnm, sw_if_index0);
	      t0 = &vxm->tunnels[hi0->dev_instance];
	      /* Note: change to always set next0 if it may be set to drop */
	      next0 = t0->next_dpo.dpoi_next_node;
	    }
	  vnet_buffer(b0)->ip.adj_index[VLIB_TX] = t0->next_dpo.dpoi_index;

          /* Apply the rewrite string. $$$$ vnet_rewrite? */
          vlib_buffer_advance (b0, -(word)_vec_len(t0->rewrite));

	  if (is_ip4)
	    {
	      /* IP4 VXLAN header should be 36 octects */
              ASSERT(sizeof(ip4_vxlan_header_t) == 36);
              ASSERT(vec_len(t0->rewrite) == sizeof(ip4_vxlan_header_t));
	      ip4_0 = vlib_buffer_get_current(b0);

	      /* Copy the fixed header */
	      copy_dst0 = (u64 *) ip4_0;
	      copy_src0 = (u64 *) t0->rewrite;
	      /* Copy first 32 octets 8-bytes at a time */
#define _(offs) copy_dst0[offs] = copy_src0[offs];
	      foreach_fixed_header4_offset;
#undef _
	      /* Last 4 octets. Hopefully gcc will be our friend */
              copy_dst_last0 = (u32 *)(&copy_dst0[4]);
              copy_src_last0 = (u32 *)(&copy_src0[4]);
              copy_dst_last0[0] = copy_src_last0[0];

	      /* Fix the IP4 checksum and length */
	      if (csum_offload)
	        {
		  ip4_0->length =
		    clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
		  b0->flags |= 
		    VNET_BUFFER_F_OFFLOAD_IP_CKSUM | VNET_BUFFER_F_IS_IP4;
		  vnet_buffer (b0)->l3_hdr_offset = (u8 *) ip4_0 - b0->data;
	        }
	      else
	        {
		  sum0 = ip4_0->checksum;
		  new_l0 = /* old_l0 always 0, see the rewrite setup */ 
		    clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0));
		  sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t,
					 length /* changed member */);
		  ip4_0->checksum = ip_csum_fold (sum0);
		  ip4_0->length = new_l0;
	        }

	      /* Fix UDP length and set source port */
	      udp0 = (udp_header_t *)(ip4_0+1);
	      new_l0 = clib_host_to_net_u16 (vlib_buffer_length_in_chain(vm, b0)
					     - sizeof (*ip4_0));
	      udp0->length = new_l0;
	      udp0->src_port = flow_hash0;

	      /* UDP checksum only if checksum offload is used */
	      if (csum_offload)
	        {
		  b0->flags |= VNET_BUFFER_F_OFFLOAD_UDP_CKSUM;
		  vnet_buffer (b0)->l4_hdr_offset = (u8 *) udp0 - b0->data;
	        }
	    }

	  else /* ip6 path */
	    {
              int bogus = 0;

	      /* IP6 VXLAN header should be 56 octects */
              ASSERT(sizeof(ip6_vxlan_header_t) == 56);
              ASSERT(vec_len(t0->rewrite) == sizeof(ip6_vxlan_header_t));
	      ip6_0 = vlib_buffer_get_current(b0);
	      /* Copy the fixed header */
	      copy_dst0 = (u64 *) ip6_0;
	      copy_src0 = (u64 *) t0->rewrite;
	      /* Copy first 56 (ip6) octets 8-bytes at a time */
#define _(offs) copy_dst0[offs] = copy_src0[offs];
	      foreach_fixed_header6_offset;
#undef _
	      /* Fix IP6 payload length */
	      new_l0 =
                clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)
				      - sizeof(*ip6_0));
	      ip6_0->payload_length = new_l0;

	      /* Fix UDP length  and set source port */
	      udp0 = (udp_header_t *)(ip6_0+1);
	      udp0->length = new_l0;
	      udp0->src_port = flow_hash0;

	      /* IPv6 UDP checksum is mandatory */
	      if (csum_offload)
	        {
		  b0->flags |= VNET_BUFFER_F_OFFLOAD_UDP_CKSUM;
		  vnet_buffer (b0)->l3_hdr_offset = (u8 *) ip6_0 - b0->data;
		  vnet_buffer (b0)->l4_hdr_offset = (u8 *) udp0 - b0->data;
	        }
	      else
	        {
		  udp0->checksum = ip6_tcp_udp_icmp_compute_checksum
		    (vm, b0, ip6_0, &bogus);
		  ASSERT(bogus == 0);
		  if (udp0->checksum == 0)
		    udp0->checksum = 0xffff;
	        }
	    }

          pkts_encapsulated ++;
	  len0 = vlib_buffer_length_in_chain (vm, b0);
	  stats_n_packets += 1;
	  stats_n_bytes += len0;

	  /* Batch stats increment on the same vxlan tunnel so counter is not
	     incremented per packet. Note stats are still incremented for deleted
	     and admin-down tunnel where packets are dropped. It is not worthwhile
	     to check for this rare case and affect normal path performance. */
	  if (PREDICT_FALSE (sw_if_index0 != stats_sw_if_index)) 
	    {
	      stats_n_packets -= 1;
	      stats_n_bytes -= len0;
	      if (stats_n_packets)
		vlib_increment_combined_counter 
		  (im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
		   thread_index, stats_sw_if_index, 
		   stats_n_packets, stats_n_bytes);
	      stats_n_packets = 1;
	      stats_n_bytes = len0;
	      stats_sw_if_index = sw_if_index0;
	    }

          if (PREDICT_FALSE(b0->flags & VLIB_BUFFER_IS_TRACED)) 
            {
              vxlan_encap_trace_t *tr = 
                vlib_add_trace (vm, node, b0, sizeof (*tr));
              tr->tunnel_index = t0 - vxm->tunnels;
              tr->vni = t0->vni;
            }
	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index,
					   to_next, n_left_to_next,
					   bi0, next0);
	}

      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
    }

  /* Do we still need this now that tunnel tx stats is kept? */
  vlib_node_increment_counter (vm, node->node_index, 
                               VXLAN_ENCAP_ERROR_ENCAPSULATED, 
                               pkts_encapsulated);

  /* Increment any remaining batch stats */
  if (stats_n_packets)
    {
      vlib_increment_combined_counter 
	(im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX,
	 thread_index, stats_sw_if_index, stats_n_packets, stats_n_bytes);
      node->runtime_data[0] = stats_sw_if_index;
    }

  return from_frame->n_vectors;
}

static uword
vxlan4_encap (vlib_main_t * vm,
	      vlib_node_runtime_t * node,
	      vlib_frame_t * from_frame)
{
  /* Disable chksum offload as setup overhead in tx node is not worthwhile
     for ip4 header checksum only, unless udp checksum is also required */
  return vxlan_encap_inline (vm, node, from_frame, /* is_ip4 */ 1, 
			     /* csum_offload */ 0);
}

static uword
vxlan6_encap (vlib_main_t * vm,
	      vlib_node_runtime_t * node,
	      vlib_frame_t * from_frame)
{
  /* Enable checksum offload for ip6 as udp checksum is mandatory, */
  return vxlan_encap_inline (vm, node, from_frame, /* is_ip4 */ 0, 
			     /* csum_offload */ 1);
}

VLIB_REGISTER_NODE (vxlan4_encap_node) = {
  .function = vxlan4_encap,
  .name = "vxlan4-encap",
  .vector_size = sizeof (u32),
  .format_trace = format_vxlan_encap_trace,
  .type = VLIB_NODE_TYPE_INTERNAL,
  .n_errors = ARRAY_LEN(vxlan_encap_error_strings),
  .error_strings = vxlan_encap_error_strings,
  .n_next_nodes = VXLAN_ENCAP_N_NEXT,
  .next_nodes = {
        [VXLAN_ENCAP_NEXT_DROP] = "error-drop",
  },
};

VLIB_NODE_FUNCTION_MULTIARCH (vxlan4_encap_node, vxlan4_encap)

VLIB_REGISTER_NODE (vxlan6_encap_node) = {
  .function = vxlan6_encap,
  .name = "vxlan6-encap",
  .vector_size = sizeof (u32),
  .format_trace = format_vxlan_encap_trace,
  .type = VLIB_NODE_TYPE_INTERNAL,
  .n_errors = ARRAY_LEN(vxlan_encap_error_strings),
  .error_strings = vxlan_encap_error_strings,
  .n_next_nodes = VXLAN_ENCAP_N_NEXT,
  .next_nodes = {
        [VXLAN_ENCAP_NEXT_DROP] = "error-drop",
  },
};

VLIB_NODE_FUNCTION_MULTIARCH (vxlan6_encap_node, vxlan6_encap)