4 files changed, 42 insertions, 11 deletions

diff --git a/erts/doc/src/erlang.xml b/erts/doc/src/erlang.xml
index fde04504c8..423ccdf98f 100644
--- a/erts/doc/src/erlang.xml
+++ b/erts/doc/src/erlang.xml
@@ -8958,7 +8958,11 @@ timestamp() ->
       <name name="trace_delivered" arity="1"/>
       <fsummary>Notification when trace has been delivered.</fsummary>
       <desc>
-        <p>The delivery of trace messages is dislocated on the time-line
+        <p>The delivery of trace messages (generated by
+         <seealso marker="#trace/3"><c>erlang:trace/3</c></seealso>,
+         <seealso marker="kernel:seq_trace"><c>seq_trace</c></seealso> or
+         <seealso marker="#system_profile/2"><c>erlang:system_profile/2</c></seealso>)
+          is dislocated on the time-line
           compared to other events in the system. If you know that
           <c><anno>Tracee</anno></c> has passed some specific point
           in its execution,
@@ -8984,8 +8988,11 @@ timestamp() ->
           or previously existing on the same node as the caller of
           <c>erlang:trace_delivered(<anno>Tracee</anno>)</c> resides on.
           The special <c><anno>Tracee</anno></c> atom <c>all</c>
-          denotes all processes
-          that currently are traced in the node.</p>
+          denotes all processes that currently are traced in the node.</p>
+        <p>When used together with an <seealso marker="#erl_tracer">
+          Tracer Module</seealso> any message sent in the trace callback
+          is guaranteed to have reached it's recipient before the
+          <c>trace_delivered</c> message is sent.</p>
         <p>Example: Process <c>A</c> is <c><anno>Tracee</anno></c>,
 	  port <c>B</c> is tracer, and process <c>C</c> is the port
 	  owner of <c>B</c>. <c>C</c> wants to close <c>B</c> when
diff --git a/erts/emulator/beam/erl_bif_trace.c b/erts/emulator/beam/erl_bif_trace.c
index 7e2fe7fcd4..ff2018aa27 100644
--- a/erts/emulator/beam/erl_bif_trace.c
+++ b/erts/emulator/beam/erl_bif_trace.c
@@ -2253,25 +2253,32 @@ static void
 reply_trace_delivered_all(void *vtdarp)
 {
     ErtsTraceDeliveredAll *tdarp = (ErtsTraceDeliveredAll *) vtdarp;
-    ErtsProcLocks rp_locks = 0;
 
     if (erts_smp_atomic32_dec_read_nob(&tdarp->refc) == 0) {
+        Eterm ref_copy, msg;
         Process *rp = tdarp->proc;
         Eterm *hp = NULL;
-        ErlOffHeap *ohp = NULL;
-        ErtsMessage *mp = NULL;
-        Eterm ref_copy, msg;
-
+        ErlOffHeap *ohp;
+#ifdef ERTS_SMP
+        ErlHeapFragment *bp;
+        bp = new_message_buffer(4 + NC_HEAP_SIZE(tdarp->ref));
+        hp = &bp->mem[0];
+        ohp = &bp->off_heap;
+#else
+        ErtsProcLocks rp_locks = 0;
+        ErtsMessage *mp;
         mp = erts_alloc_message_heap(
             rp, &rp_locks, 4 + NC_HEAP_SIZE(tdarp->ref), &hp, &ohp);
+#endif
 
         ref_copy = STORE_NC(&hp, ohp, tdarp->ref);
         msg = TUPLE3(hp, am_trace_delivered, tdarp->target, ref_copy);
 
+#ifdef ERTS_SMP
+        erts_send_sys_msg_proc(rp->common.id, rp->common.id, msg, bp);
+#else
         erts_queue_message(rp, &rp_locks, mp, msg);
-
-        if (rp_locks)
-            erts_smp_proc_unlock(rp, rp_locks);
+#endif
 
 	erts_free(ERTS_ALC_T_MISC_AUX_WORK, vtdarp);
         erts_proc_dec_refc(rp);
diff --git a/erts/emulator/beam/erl_trace.c b/erts/emulator/beam/erl_trace.c
index cd3c14e213..bd88769dfc 100644
--- a/erts/emulator/beam/erl_trace.c
+++ b/erts/emulator/beam/erl_trace.c
@@ -84,6 +84,7 @@ enum ErtsSysMsgType {
     SYS_MSG_TYPE_UNDEFINED,
     SYS_MSG_TYPE_SYSMON,
     SYS_MSG_TYPE_ERRLGR,
+    SYS_MSG_TYPE_PROC_MSG,
     SYS_MSG_TYPE_SYSPROF
 };
 
@@ -2157,6 +2158,13 @@ erts_queue_error_logger_message(Eterm from, Eterm msg, ErlHeapFragment *bp)
     enqueue_sys_msg(SYS_MSG_TYPE_ERRLGR, from, am_error_logger, msg, bp);
 }
 
+void
+erts_send_sys_msg_proc(Eterm from, Eterm to, Eterm msg, ErlHeapFragment *bp)
+{
+    ASSERT(is_internal_pid(to));
+    enqueue_sys_msg(SYS_MSG_TYPE_PROC_MSG, from, to, msg, bp);
+}
+
 #ifdef DEBUG_PRINTOUTS
 static void
 print_msg_type(ErtsSysMsgQ *smqp)
@@ -2171,6 +2179,9 @@ print_msg_type(ErtsSysMsgQ *smqp)
     case SYS_MSG_TYPE_ERRLGR:
 	erts_fprintf(stderr, "ERRLGR ");
 	break;
+    case SYS_MSG_TYPE_PROC_MSG:
+       erts_fprintf(stderr, "PROC_MSG ");
+       break;
     default:
 	erts_fprintf(stderr, "??? ");
 	break;
@@ -2241,6 +2252,8 @@ sys_msg_disp_failure(ErtsSysMsgQ *smqp, Eterm receiver)
 		     no_elgger, tag, CAR(list_val(tp[3])));
 	break;
     }
+    case SYS_MSG_TYPE_PROC_MSG:
+        break;
     default:
 	ASSERT(0);
     }
@@ -2358,6 +2371,9 @@ sys_msg_dispatcher_func(void *unused)
 	    print_msg_type(smqp);
 #endif
 	    switch (smqp->type) {
+            case SYS_MSG_TYPE_PROC_MSG:
+                receiver = smqp->to;
+                break;
 	    case SYS_MSG_TYPE_SYSMON:
 		receiver = erts_get_system_monitor();
 		if (smqp->from == receiver) {
diff --git a/erts/emulator/beam/erl_trace.h b/erts/emulator/beam/erl_trace.h
index 14d0c36016..177fd373a6 100644
--- a/erts/emulator/beam/erl_trace.h
+++ b/erts/emulator/beam/erl_trace.h
@@ -87,6 +87,7 @@ void erts_foreach_sys_msg_in_q(void (*func)(Eterm,
 					    Eterm,
 					    ErlHeapFragment *));
 void erts_queue_error_logger_message(Eterm, Eterm, ErlHeapFragment *);
+void erts_send_sys_msg_proc(Eterm, Eterm, Eterm, ErlHeapFragment *);
 #endif
 
 void trace_send(Process*, Eterm, Eterm);