gso: Add gso feature arc

Type: refactor

Change-Id: Ib4a8e7e64093b2db1deda6663325080bc8337605
Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com>

9 files changed, 1051 insertions, 377 deletions

diff --git a/src/vnet/CMakeLists.txt b/src/vnet/CMakeLists.txt
index 9c1bceb5753..7120809636e 100644
--- a/src/vnet/CMakeLists.txt
+++ b/src/vnet/CMakeLists.txt
@@ -941,6 +941,24 @@ list(APPEND VNET_HEADERS
 list(APPEND VNET_API_FILES ipfix-export/ipfix_export.api)
 
 ##############################################################################
+# GSO
+##############################################################################
+list(APPEND VNET_SOURCES
+  gso/cli.c
+  gso/gso.c
+  gso/gso_api.c
+  gso/node.c
+)
+
+list(APPEND VNET_HEADERS
+  gso/gso.h
+)
+
+list(APPEND VNET_API_FILES
+  gso/gso.api
+)
+
+##############################################################################
 # IPFIX classify code
 ##############################################################################
 
diff --git a/src/vnet/gso/cli.c b/src/vnet/gso/cli.c
new file mode 100644
index 00000000000..060ce812fad
--- /dev/null
+++ b/src/vnet/gso/cli.c
@@ -0,0 +1,93 @@
+/*
+ * Copyright (c) 2019 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 <vlib/vlib.h>
+#include <vnet/vnet.h>
+#include <vppinfra/error.h>
+#include <vnet/ethernet/ethernet.h>
+#include <vnet/feature/feature.h>
+#include <vnet/gso/gso.h>
+
+static clib_error_t *
+set_interface_feature_gso_command_fn (vlib_main_t * vm,
+				      unformat_input_t * input,
+				      vlib_cli_command_t * cmd)
+{
+  vnet_main_t *vnm = vnet_get_main ();
+  unformat_input_t _line_input, *line_input = &_line_input;
+  clib_error_t *error = 0;
+
+  u32 sw_if_index = ~0;
+  u8 enable = 0;
+
+  /* Get a line of input. */
+  if (!unformat_user (input, unformat_line_input, line_input))
+    return 0;
+
+  while (unformat_check_input (line_input) != UNFORMAT_END_OF_INPUT)
+    {
+      if (unformat
+	  (line_input, "%U", unformat_vnet_sw_interface, vnm, &sw_if_index))
+	;
+      else if (unformat (line_input, "enable"))
+	enable = 1;
+      else if (unformat (line_input, "disable"))
+	enable = 0;
+      else
+	{
+	  error = unformat_parse_error (line_input);
+	  goto done;
+	}
+    }
+
+  if (sw_if_index == ~0)
+    {
+      error = clib_error_return (0, "Interface not specified...");
+      goto done;
+    }
+  int rv = vnet_sw_interface_gso_enable_disable (sw_if_index, enable);
+
+  switch (rv)
+    {
+    case VNET_API_ERROR_INVALID_VALUE:
+      error = clib_error_return (0, "interface type is not hardware");
+      break;
+    case VNET_API_ERROR_FEATURE_DISABLED:
+      error = clib_error_return (0, "interface should be ethernet interface");
+      break;
+    default:
+      ;
+    }
+
+done:
+  unformat_free (line_input);
+  return error;
+}
+
+/* *INDENT-OFF* */
+VLIB_CLI_COMMAND (set_interface_feature_gso_command, static) = {
+  .path = "set interface feature gso",
+  .short_help = "set interface feature gso <intfc> [enable | disable]",
+  .function = set_interface_feature_gso_command_fn,
+};
+/* *INDENT-ON* */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/gso/gso.api b/src/vnet/gso/gso.api
new file mode 100644
index 00000000000..8b1da88f040
--- /dev/null
+++ b/src/vnet/gso/gso.api
@@ -0,0 +1,39 @@
+/*
+ * Copyright (c) 2019 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.
+ */
+
+option version = "1.0.0";
+
+import "vnet/interface_types.api";
+
+/** \brief Enable or disable interface feature gso arc
+    @param client_index - opaque cookie to identify the sender
+    @param context - sender context, to match reply w/ request
+    @param sw_if_index - The interface to enable/disable gso feature arc.
+    @param enable_disable - set to 1 to enable, 0 to disable gso feature arc
+*/
+autoreply define feature_gso_enable_disable
+{
+  u32 client_index;
+  u32 context;
+  vl_api_interface_index_t sw_if_index;
+  bool  enable_disable;
+  option vat_help = "<intfc> | sw_if_index <nn> [enable | disable]";
+};
+
+/*
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/gso/gso.c b/src/vnet/gso/gso.c
new file mode 100644
index 00000000000..c741b17bdd4
--- /dev/null
+++ b/src/vnet/gso/gso.c
@@ -0,0 +1,88 @@
+/*
+ * Copyright (c) 2019 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 <vlib/vlib.h>
+#include <vnet/vnet.h>
+#include <vppinfra/error.h>
+#include <vnet/ethernet/ethernet.h>
+#include <vnet/feature/feature.h>
+#include <vnet/l2/l2_in_out_feat_arc.h>
+#include <vnet/gso/gso.h>
+
+gso_main_t gso_main;
+
+int
+vnet_sw_interface_gso_enable_disable (u32 sw_if_index, u8 enable)
+{
+  ethernet_interface_t *eif;
+  vnet_sw_interface_t *si;
+  ethernet_main_t *em;
+  vnet_main_t *vnm;
+
+  vnm = vnet_get_main ();
+  em = &ethernet_main;
+  si = vnet_get_sw_interface (vnm, sw_if_index);
+
+  /*
+   * only ethernet HW interfaces are supported at this time
+   */
+  if (si->type != VNET_SW_INTERFACE_TYPE_HARDWARE)
+    {
+      return (VNET_API_ERROR_INVALID_VALUE);
+    }
+
+  eif = ethernet_get_interface (em, si->hw_if_index);
+
+  if (!eif)
+    {
+      return (VNET_API_ERROR_FEATURE_DISABLED);
+    }
+
+  vnet_feature_enable_disable ("ip4-output", "gso-ip4", sw_if_index, enable,
+			       0, 0);
+  vnet_feature_enable_disable ("ip6-output", "gso-ip6", sw_if_index, enable,
+			       0, 0);
+
+  vnet_l2_feature_enable_disable ("l2-output-nonip", "gso-l2-nonip",
+				  sw_if_index, enable, 0, 0);
+  vnet_l2_feature_enable_disable ("l2-output-ip4", "gso-l2-ip4",
+				  sw_if_index, enable, 0, 0);
+  vnet_l2_feature_enable_disable ("l2-output-ip6", "gso-l2-ip6",
+				  sw_if_index, enable, 0, 0);
+
+  return (0);
+}
+
+static clib_error_t *
+gso_init (vlib_main_t * vm)
+{
+  gso_main_t *gm = &gso_main;
+
+  clib_memset (gm, 0, sizeof (gm[0]));
+  gm->vlib_main = vm;
+  gm->vnet_main = vnet_get_main ();
+
+  return 0;
+}
+
+VLIB_INIT_FUNCTION (gso_init);
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/gso/gso.h b/src/vnet/gso/gso.h
new file mode 100644
index 00000000000..8e174dfd1f6
--- /dev/null
+++ b/src/vnet/gso/gso.h
@@ -0,0 +1,40 @@
+/*
+ * Copyright (c) 2019 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.
+ */
+
+#ifndef included_gso_h
+#define included_gso_h
+
+#include <vnet/vnet.h>
+
+typedef struct
+{
+  vlib_main_t *vlib_main;
+  vnet_main_t *vnet_main;
+  u16 msg_id_base;
+} gso_main_t;
+
+extern gso_main_t gso_main;
+
+int vnet_sw_interface_gso_enable_disable (u32 sw_if_index, u8 enable);
+
+#endif /* included_gso_h */
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/gso/gso_api.c b/src/vnet/gso/gso_api.c
new file mode 100644
index 00000000000..9991c5fd289
--- /dev/null
+++ b/src/vnet/gso/gso_api.c
@@ -0,0 +1,106 @@
+/*
+ *------------------------------------------------------------------
+ * gso_api.c - Generic Segmentation Offload api
+ *
+ * Copyright (c) 2019 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 <vnet/vnet.h>
+#include <vlibmemory/api.h>
+#include <vnet/gso/gso.h>
+
+#include <vnet/vnet_msg_enum.h>
+
+#define vl_typedefs		/* define message structures */
+#include <vnet/vnet_all_api_h.h>
+#undef vl_typedefs
+
+#define vl_endianfun		/* define message structures */
+#include <vnet/vnet_all_api_h.h>
+#undef vl_endianfun
+
+/* instantiate all the print functions we know about */
+#define vl_print(handle, ...) vlib_cli_output (handle, __VA_ARGS__)
+#define vl_printfun
+#include <vnet/vnet_all_api_h.h>
+#undef vl_printfun
+
+#include <vlibapi/api_helper_macros.h>
+
+#define foreach_feature_gso_api_msg                                              \
+_(FEATURE_GSO_ENABLE_DISABLE, feature_gso_enable_disable)
+
+static void
+  vl_api_feature_gso_enable_disable_t_handler
+  (vl_api_feature_gso_enable_disable_t * mp)
+{
+  vl_api_feature_gso_enable_disable_reply_t *rmp;
+  int rv = 0;
+
+  VALIDATE_SW_IF_INDEX (mp);
+
+  rv =
+    vnet_sw_interface_gso_enable_disable (ntohl (mp->sw_if_index),
+					  mp->enable_disable);
+
+  BAD_SW_IF_INDEX_LABEL;
+
+  REPLY_MACRO (VL_API_FEATURE_GSO_ENABLE_DISABLE_REPLY);
+}
+
+#define vl_msg_name_crc_list
+#include <vnet/gso/gso.api.h>
+#undef vl_msg_name_crc_list
+
+static void
+setup_message_id_table (api_main_t * am)
+{
+#define _(id,n,crc) vl_msg_api_add_msg_name_crc (am, #n "_" #crc, id);
+  foreach_vl_msg_name_crc_gso;
+#undef _
+}
+
+static clib_error_t *
+feature_gso_api_hookup (vlib_main_t * vm)
+{
+  api_main_t *am = &api_main;
+
+#define _(N,n)                                                  \
+    vl_msg_api_set_handlers(VL_API_##N, #n,                     \
+                           vl_api_##n##_t_handler,              \
+                           vl_noop_handler,                     \
+                           vl_api_##n##_t_endian,               \
+                           vl_api_##n##_t_print,                \
+                           sizeof(vl_api_##n##_t), 1);
+  foreach_feature_gso_api_msg;
+#undef _
+
+  /*
+   * Set up the (msg_name, crc, message-id) table
+   */
+  setup_message_id_table (am);
+
+  return 0;
+}
+
+VLIB_API_INIT_FUNCTION (feature_gso_api_hookup);
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/gso/node.c b/src/vnet/gso/node.c
new file mode 100644
index 00000000000..5998a4f8387
--- /dev/null
+++ b/src/vnet/gso/node.c
@@ -0,0 +1,652 @@
+/*
+ * Copyright (c) 2018 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 <vlib/vlib.h>
+#include <vnet/vnet.h>
+#include <vppinfra/error.h>
+#include <vnet/ethernet/ethernet.h>
+#include <vnet/feature/feature.h>
+#include <vnet/gso/gso.h>
+#include <vnet/ip/icmp46_packet.h>
+#include <vnet/ip/ip4.h>
+#include <vnet/ip/ip6.h>
+#include <vnet/udp/udp_packet.h>
+
+typedef struct
+{
+  u32 flags;
+  u16 gso_size;
+  u8 gso_l4_hdr_sz;
+} gso_trace_t;
+
+static u8 *
+format_gso_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 *);
+  gso_trace_t *t = va_arg (*args, gso_trace_t *);
+
+  if (t->flags & VNET_BUFFER_F_GSO)
+    {
+      s = format (s, "gso_sz %d gso_l4_hdr_sz %d",
+		  t->gso_size, t->gso_l4_hdr_sz);
+    }
+
+  return s;
+}
+
+static_always_inline u16
+tso_alloc_tx_bufs (vlib_main_t * vm,
+		   vnet_interface_per_thread_data_t * ptd,
+		   vlib_buffer_t * b0, u32 n_bytes_b0, u16 l234_sz,
+		   u16 gso_size)
+{
+  u16 size =
+    clib_min (gso_size, vlib_buffer_get_default_data_size (vm) - l234_sz);
+
+  /* rounded-up division */
+  u16 n_bufs = (n_bytes_b0 - l234_sz + (size - 1)) / size;
+  u16 n_alloc;
+
+  ASSERT (n_bufs > 0);
+  vec_validate (ptd->split_buffers, n_bufs - 1);
+
+  n_alloc = vlib_buffer_alloc (vm, ptd->split_buffers, n_bufs);
+  if (n_alloc < n_bufs)
+    {
+      vlib_buffer_free (vm, ptd->split_buffers, n_alloc);
+      return 0;
+    }
+  return n_alloc;
+}
+
+static_always_inline void
+tso_init_buf_from_template_base (vlib_buffer_t * nb0, vlib_buffer_t * b0,
+				 u32 flags, u16 length)
+{
+  nb0->current_data = b0->current_data;
+  nb0->total_length_not_including_first_buffer = 0;
+  nb0->flags = VLIB_BUFFER_TOTAL_LENGTH_VALID | flags;
+  clib_memcpy_fast (&nb0->opaque, &b0->opaque, sizeof (nb0->opaque));
+  clib_memcpy_fast (vlib_buffer_get_current (nb0),
+		    vlib_buffer_get_current (b0), length);
+  nb0->current_length = length;
+}
+
+static_always_inline void
+tso_init_buf_from_template (vlib_main_t * vm, vlib_buffer_t * nb0,
+			    vlib_buffer_t * b0, u16 template_data_sz,
+			    u16 gso_size, u8 ** p_dst_ptr, u16 * p_dst_left,
+			    u32 next_tcp_seq, u32 flags)
+{
+  tso_init_buf_from_template_base (nb0, b0, flags, template_data_sz);
+
+  *p_dst_left =
+    clib_min (gso_size,
+	      vlib_buffer_get_default_data_size (vm) - (template_data_sz +
+							nb0->current_data));
+  *p_dst_ptr = vlib_buffer_get_current (nb0) + template_data_sz;
+
+  tcp_header_t *tcp =
+    (tcp_header_t *) (nb0->data + vnet_buffer (nb0)->l4_hdr_offset);
+  tcp->seq_number = clib_host_to_net_u32 (next_tcp_seq);
+}
+
+static_always_inline void
+tso_fixup_segmented_buf (vlib_buffer_t * b0, u8 tcp_flags, int is_ip6)
+{
+  u16 l3_hdr_offset = vnet_buffer (b0)->l3_hdr_offset;
+  u16 l4_hdr_offset = vnet_buffer (b0)->l4_hdr_offset;
+  ip4_header_t *ip4 = (ip4_header_t *) (b0->data + l3_hdr_offset);
+  ip6_header_t *ip6 = (ip6_header_t *) (b0->data + l3_hdr_offset);
+  tcp_header_t *tcp = (tcp_header_t *) (b0->data + l4_hdr_offset);
+
+  tcp->flags = tcp_flags;
+
+  if (is_ip6)
+    ip6->payload_length =
+      clib_host_to_net_u16 (b0->current_length -
+			    (l4_hdr_offset - b0->current_data));
+  else
+    ip4->length =
+      clib_host_to_net_u16 (b0->current_length -
+			    (l3_hdr_offset - b0->current_data));
+}
+
+/**
+ * Allocate the necessary number of ptd->split_buffers,
+ * and segment the possibly chained buffer(s) from b0 into
+ * there.
+ *
+ * Return the cumulative number of bytes sent or zero
+ * if allocation failed.
+ */
+
+static_always_inline u32
+tso_segment_buffer (vlib_main_t * vm, vnet_interface_per_thread_data_t * ptd,
+		    u32 sbi0, vlib_buffer_t * sb0, u32 n_bytes_b0, int is_ip6)
+{
+  u32 n_tx_bytes = 0;
+  ASSERT (sb0->flags & VNET_BUFFER_F_L2_HDR_OFFSET_VALID);
+  ASSERT (sb0->flags & VNET_BUFFER_F_L3_HDR_OFFSET_VALID);
+  ASSERT (sb0->flags & VNET_BUFFER_F_L4_HDR_OFFSET_VALID);
+  u16 gso_size = vnet_buffer2 (sb0)->gso_size;
+
+  int l4_hdr_sz = vnet_buffer2 (sb0)->gso_l4_hdr_sz;
+  u8 save_tcp_flags = 0;
+  u8 tcp_flags_no_fin_psh = 0;
+  u32 next_tcp_seq = 0;
+
+  tcp_header_t *tcp =
+    (tcp_header_t *) (sb0->data + vnet_buffer (sb0)->l4_hdr_offset);
+  next_tcp_seq = clib_net_to_host_u32 (tcp->seq_number);
+  /* store original flags for last packet and reset FIN and PSH */
+  save_tcp_flags = tcp->flags;
+  tcp_flags_no_fin_psh = tcp->flags & ~(TCP_FLAG_FIN | TCP_FLAG_PSH);
+  tcp->checksum = 0;
+
+  u32 default_bflags =
+    sb0->flags & ~(VNET_BUFFER_F_GSO | VLIB_BUFFER_NEXT_PRESENT);
+  u16 l234_sz = vnet_buffer (sb0)->l4_hdr_offset + l4_hdr_sz
+    - sb0->current_data;
+  int first_data_size = clib_min (gso_size, sb0->current_length - l234_sz);
+  next_tcp_seq += first_data_size;
+
+  if (PREDICT_FALSE
+      (!tso_alloc_tx_bufs (vm, ptd, sb0, n_bytes_b0, l234_sz, gso_size)))
+    return 0;
+
+  vlib_buffer_t *b0 = vlib_get_buffer (vm, ptd->split_buffers[0]);
+  tso_init_buf_from_template_base (b0, sb0, default_bflags,
+				   l234_sz + first_data_size);
+
+  u32 total_src_left = n_bytes_b0 - l234_sz - first_data_size;
+  if (total_src_left)
+    {
+      /* Need to copy more segments */
+      u8 *src_ptr, *dst_ptr;
+      u16 src_left, dst_left;
+      /* current source buffer */
+      vlib_buffer_t *csb0 = sb0;
+      u32 csbi0 = sbi0;
+      /* current dest buffer */
+      vlib_buffer_t *cdb0;
+      u16 dbi = 1;		/* the buffer [0] is b0 */
+
+      src_ptr = vlib_buffer_get_current (sb0) + l234_sz + first_data_size;
+      src_left = sb0->current_length - l234_sz - first_data_size;
+
+      tso_fixup_segmented_buf (b0, tcp_flags_no_fin_psh, is_ip6);
+
+      /* grab a second buffer and prepare the loop */
+      ASSERT (dbi < vec_len (ptd->split_buffers));
+      cdb0 = vlib_get_buffer (vm, ptd->split_buffers[dbi++]);
+      tso_init_buf_from_template (vm, cdb0, b0, l234_sz, gso_size, &dst_ptr,
+				  &dst_left, next_tcp_seq, default_bflags);
+
+      /* an arbitrary large number to catch the runaway loops */
+      int nloops = 2000;
+      while (total_src_left)
+	{
+	  if (nloops-- <= 0)
+	    clib_panic ("infinite loop detected");
+	  u16 bytes_to_copy = clib_min (src_left, dst_left);
+
+	  clib_memcpy_fast (dst_ptr, src_ptr, bytes_to_copy);
+
+	  src_left -= bytes_to_copy;
+	  src_ptr += bytes_to_copy;
+	  total_src_left -= bytes_to_copy;
+	  dst_left -= bytes_to_copy;
+	  dst_ptr += bytes_to_copy;
+	  next_tcp_seq += bytes_to_copy;
+	  cdb0->current_length += bytes_to_copy;
+
+	  if (0 == src_left)
+	    {
+	      int has_next = (csb0->flags & VLIB_BUFFER_NEXT_PRESENT);
+	      u32 next_bi = csb0->next_buffer;
+
+	      /* init src to the next buffer in chain */
+	      if (has_next)
+		{
+		  csbi0 = next_bi;
+		  csb0 = vlib_get_buffer (vm, csbi0);
+		  src_left = csb0->current_length;
+		  src_ptr = vlib_buffer_get_current (csb0);
+		}
+	      else
+		{
+		  ASSERT (total_src_left == 0);
+		  break;
+		}
+	    }
+	  if (0 == dst_left && total_src_left)
+	    {
+	      n_tx_bytes += cdb0->current_length;
+	      ASSERT (dbi < vec_len (ptd->split_buffers));
+	      cdb0 = vlib_get_buffer (vm, ptd->split_buffers[dbi++]);
+	      tso_init_buf_from_template (vm, cdb0, b0, l234_sz,
+					  gso_size, &dst_ptr, &dst_left,
+					  next_tcp_seq, default_bflags);
+	    }
+	}
+
+      tso_fixup_segmented_buf (cdb0, save_tcp_flags, is_ip6);
+
+      n_tx_bytes += cdb0->current_length;
+    }
+  n_tx_bytes += b0->current_length;
+  return n_tx_bytes;
+}
+
+static_always_inline void
+drop_one_buffer_and_count (vlib_main_t * vm, vnet_main_t * vnm,
+			   vlib_node_runtime_t * node, u32 * pbi0,
+			   u32 sw_if_index, u32 drop_error_code)
+{
+  u32 thread_index = vm->thread_index;
+
+  vlib_simple_counter_main_t *cm;
+  cm =
+    vec_elt_at_index (vnm->interface_main.sw_if_counters,
+		      VNET_INTERFACE_COUNTER_TX_ERROR);
+  vlib_increment_simple_counter (cm, thread_index, sw_if_index, 1);
+
+  vlib_error_drop_buffers (vm, node, pbi0,
+			   /* buffer stride */ 1,
+			   /* n_buffers */ 1,
+			   VNET_INTERFACE_OUTPUT_NEXT_DROP,
+			   node->node_index, drop_error_code);
+}
+
+static_always_inline uword
+vnet_gso_node_inline (vlib_main_t * vm,
+		      vlib_node_runtime_t * node,
+		      vlib_frame_t * frame,
+		      vnet_main_t * vnm,
+		      vnet_hw_interface_t * hi,
+		      int is_ip6, int do_segmentation)
+{
+  u32 *to_next;
+  u32 next_index = node->cached_next_index;
+  u32 *from = vlib_frame_vector_args (frame);
+  u32 n_left_from = frame->n_vectors;
+  u32 *from_end = from + n_left_from;
+  u32 thread_index = vm->thread_index;
+  vnet_interface_main_t *im = &vnm->interface_main;
+  vnet_interface_per_thread_data_t *ptd =
+    vec_elt_at_index (im->per_thread_data, thread_index);
+  vlib_buffer_t *bufs[VLIB_FRAME_SIZE], **b = bufs;
+
+  vlib_get_buffers (vm, from, b, n_left_from);
+
+  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 (from + 8 <= from_end && n_left_to_next >= 4)
+	{
+	  u32 bi0, bi1, bi2, bi3;
+	  u32 next0, next1, next2, next3;
+	  u32 swif0, swif1, swif2, swif3;
+	  gso_trace_t *t0, *t1, *t2, *t3;
+	  vnet_hw_interface_t *hi0, *hi1, *hi2, *hi3;
+
+	  /* Prefetch next iteration. */
+	  vlib_prefetch_buffer_header (b[4], LOAD);
+	  vlib_prefetch_buffer_header (b[5], LOAD);
+	  vlib_prefetch_buffer_header (b[6], LOAD);
+	  vlib_prefetch_buffer_header (b[7], LOAD);
+
+	  bi0 = from[0];
+	  bi1 = from[1];
+	  bi2 = from[2];
+	  bi3 = from[3];
+	  to_next[0] = bi0;
+	  to_next[1] = bi1;
+	  to_next[2] = bi2;
+	  to_next[3] = bi3;
+
+	  swif0 = vnet_buffer (b[0])->sw_if_index[VLIB_TX];
+	  swif1 = vnet_buffer (b[1])->sw_if_index[VLIB_TX];
+	  swif2 = vnet_buffer (b[2])->sw_if_index[VLIB_TX];
+	  swif3 = vnet_buffer (b[3])->sw_if_index[VLIB_TX];
+
+	  if (PREDICT_FALSE (hi->sw_if_index != swif0))
+	    {
+	      hi0 = vnet_get_sup_hw_interface (vnm, swif0);
+	      if ((hi0->flags & VNET_HW_INTERFACE_FLAG_SUPPORTS_GSO) == 0 &&
+		  (b[0]->flags & VNET_BUFFER_F_GSO))
+		break;
+	    }
+	  if (PREDICT_FALSE (hi->sw_if_index != swif1))
+	    {
+	      hi1 = vnet_get_sup_hw_interface (vnm, swif0);
+	      if (!(hi1->flags & VNET_HW_INTERFACE_FLAG_SUPPORTS_GSO) &&
+		  (b[1]->flags & VNET_BUFFER_F_GSO))
+		break;
+	    }
+	  if (PREDICT_FALSE (hi->sw_if_index != swif2))
+	    {
+	      hi2 = vnet_get_sup_hw_interface (vnm, swif0);
+	      if ((hi2->flags & VNET_HW_INTERFACE_FLAG_SUPPORTS_GSO) == 0 &&
+		  (b[2]->flags & VNET_BUFFER_F_GSO))
+		break;
+	    }
+	  if (PREDICT_FALSE (hi->sw_if_index != swif3))
+	    {
+	      hi3 = vnet_get_sup_hw_interface (vnm, swif0);
+	      if (!(hi3->flags & VNET_HW_INTERFACE_FLAG_SUPPORTS_GSO) &&
+		  (b[3]->flags & VNET_BUFFER_F_GSO))
+		break;
+	    }
+
+	  if (b[0]->flags & VLIB_BUFFER_IS_TRACED)
+	    {
+	      t0 = vlib_add_trace (vm, node, b[0], sizeof (t0[0]));
+	      t0->flags = b[0]->flags & VNET_BUFFER_F_GSO;
+	      t0->gso_size = vnet_buffer2 (b[0])->gso_size;
+	      t0->gso_l4_hdr_sz = vnet_buffer2 (b[0])->gso_l4_hdr_sz;
+	    }
+	  if (b[1]->flags & VLIB_BUFFER_IS_TRACED)
+	    {
+	      t1 = vlib_add_trace (vm, node, b[1], sizeof (t1[0]));
+	      t1->flags = b[1]->flags & VNET_BUFFER_F_GSO;
+	      t1->gso_size = vnet_buffer2 (b[1])->gso_size;
+	      t1->gso_l4_hdr_sz = vnet_buffer2 (b[1])->gso_l4_hdr_sz;
+	    }
+	  if (b[2]->flags & VLIB_BUFFER_IS_TRACED)
+	    {
+	      t2 = vlib_add_trace (vm, node, b[2], sizeof (t2[0]));
+	      t2->flags = b[2]->flags & VNET_BUFFER_F_GSO;
+	      t2->gso_size = vnet_buffer2 (b[2])->gso_size;
+	      t2->gso_l4_hdr_sz = vnet_buffer2 (b[2])->gso_l4_hdr_sz;
+	    }
+	  if (b[3]->flags & VLIB_BUFFER_IS_TRACED)
+	    {
+	      t3 = vlib_add_trace (vm, node, b[3], sizeof (t3[0]));
+	      t3->flags = b[3]->flags & VNET_BUFFER_F_GSO;
+	      t3->gso_size = vnet_buffer2 (b[3])->gso_size;
+	      t3->gso_l4_hdr_sz = vnet_buffer2 (b[3])->gso_l4_hdr_sz;
+	    }
+
+	  from += 4;
+	  to_next += 4;
+	  n_left_to_next -= 4;
+	  n_left_from -= 4;
+
+	  next0 = next1 = 0;
+	  next2 = next3 = 0;
+	  vnet_feature_next (&next0, b[0]);
+	  vnet_feature_next (&next1, b[1]);
+	  vnet_feature_next (&next2, b[2]);
+	  vnet_feature_next (&next3, b[3]);
+	  vlib_validate_buffer_enqueue_x4 (vm, node, next_index, to_next,
+					   n_left_to_next, bi0, bi1, bi2, bi3,
+					   next0, next1, next2, next3);
+	  b += 4;
+	}
+
+      while (from + 1 <= from_end && n_left_to_next > 0)
+	{
+	  u32 bi0, swif0;
+	  gso_trace_t *t0;
+	  vnet_hw_interface_t *hi0;
+	  u32 next0 = 0;
+
+	  swif0 = vnet_buffer (b[0])->sw_if_index[VLIB_TX];
+	  if (PREDICT_FALSE (hi->sw_if_index != swif0))
+	    {
+	      hi0 = vnet_get_sup_hw_interface (vnm, swif0);
+	      if ((hi0->flags & VNET_HW_INTERFACE_FLAG_SUPPORTS_GSO) == 0 &&
+		  (b[0]->flags & VNET_BUFFER_F_GSO))
+		do_segmentation = 1;
+	    }
+
+	  /* speculatively enqueue b0 to the current next frame */
+	  to_next[0] = bi0 = from[0];
+	  to_next += 1;
+	  n_left_to_next -= 1;
+	  from += 1;
+	  n_left_from -= 1;
+
+	  if (b[0]->flags & VLIB_BUFFER_IS_TRACED)
+	    {
+	      t0 = vlib_add_trace (vm, node, b[0], sizeof (t0[0]));
+	      t0->flags = b[0]->flags & VNET_BUFFER_F_GSO;
+	      t0->gso_size = vnet_buffer2 (b[0])->gso_size;
+	      t0->gso_l4_hdr_sz = vnet_buffer2 (b[0])->gso_l4_hdr_sz;
+	    }
+
+	  if (do_segmentation)
+	    {
+	      if (PREDICT_FALSE (b[0]->flags & VNET_BUFFER_F_GSO))
+		{
+		  /*
+		   * Undo the enqueue of the b0 - it is not going anywhere,
+		   * and will be freed either after it's segmented or
+		   * when dropped, if there is no buffers to segment into.
+		   */
+		  to_next -= 1;
+		  n_left_to_next += 1;
+		  /* undo the counting. */
+		  u32 n_bytes_b0 = vlib_buffer_length_in_chain (vm, b[0]);
+		  u32 n_tx_bytes = 0;
+
+		  n_tx_bytes =
+		    tso_segment_buffer (vm, ptd, bi0, b[0], n_bytes_b0,
+					is_ip6);
+
+		  if (PREDICT_FALSE (n_tx_bytes == 0))
+		    {
+		      drop_one_buffer_and_count (vm, vnm, node, from - 1,
+						 hi->sw_if_index,
+						 VNET_INTERFACE_OUTPUT_ERROR_NO_BUFFERS_FOR_GSO);
+		      b += 1;
+		      continue;
+		    }
+
+		  u16 n_tx_bufs = vec_len (ptd->split_buffers);
+		  u32 *from_seg = ptd->split_buffers;
+
+		  while (n_tx_bufs > 0)
+		    {
+		      u32 sbi0;
+		      vlib_buffer_t *sb0;
+		      if (n_tx_bufs >= n_left_to_next)
+			{
+			  while (n_left_to_next > 0)
+			    {
+			      sbi0 = to_next[0] = from_seg[0];
+			      sb0 = vlib_get_buffer (vm, sbi0);
+			      to_next += 1;
+			      from_seg += 1;
+			      n_left_to_next -= 1;
+			      n_tx_bufs -= 1;
+			      vnet_feature_next (&next0, sb0);
+			      vlib_validate_buffer_enqueue_x1 (vm, node,
+							       next_index,
+							       to_next,
+							       n_left_to_next,
+							       sbi0, next0);
+			    }
+			  vlib_put_next_frame (vm, node, next_index,
+					       n_left_to_next);
+			  vlib_get_new_next_frame (vm, node, next_index,
+						   to_next, n_left_to_next);
+			}
+		      while (n_tx_bufs > 0)
+			{
+			  sbi0 = to_next[0] = from_seg[0];
+			  sb0 = vlib_get_buffer (vm, sbi0);
+			  to_next += 1;
+			  from_seg += 1;
+			  n_left_to_next -= 1;
+			  n_tx_bufs -= 1;
+			  vnet_feature_next (&next0, sb0);
+			  vlib_validate_buffer_enqueue_x1 (vm, node,
+							   next_index,
+							   to_next,
+							   n_left_to_next,
+							   sbi0, next0);
+			}
+		    }
+		  /* The buffers were enqueued. Reset the length */
+		  _vec_len (ptd->split_buffers) = 0;
+		  /* Free the now segmented buffer */
+		  vlib_buffer_free_one (vm, bi0);
+		  b += 1;
+		  continue;
+		}
+	    }
+
+	  vnet_feature_next (&next0, b[0]);
+	  vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next,
+					   n_left_to_next, bi0, next0);
+	  b += 1;
+	}
+      vlib_put_next_frame (vm, node, next_index, n_left_to_next);
+    }
+
+  return frame->n_vectors;
+}
+
+static_always_inline uword
+vnet_gso_inline (vlib_main_t * vm,
+		 vlib_node_runtime_t * node, vlib_frame_t * frame, int is_ip6)
+{
+  vnet_main_t *vnm = vnet_get_main ();
+  vnet_hw_interface_t *hi;
+
+  if (frame->n_vectors > 0)
+    {
+      u32 *from = vlib_frame_vector_args (frame);
+      vlib_buffer_t *b = vlib_get_buffer (vm, from[0]);
+      hi = vnet_get_sup_hw_interface (vnm,
+				      vnet_buffer (b)->sw_if_index[VLIB_TX]);
+      /*
+       * The 3-headed "if" is here because we want to err on the side
+       * of not impacting the non-GSO performance - so for the more
+       * common case of no GSO interfaces we want to prevent the
+       * segmentation codepath from being there altogether.
+       */
+      if (hi->flags & VNET_HW_INTERFACE_FLAG_SUPPORTS_GSO)
+	return vnet_gso_node_inline (vm, node, frame, vnm, hi,
+				     is_ip6, /* do_segmentation */ 0);
+      else
+	return vnet_gso_node_inline (vm, node, frame, vnm, hi,
+				     is_ip6, /* do_segmentation */ 1);
+    }
+  return 0;
+}
+
+VLIB_NODE_FN (gso_l2_ip4_node) (vlib_main_t * vm, vlib_node_runtime_t * node,
+				vlib_frame_t * frame)
+{
+  return vnet_gso_inline (vm, node, frame, 0 /* ip6 */ );
+}
+
+VLIB_NODE_FN (gso_l2_ip6_node) (vlib_main_t * vm, vlib_node_runtime_t * node,
+				vlib_frame_t * frame)
+{
+  return vnet_gso_inline (vm, node, frame, 1 /* ip6 */ );
+}
+
+VLIB_NODE_FN (gso_ip4_node) (vlib_main_t * vm, vlib_node_runtime_t * node,
+			     vlib_frame_t * frame)
+{
+  return vnet_gso_inline (vm, node, frame, 0 /* ip6 */ );
+}
+
+VLIB_NODE_FN (gso_ip6_node) (vlib_main_t * vm, vlib_node_runtime_t * node,
+			     vlib_frame_t * frame)
+{
+  return vnet_gso_inline (vm, node, frame, 1 /* ip6 */ );
+}
+
+/* *INDENT-OFF* */
+
+VLIB_REGISTER_NODE (gso_l2_ip4_node) = {
+  .vector_size = sizeof (u32),
+  .format_trace = format_gso_trace,
+  .type = VLIB_NODE_TYPE_INTERNAL,
+  .n_errors = 0,
+  .n_next_nodes = 0,
+  .name = "gso-l2-ip4",
+};
+
+VLIB_REGISTER_NODE (gso_l2_ip6_node) = {
+  .vector_size = sizeof (u32),
+  .format_trace = format_gso_trace,
+  .type = VLIB_NODE_TYPE_INTERNAL,
+  .n_errors = 0,
+  .n_next_nodes = 0,
+  .name = "gso-l2-ip6",
+};
+
+VLIB_REGISTER_NODE (gso_ip4_node) = {
+  .vector_size = sizeof (u32),
+  .format_trace = format_gso_trace,
+  .type = VLIB_NODE_TYPE_INTERNAL,
+  .n_errors = 0,
+  .n_next_nodes = 0,
+  .name = "gso-ip4",
+};
+
+VLIB_REGISTER_NODE (gso_ip6_node) = {
+  .vector_size = sizeof (u32),
+  .format_trace = format_gso_trace,
+  .type = VLIB_NODE_TYPE_INTERNAL,
+  .n_errors = 0,
+  .n_next_nodes = 0,
+  .name = "gso-ip6",
+};
+
+VNET_FEATURE_INIT (gso_l2_ip4_node, static) = {
+  .arc_name = "l2-output-ip4",
+  .node_name = "gso-l2-ip4",
+  .runs_before = VNET_FEATURES ("l2-output-feat-arc-end"),
+};
+
+VNET_FEATURE_INIT (gso_l2_ip6_node, static) = {
+  .arc_name = "l2-output-ip6",
+  .node_name = "gso-l2-ip6",
+  .runs_before = VNET_FEATURES ("l2-output-feat-arc-end"),
+};
+
+VNET_FEATURE_INIT (gso_ip4_node, static) = {
+  .arc_name = "ip4-output",
+  .node_name = "gso-ip4",
+  .runs_after = VNET_FEATURES ("ipsec4-output-feature"),
+  .runs_before = VNET_FEATURES ("interface-output"),
+};
+
+VNET_FEATURE_INIT (gso_ip6_node, static) = {
+  .arc_name = "ip6-output",
+  .node_name = "gso-ip6",
+  .runs_after = VNET_FEATURES ("ipsec6-output-feature"),
+  .runs_before = VNET_FEATURES ("interface-output"),
+};
+
+/*
+ * fd.io coding-style-patch-verification: ON
+ *
+ * Local Variables:
+ * eval: (c-set-style "gnu")
+ * End:
+ */
diff --git a/src/vnet/interface_output.c b/src/vnet/interface_output.c
index 707b1167a48..5e702e3b96c 100644
--- a/src/vnet/interface_output.c
+++ b/src/vnet/interface_output.c
@@ -49,9 +49,7 @@ typedef struct
 {
   u32 sw_if_index;
   u32 flags;
-  u16 gso_size;
-  u8 gso_l4_hdr_sz;
-  u8 data[128 - 3 * sizeof (u32)];
+  u8 data[128 - 2 * sizeof (u32)];
 }
 interface_output_trace_t;
 
@@ -83,17 +81,6 @@ format_vnet_interface_output_trace (u8 * s, va_list * va)
 	    format (s, "%U ", format_vnet_sw_interface_name, vnm, si,
 		    t->flags);
 	}
-#define _(bit, name, v, x) \
-          if (v && (t->flags & VNET_BUFFER_F_##name)) \
-            s = format (s, "%s ", v);
-      foreach_vnet_buffer_flag
-#undef _
-	if (t->flags & VNET_BUFFER_F_GSO)
-	{
-	  s = format (s, "\n%Ugso_sz %d gso_l4_hdr_sz %d",
-		      format_white_space, indent + 2, t->gso_size,
-		      t->gso_l4_hdr_sz);
-	}
       s =
 	format (s, "\n%U%U", format_white_space, indent,
 		node->format_buffer ? node->format_buffer : format_hex_bytes,
@@ -133,8 +120,6 @@ vnet_interface_output_trace (vlib_main_t * vm,
 	  t0 = vlib_add_trace (vm, node, b0, sizeof (t0[0]));
 	  t0->sw_if_index = vnet_buffer (b0)->sw_if_index[VLIB_TX];
 	  t0->flags = b0->flags;
-	  t0->gso_size = vnet_buffer2 (b0)->gso_size;
-	  t0->gso_l4_hdr_sz = vnet_buffer2 (b0)->gso_l4_hdr_sz;
 	  clib_memcpy_fast (t0->data, vlib_buffer_get_current (b0),
 			    sizeof (t0->data));
 	}
@@ -143,8 +128,6 @@ vnet_interface_output_trace (vlib_main_t * vm,
 	  t1 = vlib_add_trace (vm, node, b1, sizeof (t1[0]));
 	  t1->sw_if_index = vnet_buffer (b1)->sw_if_index[VLIB_TX];
 	  t1->flags = b1->flags;
-	  t1->gso_size = vnet_buffer2 (b1)->gso_size;
-	  t1->gso_l4_hdr_sz = vnet_buffer2 (b1)->gso_l4_hdr_sz;
 	  clib_memcpy_fast (t1->data, vlib_buffer_get_current (b1),
 			    sizeof (t1->data));
 	}
@@ -167,8 +150,6 @@ vnet_interface_output_trace (vlib_main_t * vm,
 	  t0 = vlib_add_trace (vm, node, b0, sizeof (t0[0]));
 	  t0->sw_if_index = vnet_buffer (b0)->sw_if_index[VLIB_TX];
 	  t0->flags = b0->flags;
-	  t0->gso_size = vnet_buffer2 (b0)->gso_size;
-	  t0->gso_l4_hdr_sz = vnet_buffer2 (b0)->gso_l4_hdr_sz;
 	  clib_memcpy_fast (t0->data, vlib_buffer_get_current (b0),
 			    sizeof (t0->data));
 	}
@@ -230,250 +211,13 @@ calc_checksums (vlib_main_t * vm, vlib_buffer_t * b)
   b->flags &= ~VNET_BUFFER_F_OFFLOAD_IP_CKSUM;
 }
 
-static_always_inline u16
-tso_alloc_tx_bufs (vlib_main_t * vm,
-		   vnet_interface_per_thread_data_t * ptd,
-		   vlib_buffer_t * b0, u32 n_bytes_b0, u16 l234_sz,
-		   u16 gso_size)
-{
-  u16 size =
-    clib_min (gso_size, vlib_buffer_get_default_data_size (vm) - l234_sz);
-
-  /* rounded-up division */
-  u16 n_bufs = (n_bytes_b0 - l234_sz + (size - 1)) / size;
-  u16 n_alloc;
-
-  ASSERT (n_bufs > 0);
-  vec_validate (ptd->split_buffers, n_bufs - 1);
-
-  n_alloc = vlib_buffer_alloc (vm, ptd->split_buffers, n_bufs);
-  if (n_alloc < n_bufs)
-    {
-      vlib_buffer_free (vm, ptd->split_buffers, n_alloc);
-      return 0;
-    }
-  return n_alloc;
-}
-
-static_always_inline void
-tso_init_buf_from_template_base (vlib_buffer_t * nb0, vlib_buffer_t * b0,
-				 u32 flags, u16 length)
-{
-  nb0->current_data = b0->current_data;
-  nb0->total_length_not_including_first_buffer = 0;
-  nb0->flags = VLIB_BUFFER_TOTAL_LENGTH_VALID | flags;
-  clib_memcpy_fast (&nb0->opaque, &b0->opaque, sizeof (nb0->opaque));
-  clib_memcpy_fast (vlib_buffer_get_current (nb0),
-		    vlib_buffer_get_current (b0), length);
-  nb0->current_length = length;
-}
-
-static_always_inline void
-tso_init_buf_from_template (vlib_main_t * vm, vlib_buffer_t * nb0,
-			    vlib_buffer_t * b0, u16 template_data_sz,
-			    u16 gso_size, u8 ** p_dst_ptr, u16 * p_dst_left,
-			    u32 next_tcp_seq, u32 flags)
-{
-  tso_init_buf_from_template_base (nb0, b0, flags, template_data_sz);
-
-  *p_dst_left =
-    clib_min (gso_size,
-	      vlib_buffer_get_default_data_size (vm) - (template_data_sz +
-							nb0->current_data));
-  *p_dst_ptr = vlib_buffer_get_current (nb0) + template_data_sz;
-
-  tcp_header_t *tcp =
-    (tcp_header_t *) (nb0->data + vnet_buffer (nb0)->l4_hdr_offset);
-  tcp->seq_number = clib_host_to_net_u32 (next_tcp_seq);
-}
-
-static_always_inline void
-tso_fixup_segmented_buf (vlib_buffer_t * b0, u8 tcp_flags, int is_ip6)
-{
-  u16 l3_hdr_offset = vnet_buffer (b0)->l3_hdr_offset;
-  u16 l4_hdr_offset = vnet_buffer (b0)->l4_hdr_offset;
-  ip4_header_t *ip4 = (ip4_header_t *) (b0->data + l3_hdr_offset);
-  ip6_header_t *ip6 = (ip6_header_t *) (b0->data + l3_hdr_offset);
-  tcp_header_t *tcp = (tcp_header_t *) (b0->data + l4_hdr_offset);
-
-  tcp->flags = tcp_flags;
-
-  if (is_ip6)
-    ip6->payload_length =
-      clib_host_to_net_u16 (b0->current_length -
-			    (l4_hdr_offset - b0->current_data));
-  else
-    ip4->length =
-      clib_host_to_net_u16 (b0->current_length -
-			    (l3_hdr_offset - b0->current_data));
-}
-
-/**
- * Allocate the necessary number of ptd->split_buffers,
- * and segment the possibly chained buffer(s) from b0 into
- * there.
- *
- * Return the cumulative number of bytes sent or zero
- * if allocation failed.
- */
-
-static_always_inline u32
-tso_segment_buffer (vlib_main_t * vm, vnet_interface_per_thread_data_t * ptd,
-		    int do_tx_offloads, u32 sbi0, vlib_buffer_t * sb0,
-		    u32 n_bytes_b0)
-{
-  u32 n_tx_bytes = 0;
-  int is_ip4 = sb0->flags & VNET_BUFFER_F_IS_IP4;
-  int is_ip6 = sb0->flags & VNET_BUFFER_F_IS_IP6;
-  ASSERT (is_ip4 || is_ip6);
-  ASSERT (sb0->flags & VNET_BUFFER_F_L2_HDR_OFFSET_VALID);
-  ASSERT (sb0->flags & VNET_BUFFER_F_L3_HDR_OFFSET_VALID);
-  ASSERT (sb0->flags & VNET_BUFFER_F_L4_HDR_OFFSET_VALID);
-  u16 gso_size = vnet_buffer2 (sb0)->gso_size;
-
-  int l4_hdr_sz = vnet_buffer2 (sb0)->gso_l4_hdr_sz;
-  u8 save_tcp_flags = 0;
-  u8 tcp_flags_no_fin_psh = 0;
-  u32 next_tcp_seq = 0;
-
-  tcp_header_t *tcp =
-    (tcp_header_t *) (sb0->data + vnet_buffer (sb0)->l4_hdr_offset);
-  next_tcp_seq = clib_net_to_host_u32 (tcp->seq_number);
-  /* store original flags for last packet and reset FIN and PSH */
-  save_tcp_flags = tcp->flags;
-  tcp_flags_no_fin_psh = tcp->flags & ~(TCP_FLAG_FIN | TCP_FLAG_PSH);
-  tcp->checksum = 0;
-
-  u32 default_bflags =
-    sb0->flags & ~(VNET_BUFFER_F_GSO | VLIB_BUFFER_NEXT_PRESENT);
-  u16 l234_sz = vnet_buffer (sb0)->l4_hdr_offset + l4_hdr_sz
-    - sb0->current_data;
-  int first_data_size = clib_min (gso_size, sb0->current_length - l234_sz);
-  next_tcp_seq += first_data_size;
-
-  if (PREDICT_FALSE
-      (!tso_alloc_tx_bufs (vm, ptd, sb0, n_bytes_b0, l234_sz, gso_size)))
-    return 0;
-
-  vlib_buffer_t *b0 = vlib_get_buffer (vm, ptd->split_buffers[0]);
-  tso_init_buf_from_template_base (b0, sb0, default_bflags,
-				   l234_sz + first_data_size);
-
-  u32 total_src_left = n_bytes_b0 - l234_sz - first_data_size;
-  if (total_src_left)
-    {
-      /* Need to copy more segments */
-      u8 *src_ptr, *dst_ptr;
-      u16 src_left, dst_left;
-      /* current source buffer */
-      vlib_buffer_t *csb0 = sb0;
-      u32 csbi0 = sbi0;
-      /* current dest buffer */
-      vlib_buffer_t *cdb0;
-      u16 dbi = 1;		/* the buffer [0] is b0 */
-
-      src_ptr = vlib_buffer_get_current (sb0) + l234_sz + first_data_size;
-      src_left = sb0->current_length - l234_sz - first_data_size;
-
-      tso_fixup_segmented_buf (b0, tcp_flags_no_fin_psh, is_ip6);
-      if (do_tx_offloads)
-	calc_checksums (vm, b0);
-
-      /* grab a second buffer and prepare the loop */
-      ASSERT (dbi < vec_len (ptd->split_buffers));
-      cdb0 = vlib_get_buffer (vm, ptd->split_buffers[dbi++]);
-      tso_init_buf_from_template (vm, cdb0, b0, l234_sz, gso_size, &dst_ptr,
-				  &dst_left, next_tcp_seq, default_bflags);
-
-      /* an arbitrary large number to catch the runaway loops */
-      int nloops = 2000;
-      while (total_src_left)
-	{
-	  if (nloops-- <= 0)
-	    clib_panic ("infinite loop detected");
-	  u16 bytes_to_copy = clib_min (src_left, dst_left);
-
-	  clib_memcpy_fast (dst_ptr, src_ptr, bytes_to_copy);
-
-	  src_left -= bytes_to_copy;
-	  src_ptr += bytes_to_copy;
-	  total_src_left -= bytes_to_copy;
-	  dst_left -= bytes_to_copy;
-	  dst_ptr += bytes_to_copy;
-	  next_tcp_seq += bytes_to_copy;
-	  cdb0->current_length += bytes_to_copy;
-
-	  if (0 == src_left)
-	    {
-	      int has_next = (csb0->flags & VLIB_BUFFER_NEXT_PRESENT);
-	      u32 next_bi = csb0->next_buffer;
-
-	      /* init src to the next buffer in chain */
-	      if (has_next)
-		{
-		  csbi0 = next_bi;
-		  csb0 = vlib_get_buffer (vm, csbi0);
-		  src_left = csb0->current_length;
-		  src_ptr = vlib_buffer_get_current (csb0);
-		}
-	      else
-		{
-		  ASSERT (total_src_left == 0);
-		  break;
-		}
-	    }
-	  if (0 == dst_left && total_src_left)
-	    {
-	      if (do_tx_offloads)
-		calc_checksums (vm, cdb0);
-	      n_tx_bytes += cdb0->current_length;
-	      ASSERT (dbi < vec_len (ptd->split_buffers));
-	      cdb0 = vlib_get_buffer (vm, ptd->split_buffers[dbi++]);
-	      tso_init_buf_from_template (vm, cdb0, b0, l234_sz,
-					  gso_size, &dst_ptr, &dst_left,
-					  next_tcp_seq, default_bflags);
-	    }
-	}
-
-      tso_fixup_segmented_buf (cdb0, save_tcp_flags, is_ip6);
-      if (do_tx_offloads)
-	calc_checksums (vm, cdb0);
-
-      n_tx_bytes += cdb0->current_length;
-    }
-  n_tx_bytes += b0->current_length;
-  return n_tx_bytes;
-}
-
-static_always_inline void
-drop_one_buffer_and_count (vlib_main_t * vm, vnet_main_t * vnm,
-			   vlib_node_runtime_t * node, u32 * pbi0,
-			   u32 drop_error_code)
-{
-  u32 thread_index = vm->thread_index;
-  vnet_interface_output_runtime_t *rt = (void *) node->runtime_data;
-
-  vlib_simple_counter_main_t *cm;
-  cm =
-    vec_elt_at_index (vnm->interface_main.sw_if_counters,
-		      VNET_INTERFACE_COUNTER_TX_ERROR);
-  vlib_increment_simple_counter (cm, thread_index, rt->sw_if_index, 1);
-
-  vlib_error_drop_buffers (vm, node, pbi0,
-			   /* buffer stride */ 1,
-			   /* n_buffers */ 1,
-			   VNET_INTERFACE_OUTPUT_NEXT_DROP,
-			   node->node_index, drop_error_code);
-}
-
 static_always_inline uword
-vnet_interface_output_node_inline_gso (vlib_main_t * vm,
-				       vlib_node_runtime_t * node,
-				       vlib_frame_t * frame,
-				       vnet_main_t * vnm,
-				       vnet_hw_interface_t * hi,
-				       int do_tx_offloads,
-				       int do_segmentation)
+vnet_interface_output_node_inline (vlib_main_t * vm,
+				   vlib_node_runtime_t * node,
+				   vlib_frame_t * frame,
+				   vnet_main_t * vnm,
+				   vnet_hw_interface_t * hi,
+				   int do_tx_offloads)
 {
   vnet_interface_output_runtime_t *rt = (void *) node->runtime_data;
   vnet_sw_interface_t *si;
@@ -485,8 +229,6 @@ vnet_interface_output_node_inline_gso (vlib_main_t * vm,
   u32 next_index = VNET_INTERFACE_OUTPUT_NEXT_TX;
   u32 current_config_index = ~0;
   u8 arc = im->output_feature_arc_index;
-  vnet_interface_per_thread_data_t *ptd =
-    vec_elt_at_index (im->per_thread_data, thread_index);
   vlib_buffer_t *bufs[VLIB_FRAME_SIZE], **b = bufs;
 
   n_buffers = frame->n_vectors;
@@ -571,12 +313,6 @@ vnet_interface_output_node_inline_gso (vlib_main_t * vm,
 
 	  or_flags = b[0]->flags | b[1]->flags | b[2]->flags | b[3]->flags;
 
-	  if (do_segmentation)
-	    {
-	      /* go to single loop if we need TSO segmentation */
-	      if (PREDICT_FALSE (or_flags & VNET_BUFFER_F_GSO))
-		break;
-	    }
 	  from += 4;
 	  to_tx += 4;
 	  n_left_to_tx -= 4;
@@ -691,84 +427,6 @@ vnet_interface_output_node_inline_gso (vlib_main_t * vm,
 	      b[0]->current_config_index = current_config_index;
 	    }
 
-	  if (do_segmentation)
-	    {
-	      if (PREDICT_FALSE (b[0]->flags & VNET_BUFFER_F_GSO))
-		{
-		  /*
-		   * Undo the enqueue of the b0 - it is not going anywhere,
-		   * and will be freed either after it's segmented or
-		   * when dropped, if there is no buffers to segment into.
-		   */
-		  to_tx -= 1;
-		  n_left_to_tx += 1;
-		  /* undo the counting. */
-		  n_bytes -= n_bytes_b0;
-		  n_packets -= 1;
-
-		  u32 n_tx_bytes = 0;
-
-		  n_tx_bytes =
-		    tso_segment_buffer (vm, ptd, do_tx_offloads, bi0, b[0],
-					n_bytes_b0);
-
-		  if (PREDICT_FALSE (n_tx_bytes == 0))
-		    {
-		      drop_one_buffer_and_count (vm, vnm, node, from - 1,
-						 VNET_INTERFACE_OUTPUT_ERROR_NO_BUFFERS_FOR_GSO);
-		      b += 1;
-		      continue;
-		    }
-
-		  u16 n_tx_bufs = vec_len (ptd->split_buffers);
-		  u32 *from_tx_seg = ptd->split_buffers;
-
-		  while (n_tx_bufs > 0)
-		    {
-		      if (n_tx_bufs >= n_left_to_tx)
-			{
-			  while (n_left_to_tx > 0)
-			    {
-			      to_tx[0] = from_tx_seg[0];
-			      to_tx += 1;
-			      from_tx_seg += 1;
-			      n_left_to_tx -= 1;
-			      n_tx_bufs -= 1;
-			      n_packets += 1;
-			    }
-			  vlib_put_next_frame (vm, node, next_index,
-					       n_left_to_tx);
-			  vlib_get_new_next_frame (vm, node, next_index,
-						   to_tx, n_left_to_tx);
-			}
-		      while (n_tx_bufs > 0)
-			{
-			  to_tx[0] = from_tx_seg[0];
-			  to_tx += 1;
-			  from_tx_seg += 1;
-			  n_left_to_tx -= 1;
-			  n_tx_bufs -= 1;
-			  n_packets += 1;
-			}
-		    }
-		  n_bytes += n_tx_bytes;
-		  if (PREDICT_FALSE (tx_swif0 != rt->sw_if_index))
-		    {
-
-		      vlib_increment_combined_counter
-			(im->combined_sw_if_counters +
-			 VNET_INTERFACE_COUNTER_TX, thread_index, tx_swif0,
-			 _vec_len (ptd->split_buffers), n_tx_bytes);
-		    }
-		  /* The buffers were enqueued. Reset the length */
-		  _vec_len (ptd->split_buffers) = 0;
-		  /* Free the now segmented buffer */
-		  vlib_buffer_free_one (vm, bi0);
-		  b += 1;
-		  continue;
-		}
-	    }
-
 	  if (PREDICT_FALSE (tx_swif0 != rt->sw_if_index))
 	    {
 
@@ -779,8 +437,13 @@ vnet_interface_output_node_inline_gso (vlib_main_t * vm,
 	    }
 
 	  if (do_tx_offloads)
-	    calc_checksums (vm, b[0]);
-
+	    {
+	      if (b[0]->flags &
+		  (VNET_BUFFER_F_OFFLOAD_TCP_CKSUM |
+		   VNET_BUFFER_F_OFFLOAD_UDP_CKSUM |
+		   VNET_BUFFER_F_OFFLOAD_IP_CKSUM))
+		calc_checksums (vm, b[0]);
+	    }
 	  b += 1;
 	}
 
@@ -845,32 +508,6 @@ static_always_inline void vnet_interface_pcap_tx_trace
 }
 
 #ifndef CLIB_MARCH_VARIANT
-static_always_inline uword
-vnet_interface_output_node_inline (vlib_main_t * vm,
-				   vlib_node_runtime_t * node,
-				   vlib_frame_t * frame, vnet_main_t * vnm,
-				   vnet_hw_interface_t * hi,
-				   int do_tx_offloads)
-{
-  /*
-   * The 3-headed "if" is here because we want to err on the side
-   * of not impacting the non-GSO performance - so for the more
-   * common case of no GSO interfaces we want to prevent the
-   * segmentation codepath from being there altogether.
-   */
-  if (PREDICT_TRUE (vnm->interface_main.gso_interface_count == 0))
-    return vnet_interface_output_node_inline_gso (vm, node, frame, vnm, hi,
-						  do_tx_offloads,
-						  /* do_segmentation */ 0);
-  else if (hi->flags & VNET_HW_INTERFACE_FLAG_SUPPORTS_GSO)
-    return vnet_interface_output_node_inline_gso (vm, node, frame, vnm, hi,
-						  do_tx_offloads,
-						  /* do_segmentation */ 0);
-  else
-    return vnet_interface_output_node_inline_gso (vm, node, frame, vnm, hi,
-						  do_tx_offloads,
-						  /* do_segmentation */ 1);
-}
 
 uword
 vnet_interface_output_node (vlib_main_t * vm, vlib_node_runtime_t * node,
diff --git a/src/vnet/vnet_all_api_h.h b/src/vnet/vnet_all_api_h.h
index 519f5219042..7395d1c4756 100644
--- a/src/vnet/vnet_all_api_h.h
+++ b/src/vnet/vnet_all_api_h.h
@@ -74,6 +74,7 @@
 #include <vnet/vxlan-gbp/vxlan_gbp.api.h>
 #include <vnet/syslog/syslog.api.h>
 #include <vnet/devices/virtio/virtio.api.h>
+#include <vnet/gso/gso.api.h>
 
 /*
  * fd.io coding-style-patch-verification: ON