diff --git a/src/decode.h b/src/decode.h
index 50036192ba..4b097a2123 100644
--- a/src/decode.h
+++ b/src/decode.h
@@ -627,6 +627,16 @@ typedef struct DecodeThreadVars_
 #endif
 
 
+/* \todo there is another copy of this same macro inside flow-util.h.  The
+ *       reason we have this duplicate is because we couldn't solve endless
+ *       header files cross-reference. */
+#define FlowDeReference(src_f_ptr) do {               \
+        if (*(src_f_ptr) != NULL) {                   \
+            FlowDecrUsecnt(*(src_f_ptr));             \
+            *(src_f_ptr) = NULL;                      \
+        }                                             \
+    } while (0)
+
 /**
  *  \brief Recycle a packet structure for reuse.
  *  \todo the mutex destroy & init is necessary because of the memset, reconsider
@@ -641,7 +651,7 @@ typedef struct DecodeThreadVars_
         (p)->flags = 0;                         \
         (p)->flowflags = 0;                     \
         (p)->pkt_src = 0;                       \
-        (p)->flow = NULL;                       \
+        FlowDeReference(&((p)->flow));          \
         (p)->ts.tv_sec = 0;                     \
         (p)->ts.tv_usec = 0;                    \
         (p)->datalink = 0;                      \
diff --git a/src/flow-util.h b/src/flow-util.h
index 21768bd327..c6e15b3e21 100644
--- a/src/flow-util.h
+++ b/src/flow-util.h
@@ -134,6 +134,9 @@
         }                                           \
     } while (0)
 
+/* \todo there is another copy of this same macro inside decode.h.  The reason
+ *       we have this duplicate is because we couldn't solve endless header
+ *       files cross-reference. */
 #define FlowDeReference(src_f_ptr) do {               \
         if (*(src_f_ptr) != NULL) {                   \
             FlowDecrUsecnt(*(src_f_ptr));             \
diff --git a/src/stream-tcp.c b/src/stream-tcp.c
index dbc6ab85c5..8c9d6e5b41 100644
--- a/src/stream-tcp.c
+++ b/src/stream-tcp.c
@@ -190,7 +190,7 @@ void StreamTcpSessionClear(void *ssnptr)
         SCLogDebug("returning smsg %p to pool", smsg);
         smsg->next = NULL;
         smsg->prev = NULL;
-        smsg->flow = NULL;
+        FlowDeReference(&smsg->flow);
         StreamMsgReturnToPool(smsg);
         smsg = smsg_next;
     }
@@ -202,7 +202,7 @@ void StreamTcpSessionClear(void *ssnptr)
         SCLogDebug("returning smsg %p to pool", smsg);
         smsg->next = NULL;
         smsg->prev = NULL;
-        smsg->flow = NULL;
+        FlowDeReference(&smsg->flow);
         StreamMsgReturnToPool(smsg);
         smsg = smsg_next;
     }
@@ -289,7 +289,7 @@ void StreamTcpSessionPoolCleanup(void *s)
         SCLogDebug("returning smsg %p to pool", smsg);
         smsg->next = NULL;
         smsg->prev = NULL;
-        smsg->flow = NULL;
+        FlowDeReference(&smsg->flow);
         StreamMsgReturnToPool(smsg);
         smsg = smsg_next;
     }
@@ -301,7 +301,7 @@ void StreamTcpSessionPoolCleanup(void *s)
         SCLogDebug("returning smsg %p to pool", smsg);
         smsg->next = NULL;
         smsg->prev = NULL;
-        smsg->flow = NULL;
+        FlowDeReference(&smsg->flow);
         StreamMsgReturnToPool(smsg);
         smsg = smsg_next;
     }
diff --git a/src/stream.c b/src/stream.c
index d6f8a9a79c..6bb43dfea7 100644
--- a/src/stream.c
+++ b/src/stream.c
@@ -30,6 +30,7 @@
 #include "util-pool.h"
 #include "util-debug.h"
 #include "stream-tcp.h"
+#include "flow-util.h"
 
 #ifdef DEBUG
 static SCMutex stream_pool_memuse_mutex;
@@ -215,7 +216,7 @@ void StreamMsgReturnListToPool(void *list) {
         SCLogDebug("returning smsg %p to pool", smsg);
         smsg->next = NULL;
         smsg->prev = NULL;
-        smsg->flow = NULL;
+        FlowDeReference(&smsg->flow);
         StreamMsgReturnToPool(smsg);
         smsg = smsg_next;
     }
diff --git a/src/tmqh-packetpool.c b/src/tmqh-packetpool.c
index 84a7f6a021..9310f88075 100644
--- a/src/tmqh-packetpool.c
+++ b/src/tmqh-packetpool.c
@@ -36,6 +36,7 @@
 #include "threads.h"
 #include "threadvars.h"
 #include "flow.h"
+#include "flow-util.h"
 
 #include "stream.h"
 #include "stream-tcp-reassemble.h"
@@ -242,13 +243,13 @@ void TmqhOutputPacketpool(ThreadVars *t, Packet *p)
         SCLogDebug("tunnel stuff done, move on (proot %d)", proot);
     }
 
-    FlowDecrUsecnt(p->flow);
+    FlowDeReference(&p->flow);
 
     /* we're done with the tunnel root now as well */
     if (proot == 1) {
         SCLogDebug("getting rid of root pkt... alloc'd %s", p->root->flags & PKT_ALLOC ? "true" : "false");
 
-        FlowDecrUsecnt(p->root->flow);
+        FlowDeReference(&p->root->flow);
         /* if p->root uses extended data, free them */
         if (p->root->ReleaseData) {
             if (p->root->ReleaseData(t, p->root) == TM_ECODE_FAILED) {