diff --git a/src/vlib/buffer_funcs.c b/src/vlib/buffer_funcs.c
index 4ad652b..4e1f2fd 100644
--- a/src/vlib/buffer_funcs.c
+++ b/src/vlib/buffer_funcs.c
@@ -10,26 +10,38 @@
 static_always_inline u32
 enqueue_one (vlib_main_t *vm, vlib_node_runtime_t *node,
 	     vlib_frame_bitmap_t used_elt_bmp, u16 next_index, u32 *buffers,
-	     u16 *nexts, u32 n_buffers, u32 n_left, u32 *tmp)
+	     u16 *nexts, u32 n_buffers, u32 n_left, u32 *tmp, u8 maybe_aux,
+	     u32 *aux_data, u32 *tmp_aux)
 {
   vlib_frame_bitmap_t match_bmp;
   vlib_frame_t *f;
   u32 n_extracted, n_free;
-  u32 *to;
+  u32 *to, *to_aux;
 
   f = vlib_get_next_frame_internal (vm, node, next_index, 0);
 
+  maybe_aux = maybe_aux && f->aux_offset;
+
   n_free = VLIB_FRAME_SIZE - f->n_vectors;
 
   /* if frame contains enough space for worst case scenario, we can avoid
    * use of tmp */
   if (n_free >= n_left)
-    to = (u32 *) vlib_frame_vector_args (f) + f->n_vectors;
+    {
+      to = (u32 *) vlib_frame_vector_args (f) + f->n_vectors;
+      if (maybe_aux)
+	to_aux = (u32 *) vlib_frame_aux_args (f) + f->n_vectors;
+    }
   else
-    to = tmp;
-
+    {
+      to = tmp;
+      if (maybe_aux)
+	to_aux = tmp_aux;
+    }
   clib_mask_compare_u16 (next_index, nexts, match_bmp, n_buffers);
   n_extracted = clib_compress_u32 (to, buffers, match_bmp, n_buffers);
+  if (maybe_aux)
+    clib_compress_u32 (to_aux, aux_data, match_bmp, n_buffers);
   vlib_frame_bitmap_or (used_elt_bmp, match_bmp);
 
   if (to != tmp)
@@ -42,6 +54,11 @@
       /* enough space in the existing frame */
       to = (u32 *) vlib_frame_vector_args (f) + f->n_vectors;
       vlib_buffer_copy_indices (to, tmp, n_extracted);
+      if (maybe_aux)
+	{
+	  to_aux = (u32 *) vlib_frame_aux_args (f) + f->n_vectors;
+	  vlib_buffer_copy_indices (to_aux, tmp_aux, n_extracted);
+	}
       vlib_put_next_frame (vm, node, next_index, n_free - n_extracted);
     }
   else
@@ -49,6 +66,11 @@
       /* full frame */
       to = (u32 *) vlib_frame_vector_args (f) + f->n_vectors;
       vlib_buffer_copy_indices (to, tmp, n_free);
+      if (maybe_aux)
+	{
+	  to_aux = (u32 *) vlib_frame_aux_args (f) + f->n_vectors;
+	  vlib_buffer_copy_indices (to_aux, tmp_aux, n_free);
+	}
       vlib_put_next_frame (vm, node, next_index, 0);
 
       /* second frame */
@@ -56,6 +78,11 @@
       f = vlib_get_next_frame_internal (vm, node, next_index, 1);
       to = vlib_frame_vector_args (f);
       vlib_buffer_copy_indices (to, tmp + n_free, n_2nd_frame);
+      if (maybe_aux)
+	{
+	  to_aux = vlib_frame_aux_args (f);
+	  vlib_buffer_copy_indices (to_aux, tmp_aux + n_free, n_2nd_frame);
+	}
       vlib_put_next_frame (vm, node, next_index,
 			   VLIB_FRAME_SIZE - n_2nd_frame);
     }
@@ -63,12 +90,14 @@
   return n_left - n_extracted;
 }
 
-void __clib_section (".vlib_buffer_enqueue_to_next_fn")
-CLIB_MULTIARCH_FN (vlib_buffer_enqueue_to_next_fn)
-(vlib_main_t *vm, vlib_node_runtime_t *node, u32 *buffers, u16 *nexts,
- uword count)
+static_always_inline void
+vlib_buffer_enqueue_to_next_fn_inline (vlib_main_t *vm,
+				       vlib_node_runtime_t *node, u32 *buffers,
+				       u32 *aux_data, u16 *nexts, uword count,
+				       u8 maybe_aux)
 {
   u32 tmp[VLIB_FRAME_SIZE];
+  u32 tmp_aux[VLIB_FRAME_SIZE];
   u32 n_left;
   u16 next_index;
 
@@ -80,7 +109,8 @@
 
       next_index = nexts[0];
       n_left = enqueue_one (vm, node, used_elt_bmp, next_index, buffers, nexts,
-			    VLIB_FRAME_SIZE, n_left, tmp);
+			    VLIB_FRAME_SIZE, n_left, tmp, maybe_aux, aux_data,
+			    tmp_aux);
 
       while (n_left)
 	{
@@ -93,10 +123,13 @@
 	  next_index =
 	    nexts[off * 64 + count_trailing_zeros (~used_elt_bmp[off])];
 	  n_left = enqueue_one (vm, node, used_elt_bmp, next_index, buffers,
-				nexts, VLIB_FRAME_SIZE, n_left, tmp);
+				nexts, VLIB_FRAME_SIZE, n_left, tmp, maybe_aux,
+				aux_data, tmp_aux);
 	}
 
       buffers += VLIB_FRAME_SIZE;
+      if (maybe_aux)
+	aux_data += VLIB_FRAME_SIZE;
       nexts += VLIB_FRAME_SIZE;
       count -= VLIB_FRAME_SIZE;
     }
@@ -109,7 +142,7 @@
       u32 off = 0;
 
       n_left = enqueue_one (vm, node, used_elt_bmp, next_index, buffers, nexts,
-			    count, n_left, tmp);
+			    count, n_left, tmp, maybe_aux, aux_data, tmp_aux);
 
       while (n_left)
 	{
@@ -121,26 +154,55 @@
 
 	  next_index =
 	    nexts[off * 64 + count_trailing_zeros (~used_elt_bmp[off])];
-	  n_left = enqueue_one (vm, node, used_elt_bmp, next_index, buffers,
-				nexts, count, n_left, tmp);
+	  n_left =
+	    enqueue_one (vm, node, used_elt_bmp, next_index, buffers, nexts,
+			 count, n_left, tmp, maybe_aux, aux_data, tmp_aux);
 	}
     }
 }
 
+void __clib_section (".vlib_buffer_enqueue_to_next_fn")
+CLIB_MULTIARCH_FN (vlib_buffer_enqueue_to_next_fn)
+(vlib_main_t *vm, vlib_node_runtime_t *node, u32 *buffers, u16 *nexts,
+ uword count)
+{
+  vlib_buffer_enqueue_to_next_fn_inline (vm, node, buffers, NULL, nexts, count,
+					 0 /* maybe_aux */);
+}
+
 CLIB_MARCH_FN_REGISTRATION (vlib_buffer_enqueue_to_next_fn);
 
-void __clib_section (".vlib_buffer_enqueue_to_single_next_fn")
-CLIB_MULTIARCH_FN (vlib_buffer_enqueue_to_single_next_fn)
-(vlib_main_t *vm, vlib_node_runtime_t *node, u32 *buffers, u16 next_index,
- u32 count)
+void __clib_section (".vlib_buffer_enqueue_to_next_with_aux_fn")
+CLIB_MULTIARCH_FN (vlib_buffer_enqueue_to_next_with_aux_fn)
+(vlib_main_t *vm, vlib_node_runtime_t *node, u32 *buffers, u32 *aux_data,
+ u16 *nexts, uword count)
 {
-  u32 *to_next, n_left_to_next, n_enq;
+  vlib_buffer_enqueue_to_next_fn_inline (vm, node, buffers, aux_data, nexts,
+					 count, 1 /* maybe_aux */);
+}
 
-  vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+CLIB_MARCH_FN_REGISTRATION (vlib_buffer_enqueue_to_next_with_aux_fn);
+
+static_always_inline void
+vlib_buffer_enqueue_to_single_next_fn_inline (vlib_main_t *vm,
+					      vlib_node_runtime_t *node,
+					      u32 *buffers, u32 *aux_data,
+					      u16 next_index, u32 count,
+					      u8 with_aux)
+{
+  u32 *to_next, *to_next_aux, n_left_to_next, n_enq;
+
+  if (with_aux)
+    vlib_get_next_frame_with_aux (vm, node, next_index, to_next, to_next_aux,
+				  n_left_to_next);
+  else
+    vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
 
   if (PREDICT_TRUE (n_left_to_next >= count))
     {
       vlib_buffer_copy_indices (to_next, buffers, count);
+      if (with_aux)
+	vlib_buffer_copy_indices (to_next_aux, aux_data, count);
       n_left_to_next -= count;
       vlib_put_next_frame (vm, node, next_index, n_left_to_next);
       return;
@@ -149,22 +211,49 @@
   n_enq = n_left_to_next;
 next:
   vlib_buffer_copy_indices (to_next, buffers, n_enq);
+  if (with_aux)
+    vlib_buffer_copy_indices (to_next_aux, aux_data, n_enq);
   n_left_to_next -= n_enq;
 
   if (PREDICT_FALSE (count > n_enq))
     {
       count -= n_enq;
       buffers += n_enq;
+      if (with_aux)
+	aux_data += n_enq;
 
       vlib_put_next_frame (vm, node, next_index, n_left_to_next);
-      vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
+      if (with_aux)
+	vlib_get_next_frame_with_aux (vm, node, next_index, to_next,
+				      to_next_aux, n_left_to_next);
+      else
+	vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next);
       n_enq = clib_min (n_left_to_next, count);
       goto next;
     }
   vlib_put_next_frame (vm, node, next_index, n_left_to_next);
 }
+
+void __clib_section (".vlib_buffer_enqueue_to_single_next_fn")
+CLIB_MULTIARCH_FN (vlib_buffer_enqueue_to_single_next_fn)
+(vlib_main_t *vm, vlib_node_runtime_t *node, u32 *buffers, u16 next_index,
+ u32 count)
+{
+  vlib_buffer_enqueue_to_single_next_fn_inline (
+    vm, node, buffers, NULL, next_index, count, 0 /* with_aux */);
+}
 CLIB_MARCH_FN_REGISTRATION (vlib_buffer_enqueue_to_single_next_fn);
 
+void __clib_section (".vlib_buffer_enqueue_to_single_next_with_aux_fn")
+CLIB_MULTIARCH_FN (vlib_buffer_enqueue_to_single_next_with_aux_fn)
+(vlib_main_t *vm, vlib_node_runtime_t *node, u32 *buffers, u32 *aux_data,
+ u16 next_index, u32 count)
+{
+  vlib_buffer_enqueue_to_single_next_fn_inline (
+    vm, node, buffers, aux_data, next_index, count, 1 /* with_aux */);
+}
+CLIB_MARCH_FN_REGISTRATION (vlib_buffer_enqueue_to_single_next_with_aux_fn);
+
 static inline vlib_frame_queue_elt_t *
 vlib_get_frame_queue_elt (vlib_frame_queue_main_t *fqm, u32 index,
 			  int dont_wait)
@@ -482,8 +571,12 @@
   vlib_buffer_func_main_t *bfm = &vlib_buffer_func_main;
   bfm->buffer_enqueue_to_next_fn =
     CLIB_MARCH_FN_POINTER (vlib_buffer_enqueue_to_next_fn);
+  bfm->buffer_enqueue_to_next_with_aux_fn =
+    CLIB_MARCH_FN_POINTER (vlib_buffer_enqueue_to_next_with_aux_fn);
   bfm->buffer_enqueue_to_single_next_fn =
     CLIB_MARCH_FN_POINTER (vlib_buffer_enqueue_to_single_next_fn);
+  bfm->buffer_enqueue_to_single_next_with_aux_fn =
+    CLIB_MARCH_FN_POINTER (vlib_buffer_enqueue_to_single_next_with_aux_fn);
   bfm->buffer_enqueue_to_thread_fn =
     CLIB_MARCH_FN_POINTER (vlib_buffer_enqueue_to_thread_fn);
   bfm->buffer_enqueue_to_thread_with_aux_fn =
diff --git a/src/vlib/buffer_funcs.h b/src/vlib/buffer_funcs.h
index 00dce80..7085399 100644
--- a/src/vlib/buffer_funcs.h
+++ b/src/vlib/buffer_funcs.h
@@ -56,10 +56,17 @@
 						 vlib_node_runtime_t *node,
 						 u32 *buffers, u16 *nexts,
 						 uword count);
+typedef void (vlib_buffer_enqueue_to_next_with_aux_fn_t) (
+  vlib_main_t *vm, vlib_node_runtime_t *node, u32 *buffers, u32 *aux_data,
+  u16 *nexts, uword count);
 typedef void (vlib_buffer_enqueue_to_single_next_fn_t) (
   vlib_main_t *vm, vlib_node_runtime_t *node, u32 *ers, u16 next_index,
   u32 count);
 
+typedef void (vlib_buffer_enqueue_to_single_next_with_aux_fn_t) (
+  vlib_main_t *vm, vlib_node_runtime_t *node, u32 *ers, u32 *aux_data,
+  u16 next_index, u32 count);
+
 typedef u32 (vlib_buffer_enqueue_to_thread_fn_t) (
   vlib_main_t *vm, vlib_node_runtime_t *node, u32 frame_queue_index,
   u32 *buffer_indices, u16 *thread_indices, u32 n_packets,
@@ -73,7 +80,11 @@
 typedef struct
 {
   vlib_buffer_enqueue_to_next_fn_t *buffer_enqueue_to_next_fn;
+  vlib_buffer_enqueue_to_next_with_aux_fn_t
+    *buffer_enqueue_to_next_with_aux_fn;
   vlib_buffer_enqueue_to_single_next_fn_t *buffer_enqueue_to_single_next_fn;
+  vlib_buffer_enqueue_to_single_next_with_aux_fn_t
+    *buffer_enqueue_to_single_next_with_aux_fn;
   vlib_buffer_enqueue_to_thread_fn_t *buffer_enqueue_to_thread_fn;
   vlib_buffer_enqueue_to_thread_with_aux_fn_t
     *buffer_enqueue_to_thread_with_aux_fn;
diff --git a/src/vlib/buffer_node.h b/src/vlib/buffer_node.h
index a4c259f..5bc547a 100644
--- a/src/vlib/buffer_node.h
+++ b/src/vlib/buffer_node.h
@@ -356,6 +356,16 @@
 }
 
 static_always_inline void
+vlib_buffer_enqueue_to_next_with_aux (vlib_main_t *vm,
+				      vlib_node_runtime_t *node, u32 *buffers,
+				      u32 *aux_data, u16 *nexts, uword count)
+{
+  vlib_buffer_enqueue_to_next_with_aux_fn_t *fn;
+  fn = vlib_buffer_func_main.buffer_enqueue_to_next_with_aux_fn;
+  (fn) (vm, node, buffers, aux_data, nexts, count);
+}
+
+static_always_inline void
 vlib_buffer_enqueue_to_next_vec (vlib_main_t *vm, vlib_node_runtime_t *node,
 				 u32 **buffers, u16 **nexts, uword count)
 {
@@ -379,6 +389,17 @@
   (fn) (vm, node, buffers, next_index, count);
 }
 
+static_always_inline void
+vlib_buffer_enqueue_to_single_next_with_aux (vlib_main_t *vm,
+					     vlib_node_runtime_t *node,
+					     u32 *buffers, u32 *aux_data,
+					     u16 next_index, u32 count)
+{
+  vlib_buffer_enqueue_to_single_next_with_aux_fn_t *fn;
+  fn = vlib_buffer_func_main.buffer_enqueue_to_single_next_with_aux_fn;
+  (fn) (vm, node, buffers, aux_data, next_index, count);
+}
+
 static_always_inline u32
 vlib_buffer_enqueue_to_thread (vlib_main_t *vm, vlib_node_runtime_t *node,
 			       u32 frame_queue_index, u32 *buffer_indices,
diff --git a/src/vlib/node_funcs.h b/src/vlib/node_funcs.h
index 5b4d2a4..62a1fec 100644
--- a/src/vlib/node_funcs.h
+++ b/src/vlib/node_funcs.h
@@ -372,16 +372,34 @@
 					    u32 next_index,
 					    u32 alloc_new_frame);
 
-#define vlib_get_next_frame_macro(vm,node,next_index,vectors,n_vectors_left,alloc_new_frame) \
-do {									\
-  vlib_frame_t * _f							\
-    = vlib_get_next_frame_internal ((vm), (node), (next_index),		\
-				    (alloc_new_frame));			\
-  u32 _n = _f->n_vectors;						\
-  (vectors) = vlib_frame_vector_args (_f) + _n * sizeof ((vectors)[0]); \
-  (n_vectors_left) = VLIB_FRAME_SIZE - _n;				\
-} while (0)
+#define vlib_get_next_frame_macro(vm, node, next_index, vectors,              \
+				  n_vectors_left, alloc_new_frame)            \
+  do                                                                          \
+    {                                                                         \
+      vlib_frame_t *_f = vlib_get_next_frame_internal (                       \
+	(vm), (node), (next_index), (alloc_new_frame));                       \
+      u32 _n = _f->n_vectors;                                                 \
+      (vectors) = vlib_frame_vector_args (_f) + _n * sizeof ((vectors)[0]);   \
+      (n_vectors_left) = VLIB_FRAME_SIZE - _n;                                \
+    }                                                                         \
+  while (0)
 
+#define vlib_get_next_frame_macro_with_aux(vm, node, next_index, vectors,     \
+					   n_vectors_left, alloc_new_frame,   \
+					   aux_data, maybe_no_aux)            \
+  do                                                                          \
+    {                                                                         \
+      vlib_frame_t *_f = vlib_get_next_frame_internal (                       \
+	(vm), (node), (next_index), (alloc_new_frame));                       \
+      u32 _n = _f->n_vectors;                                                 \
+      (vectors) = vlib_frame_vector_args (_f) + _n * sizeof ((vectors)[0]);   \
+      if ((maybe_no_aux) && (_f)->aux_offset == 0)                            \
+	(aux_data) = NULL;                                                    \
+      else                                                                    \
+	(aux_data) = vlib_frame_aux_args (_f) + _n * sizeof ((aux_data)[0]);  \
+      (n_vectors_left) = VLIB_FRAME_SIZE - _n;                                \
+    }                                                                         \
+  while (0)
 
 /** \brief Get pointer to next frame vector data by
     (@c vlib_node_runtime_t, @c next_index).
@@ -395,16 +413,69 @@
  @return @c vectors -- pointer to next available vector slot
  @return @c n_vectors_left -- number of vector slots available
 */
-#define vlib_get_next_frame(vm,node,next_index,vectors,n_vectors_left)	\
-  vlib_get_next_frame_macro (vm, node, next_index,			\
-			     vectors, n_vectors_left,			\
+#define vlib_get_next_frame(vm, node, next_index, vectors, n_vectors_left)    \
+  vlib_get_next_frame_macro (vm, node, next_index, vectors, n_vectors_left,   \
 			     /* alloc new frame */ 0)
 
-#define vlib_get_new_next_frame(vm,node,next_index,vectors,n_vectors_left) \
-  vlib_get_next_frame_macro (vm, node, next_index,			\
-			     vectors, n_vectors_left,			\
+#define vlib_get_new_next_frame(vm, node, next_index, vectors,                \
+				n_vectors_left)                               \
+  vlib_get_next_frame_macro (vm, node, next_index, vectors, n_vectors_left,   \
 			     /* alloc new frame */ 1)
 
+/** \brief Get pointer to next frame vector data and next frame aux data by
+    (@c vlib_node_runtime_t, @c next_index).
+ Standard single/dual loop boilerplate element.
+ @attention This is a MACRO, with SIDE EFFECTS.
+ @attention This MACRO is unsafe in case the next node does not support
+ aux_data
+
+ @param vm vlib_main_t pointer, varies by thread
+ @param node current node vlib_node_runtime_t pointer
+ @param next_index requested graph arc index
+
+ @return @c vectors -- pointer to next available vector slot
+ @return @c aux_data -- pointer to next available aux data slot
+ @return @c n_vectors_left -- number of vector slots available
+*/
+#define vlib_get_next_frame_with_aux(vm, node, next_index, vectors, aux_data, \
+				     n_vectors_left)                          \
+  vlib_get_next_frame_macro_with_aux (                                        \
+    vm, node, next_index, vectors, n_vectors_left, /* alloc new frame */ 0,   \
+    aux_data, /* maybe_no_aux */ 0)
+
+#define vlib_get_new_next_frame_with_aux(vm, node, next_index, vectors,       \
+					 aux_data, n_vectors_left)            \
+  vlib_get_next_frame_macro_with_aux (                                        \
+    vm, node, next_index, vectors, n_vectors_left, /* alloc new frame */ 1,   \
+    aux_data, /* maybe_no_aux */ 0)
+
+/** \brief Get pointer to next frame vector data and next frame aux data by
+    (@c vlib_node_runtime_t, @c next_index).
+ Standard single/dual loop boilerplate element.
+ @attention This is a MACRO, with SIDE EFFECTS.
+ @attention This MACRO is safe in case the next node does not support aux_data.
+ In that case aux_data is set to NULL.
+
+ @param vm vlib_main_t pointer, varies by thread
+ @param node current node vlib_node_runtime_t pointer
+ @param next_index requested graph arc index
+
+ @return @c vectors -- pointer to next available vector slot
+ @return @c aux_data -- pointer to next available aux data slot
+ @return @c n_vectors_left -- number of vector slots available
+*/
+#define vlib_get_next_frame_with_aux_safe(vm, node, next_index, vectors,      \
+					  aux_data, n_vectors_left)           \
+  vlib_get_next_frame_macro_with_aux (                                        \
+    vm, node, next_index, vectors, n_vectors_left, /* alloc new frame */ 0,   \
+    aux_data, /* maybe_no_aux */ 1)
+
+#define vlib_get_new_next_frame_with_aux_safe(vm, node, next_index, vectors,  \
+					      aux_data, n_vectors_left)       \
+  vlib_get_next_frame_macro_with_aux (                                        \
+    vm, node, next_index, vectors, n_vectors_left, /* alloc new frame */ 1,   \
+    aux_data, /* maybe_no_aux */ 1)
+
 /** \brief Release pointer to next frame vector data.
  Standard single/dual loop boilerplate element.
  @param vm vlib_main_t pointer, varies by thread