Raw export of IP6 packet with iOAM metadata

- is enabled on iOAM pop nodes with "set ioam export ipfix collector"
- ioam_export_node Hooks into vlib graph b/n ip6-hop-by-hop node and ip6-pop-hop-by-hop node
- A buffer per worker thread is created for collecting packet data to be exported
- ioam_export_node exports first 3 cachelines by collecting it in a MTU sized frame, slaps on ipfix header for export
- ioam_export_thread process node - checks for unsent record buffers for longer than 20 seconds and exports it
- Added dual loop and prefetch in add, hop-by-hop and pop functions

To be done:
- IPfix template
- Multi collector distribution of ipfix packets
- Port to be configurable

Change-Id: I959b4253036551382562bdaf10a83fd6f2f1c88b
Signed-off-by: Shwetha <shwethab@cisco.com>

1 files changed, 352 insertions, 0 deletions

diff --git a/plugins/ioam-plugin/ioam/export/node.c b/plugins/ioam-plugin/ioam/export/node.c
new file mode 100644
index 00000000000..484bcb5d265
--- /dev/null
+++ b/plugins/ioam-plugin/ioam/export/node.c
@@ -0,0 +1,352 @@
+/*
+ * Copyright (c) 2016 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 <vnet/pg/pg.h>
+#include <vppinfra/error.h>
+#include <vnet/ip/ip.h>
+#include <ioam/export/ioam_export.h>
+
+typedef struct
+{
+  u32 next_index;
+  u32 flow_label;
+} export_trace_t;
+
+/* packet trace format function */
+static u8 *
+format_export_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 *);
+  export_trace_t *t = va_arg (*args, export_trace_t *);
+
+  s = format (s, "EXPORT: flow_label %d, next index %d",
+	      t->flow_label, t->next_index);
+  return s;
+}
+
+vlib_node_registration_t export_node;
+
+#define foreach_export_error \
+_(RECORDED, "Packets recorded for export")
+
+typedef enum
+{
+#define _(sym,str) EXPORT_ERROR_##sym,
+  foreach_export_error
+#undef _
+    EXPORT_N_ERROR,
+} export_error_t;
+
+static char *export_error_strings[] = {
+#define _(sym,string) string,
+  foreach_export_error
+#undef _
+};
+
+typedef enum
+{
+  EXPORT_NEXT_POP_HBYH,
+  EXPORT_N_NEXT,
+} export_next_t;
+
+always_inline void
+copy3cachelines (void *dst, const void *src, size_t n)
+{
+#if 0
+  if (PREDICT_FALSE (n < DEFAULT_EXPORT_SIZE))
+    {
+      /* Copy only the first 1/2 cache lines whatever is available */
+      if (n >= 64)
+	clib_mov64 ((u8 *) dst, (const u8 *) src);
+      if (n >= 128)
+	clib_mov64 ((u8 *) dst + 64, (const u8 *) src + 64);
+      return;
+    }
+  clib_mov64 ((u8 *) dst, (const u8 *) src);
+  clib_mov64 ((u8 *) dst + 64, (const u8 *) src + 64);
+  clib_mov64 ((u8 *) dst + 128, (const u8 *) src + 128);
+#endif
+#if 1
+
+  u64 *copy_dst, *copy_src;
+  int i;
+  copy_dst = (u64 *) dst;
+  copy_src = (u64 *) src;
+  if (PREDICT_FALSE (n < DEFAULT_EXPORT_SIZE))
+    {
+      for (i = 0; i < n / 64; i++)
+	{
+	  copy_dst[0] = copy_src[0];
+	  copy_dst[1] = copy_src[1];
+	  copy_dst[2] = copy_src[2];
+	  copy_dst[3] = copy_src[3];
+	  copy_dst[4] = copy_src[4];
+	  copy_dst[5] = copy_src[5];
+	  copy_dst[6] = copy_src[6];
+	  copy_dst[7] = copy_src[7];
+	  copy_dst += 8;
+	  copy_src += 8;
+	}
+      return;
+    }
+  for (i = 0; i < 3; i++)
+    {
+      copy_dst[0] = copy_src[0];
+      copy_dst[1] = copy_src[1];
+      copy_dst[2] = copy_src[2];
+      copy_dst[3] = copy_src[3];
+      copy_dst[4] = copy_src[4];
+      copy_dst[5] = copy_src[5];
+      copy_dst[6] = copy_src[6];
+      copy_dst[7] = copy_src[7];
+      copy_dst += 8;
+      copy_src += 8;
+    }
+#endif
+}
+
+static uword
+ip6_export_node_fn (vlib_main_t * vm,
+		    vlib_node_runtime_t * node, vlib_frame_t * frame)
+{
+  ioam_export_main_t *em = &ioam_export_main;
+  u32 n_left_from, *from, *to_next;
+  export_next_t next_index;
+  u32 pkts_recorded = 0;
+  ioam_export_buffer_t *my_buf = 0;
+  vlib_buffer_t *eb0 = 0;
+  u32 ebi0 = 0;
+  from = vlib_frame_vector_args (frame);
+  n_left_from = frame->n_vectors;
+  next_index = node->cached_next_index;
+
+  while (__sync_lock_test_and_set (em->lockp[vm->cpu_index], 1))
+    ;
+  my_buf = ioam_export_get_my_buffer (vm->cpu_index);
+  my_buf->touched_at = vlib_time_now (vm);
+  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 next0 = EXPORT_NEXT_POP_HBYH;
+	  u32 next1 = EXPORT_NEXT_POP_HBYH;
+	  u32 bi0, bi1;
+	  ip6_header_t *ip60, *ip61;
+	  vlib_buffer_t *p0, *p1;
+	  u32 ip_len0, ip_len1;
+
+	  /* 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);
+
+	    /* IPv6 + HbyH header + Trace option */
+	    /* 40   +           2 + [4 hdr] + [16]* no_of_nodes */
+	    /* 3 cache lines can get v6 hdr + trace option with upto 9 node trace */
+	    CLIB_PREFETCH (p2->data, 3 * CLIB_CACHE_LINE_BYTES, LOAD);
+	    CLIB_PREFETCH (p3->data, 3 * CLIB_CACHE_LINE_BYTES, LOAD);
+	  }
+
+	  /* speculatively enqueue p0 and p1 to the current next frame */
+	  to_next[0] = bi0 = from[0];
+	  to_next[1] = bi1 = from[1];
+	  from += 2;
+	  to_next += 2;
+	  n_left_from -= 2;
+	  n_left_to_next -= 2;
+
+	  p0 = vlib_get_buffer (vm, bi0);
+	  p1 = vlib_get_buffer (vm, bi1);
+
+	  ip60 = vlib_buffer_get_current (p0);
+	  ip61 = vlib_buffer_get_current (p1);
+
+	  ip_len0 =
+	    clib_net_to_host_u16 (ip60->payload_length) +
+	    sizeof (ip6_header_t);
+	  ip_len1 =
+	    clib_net_to_host_u16 (ip61->payload_length) +
+	    sizeof (ip6_header_t);
+
+	  ebi0 = my_buf->buffer_index;
+	  eb0 = vlib_get_buffer (vm, ebi0);
+	  if (PREDICT_FALSE (eb0 == 0))
+	    goto NO_BUFFER1;
+
+	  ip_len0 =
+	    ip_len0 > DEFAULT_EXPORT_SIZE ? DEFAULT_EXPORT_SIZE : ip_len0;
+	  ip_len1 =
+	    ip_len1 > DEFAULT_EXPORT_SIZE ? DEFAULT_EXPORT_SIZE : ip_len1;
+
+	  copy3cachelines (eb0->data + eb0->current_length, ip60, ip_len0);
+	  eb0->current_length += DEFAULT_EXPORT_SIZE;
+	  /* To maintain uniform size per export, each
+	   * record is default size, ip6 hdr can be
+	   * used to parse the record correctly
+	   */
+	  my_buf->records_in_this_buffer++;
+	  /* if number of buf exceeds max that fits in a MTU sized buffer
+	   * ship it to the queue and pick new one
+	   */
+	  if (my_buf->records_in_this_buffer >= DEFAULT_EXPORT_RECORDS)
+	    {
+	      ioam_export_send_buffer (vm, my_buf);
+	      ioam_export_init_buffer (vm, my_buf);
+	    }
+
+	  ebi0 = my_buf->buffer_index;
+	  eb0 = vlib_get_buffer (vm, ebi0);
+	  if (PREDICT_FALSE (eb0 == 0))
+	    goto NO_BUFFER1;
+
+	  copy3cachelines (eb0->data + eb0->current_length, ip61, ip_len1);
+	  eb0->current_length += DEFAULT_EXPORT_SIZE;
+	  my_buf->records_in_this_buffer++;
+	  if (my_buf->records_in_this_buffer >= DEFAULT_EXPORT_RECORDS)
+	    {
+	      ioam_export_send_buffer (vm, my_buf);
+	      ioam_export_init_buffer (vm, my_buf);
+	    }
+
+	  pkts_recorded += 2;
+
+	  if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE)))
+	    {
+	      if (p0->flags & VLIB_BUFFER_IS_TRACED)
+		{
+		  export_trace_t *t =
+		    vlib_add_trace (vm, node, p0, sizeof (*t));
+		  t->flow_label =
+		    clib_net_to_host_u32 (ip60->
+					  ip_version_traffic_class_and_flow_label);
+		  t->next_index = next0;
+		}
+	      if (p1->flags & VLIB_BUFFER_IS_TRACED)
+		{
+		  export_trace_t *t =
+		    vlib_add_trace (vm, node, p1, sizeof (*t));
+		  t->flow_label =
+		    clib_net_to_host_u32 (ip61->
+					  ip_version_traffic_class_and_flow_label);
+		  t->next_index = next1;
+		}
+	    }
+	NO_BUFFER1:
+	  /* verify speculative enqueues, maybe switch current next frame */
+	  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 *p0;
+	  u32 next0 = EXPORT_NEXT_POP_HBYH;
+	  ip6_header_t *ip60;
+	  u32 ip_len0;
+
+	  /* speculatively enqueue p0 to the current next frame */
+	  bi0 = from[0];
+	  to_next[0] = bi0;
+	  from += 1;
+	  to_next += 1;
+	  n_left_from -= 1;
+	  n_left_to_next -= 1;
+
+	  p0 = vlib_get_buffer (vm, bi0);
+	  ip60 = vlib_buffer_get_current (p0);
+	  ip_len0 =
+	    clib_net_to_host_u16 (ip60->payload_length) +
+	    sizeof (ip6_header_t);
+
+	  ebi0 = my_buf->buffer_index;
+	  eb0 = vlib_get_buffer (vm, ebi0);
+	  if (PREDICT_FALSE (eb0 == 0))
+	    goto NO_BUFFER;
+
+	  ip_len0 =
+	    ip_len0 > DEFAULT_EXPORT_SIZE ? DEFAULT_EXPORT_SIZE : ip_len0;
+	  copy3cachelines (eb0->data + eb0->current_length, ip60, ip_len0);
+	  eb0->current_length += DEFAULT_EXPORT_SIZE;
+	  /* To maintain uniform size per export, each
+	   * record is default size, ip6 hdr can be
+	   * used to parse the record correctly
+	   */
+	  my_buf->records_in_this_buffer++;
+	  /* if number of buf exceeds max that fits in a MTU sized buffer
+	   * ship it to the queue and pick new one
+	   */
+	  if (my_buf->records_in_this_buffer >= DEFAULT_EXPORT_RECORDS)
+	    {
+	      ioam_export_send_buffer (vm, my_buf);
+	      ioam_export_init_buffer (vm, my_buf);
+	    }
+	  if (PREDICT_FALSE ((node->flags & VLIB_NODE_FLAG_TRACE)
+			     && (p0->flags & VLIB_BUFFER_IS_TRACED)))
+	    {
+	      export_trace_t *t = vlib_add_trace (vm, node, p0, sizeof (*t));
+	      t->flow_label =
+		clib_net_to_host_u32 (ip60->
+				      ip_version_traffic_class_and_flow_label);
+	      t->next_index = next0;
+	    }
+
+	  pkts_recorded += 1;
+	NO_BUFFER:
+	  /* verify speculative enqueue, maybe switch current next frame */
+	  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);
+    }
+
+  vlib_node_increment_counter (vm, export_node.index,
+			       EXPORT_ERROR_RECORDED, pkts_recorded);
+  *em->lockp[vm->cpu_index] = 0;
+  return frame->n_vectors;
+}
+
+/*
+ * Node for IP6 export
+ */
+VLIB_REGISTER_NODE (export_node) =
+{
+  .function = ip6_export_node_fn,
+  .name = "ip6-export",
+  .vector_size = sizeof (u32),
+  .format_trace = format_export_trace,
+  .type = VLIB_NODE_TYPE_INTERNAL,
+  .n_errors = ARRAY_LEN (export_error_strings),
+  .error_strings = export_error_strings,
+  .n_next_nodes = EXPORT_N_NEXT,
+  /* edit / add dispositions here */
+  .next_nodes =
+  {
+    [EXPORT_NEXT_POP_HBYH] = "ip6-pop-hop-by-hop"
+  },
+};