File 6401-erts-Support-for-truly-asynchronous-distri... of Package erlang

Overview Repositories Revisions Requests Users Attributes Meta

File 6401-erts-Support-for-truly-asynchronous-distributed-sign.patch of Package erlang (Revision 33fb1bfcfb3bce757d0f7cfc6a9a3eaf)

Currently displaying revision 33fb1bfcfb3bce757d0f7cfc6a9a3eaf , Show latest

From 349b244f2e63eda1579891de027e96c6dfedfd55 Mon Sep 17 00:00:00 2001
From: Rickard Green <rickard@erlang.org>
Date: Fri, 16 Dec 2022 18:36:39 +0100
Subject: [PATCH] [erts] Support for truly asynchronous distributed signaling

diff --git a/erts/doc/src/erl_cmd.xml b/erts/doc/src/erl_cmd.xml
index 9f764d2b4e..6c4919413d 100644
--- a/erts/doc/src/erl_cmd.xml
+++ b/erts/doc/src/erl_cmd.xml
@@ -1036,6 +1036,23 @@ $ <input>erl \
 Memory allocator-specific flags. For more information, see
 <seecref marker="erts_alloc"><c>erts_alloc(3)</c></seecref>.
 </item>
+ <tag><marker id="+pad"/><c>+pad true|false</c></tag>
+ <item>
+ Since: OTP 25.3
+ 
+ The boolean value used with the <c>+pad</c> parameter determines
+ the default value of the
+ <seeerl marker="erlang#process_flag_async_dist">
+ <c>async_dist</c></seeerl> process flag of newly spawned processes.
+ By default, if no <c>+pad</c> command line option is passed,
+ the <c>async_dist</c> flag will be set to <c>false</c>.
+ 
+ 
+ The value used in runtime can be inspected by calling
+ <seeerl marker="erlang#system_info_async_dist">
+ <c>erlang:system_info(async_dist)</c></seeerl>.
+ 
+ </item>
 <tag><marker id="+pc"/><marker id="printable_character_range"/>
 <c><![CDATA[+pc Range]]></c></tag>
 <item>
@@ -1781,6 +1798,11 @@ $ <input>erl \
 limit is per distribution channel. A higher limit
 gives lower latency and higher throughput at the expense
 of higher memory use.
+ 
+ This limit only affects processes that have disabled
+ <seeerl marker="erlang#process_flag_async_dist">fully
+ asynchronous distributed signaling</seeerl>.
+ 
 </item>
 <tag><marker id="+zdntgc"/><c>+zdntgc time</c></tag>
 <item>
diff --git a/erts/doc/src/erlang.xml b/erts/doc/src/erlang.xml
index 34fb78e323..ceb85415c1 100644
--- a/erts/doc/src/erlang.xml
+++ b/erts/doc/src/erlang.xml
@@ -6154,6 +6154,68 @@ receive_replies(ReqId, N, Acc) ->
 
 <func>
 <name name="process_flag" arity="2" clause_i="1"
+	 anchor="process_flag_async_dist" since="OTP 25.3"/>
+ <fsummary>
+ Enable or disable fully asynchronous distributed signaling
+ for the calling process.
+ </fsummary>
+ <desc>
+ 
+ Enable or disable fully asynchronous distributed signaling
+ for the calling process. When disabled, which is the default, the
+ process sending a distributed signal will block in the send
+ operation if the buffer for the distribution channel reach the
+ <seecom marker="erts:erl#+zdbbl">distribution buffer busy
+ limit</seecom>. The process will remain blocked until the buffer
+ shrinks enough. This might in some cases take a substantial amount
+ of time. When <c>async_dist</c> is enabled, send operations of
+ distributed signals will always buffer the signal on the outgoing
+ distribution channel and then immediately return. That is, these
+ send operations will never block the sending process.
+ 
+ <note>
+ Since no flow control is enforced by the runtime system when
+ <c>async_dist</c> process flag is enabled, you need to make sure
+ that flow control for such data is implemented, or that the amount
+ of such data is known to always be limited. Unlimited signaling with
+ <c>async_dist</c> enabled in the absence of flow control will
+ typically cause the sending runtime system to crash on an out of
+ memory condition.
+ </note>
+ 
+ Blocking due to disabled <c>async_dist</c> can be monitored by
+ <seemfa marker="#system_monitor/2"><c>erlang:system_montor()</c></seemfa>
+ using the
+ <seeerl marker="#busy_dist_port"><c>busy_dist_port</c></seeerl>
+ option. Only data buffered by processes which (at the time of sending
+ a signal) have disabled <c>async_dist</c> will be counted when
+ determining whether or not an operation should block the caller.
+ 
+ 
+ The <c>async_dist</c> flag can also be set on a new process when
+ spawning it using the
+ <seemfa marker="#spawn_opt/4"><c>spawn_opt()</c></seemfa> BIF with the
+ option <seeerl marker="#spawn_opt_async_dist"><c>{async_dist,
+ Enable}</c></seeerl>. The default <c>async_dist</c> flag to use on
+ newly spawned processes can be set by passing the command line
+ argument <seecom marker="erl#+pad"><c>+pad
+ &lt;boolean&gt;</c></seecom> when starting the runtime system. If the
+ <c>+pad &lt;boolean&gt;</c> command line argument is not passed, the
+ default value of the <c>async_dist</c> flag will be <c>false</c>.
+ 
+ 
+ You can inspect the state of the <c>async_dist</c> process flag of a
+ process by calling <seeerl marker="#process_info_async_dist">
+ <c>process_info(Pid, async_dist)</c></seeerl>.
+ 
+ 
+ Returns the old value of the <c>async_dist</c> flag.
+ 
+ </desc>
+ </func>
+
+ <func>
+ <name name="process_flag" arity="2" clause_i="2"
 	 anchor="process_flag_trap_exit" since=""/>
 <fsummary>Set process flag trap_exit for the calling process.</fsummary>
 <desc>
@@ -6171,7 +6233,7 @@ receive_replies(ReqId, N, Acc) ->
 </func>
 
 <func>
- <name name="process_flag" arity="2" clause_i="2" since=""/>
+ <name name="process_flag" arity="2" clause_i="3" since=""/>
 <fsummary>Set process flag error_handler for the calling process.
 </fsummary>
 <desc>
@@ -6185,7 +6247,7 @@ receive_replies(ReqId, N, Acc) ->
 </func>
 
 <func>
- <name name="process_flag" arity="2" clause_i="3" since="OTP 24.0"/>
+ <name name="process_flag" arity="2" clause_i="4" since="OTP 24.0"/>
 <fsummary>Set process flag fullsweep_after for the calling process.
 </fsummary>
 <desc>
@@ -6196,7 +6258,7 @@ receive_replies(ReqId, N, Acc) ->
 </func>
 
 <func>
- <name name="process_flag" arity="2" clause_i="4"
+ <name name="process_flag" arity="2" clause_i="5"
 	 anchor="process_flag_min_heap_size" since=""/>
 <fsummary>Set process flag min_heap_size for the calling process.
 </fsummary>
@@ -6207,7 +6269,7 @@ receive_replies(ReqId, N, Acc) ->
 </func>
 
 <func>
- <name name="process_flag" arity="2" clause_i="5" since="OTP R13B04"/>
+ <name name="process_flag" arity="2" clause_i="6" since="OTP R13B04"/>
 <fsummary>Set process flag min_bin_vheap_size for the calling process.
 </fsummary>
 <desc>
@@ -6218,7 +6280,7 @@ receive_replies(ReqId, N, Acc) ->
 </func>
 
 <func>
- <name name="process_flag" arity="2" clause_i="6"
+ <name name="process_flag" arity="2" clause_i="7"
 	 anchor="process_flag_max_heap_size" since="OTP 19.0"/>
 <fsummary>Set process flag max_heap_size for the calling process.
 </fsummary>
@@ -6297,7 +6359,7 @@ receive_replies(ReqId, N, Acc) ->
 </func>
 
 <func>
- <name name="process_flag" arity="2" clause_i="7"
+ <name name="process_flag" arity="2" clause_i="8"
 	 anchor="process_flag_message_queue_data" since="OTP 19.0"/>
 <fsummary>Set process flag message_queue_data for the calling process.
 </fsummary>
@@ -6339,7 +6401,7 @@ receive_replies(ReqId, N, Acc) ->
 </func>
 
 <func>
- <name name="process_flag" arity="2" clause_i="8"
+ <name name="process_flag" arity="2" clause_i="9"
 	 anchor="process_flag_priority" since=""/>
 <fsummary>Set process flag priority for the calling process.</fsummary>
 <type name="priority_level"/>
@@ -6412,7 +6474,7 @@ receive_replies(ReqId, N, Acc) ->
 </func>
 
 <func>
- <name name="process_flag" arity="2" clause_i="9" since=""/>
+ <name name="process_flag" arity="2" clause_i="10" since=""/>
 <fsummary>Set process flag save_calls for the calling process.</fsummary>
 <desc>
 <c><anno>N</anno></c> must be an integer in the interval 0..10000.
@@ -6443,7 +6505,7 @@ receive_replies(ReqId, N, Acc) ->
 </func>
 
 <func>
- <name name="process_flag" arity="2" clause_i="10" since=""/>
+ <name name="process_flag" arity="2" clause_i="11" since=""/>
 <fsummary>Set process flag sensitive for the calling process.</fsummary>
 <desc>
 Sets or clears flag <c>sensitive</c> for the current process.
@@ -6588,6 +6650,18 @@ receive_replies(ReqId, N, Acc) ->
 Valid <c><anno>InfoTuple</anno></c>s with corresponding
 <c><anno>Item</anno></c>s:
 <taglist>
+ <tag>
+ <marker id="process_info_async_dist"/>
+ <c>{async_dist, Enabled}</c>
+ </tag>
+ <item>
+ Since: OTP 25.3
+ 
+ Current value of the
+ <seeerl marker="erlang#process_flag_async_dist">
+ <c>async_dist</c></seeerl> process flag.
+ 
+ </item>
 <tag><c>{backtrace, <anno>Bin</anno>}</c></tag>
 <item>
 Binary <c><anno>Bin</anno></c> contains the same information
@@ -7831,6 +7905,21 @@ true</pre>
 <c>process_flag(message_queue_data,
 <anno>MQD</anno>)</c></seeerl>.
 </item>
+ <tag>
+ <marker id="spawn_opt_async_dist"/>
+ <c>{async_dist, Enabled}</c>
+ </tag>
+ <item>
+ Since: OTP 25.3
+ 
+ Set the
+ <seeerl marker="erlang#process_flag_async_dist">
+ <c>async_dist</c></seeerl> process flag of the spawned process.
+ This option will override the default value set by the command
+ line argument
+ <seecom marker="erl#+pad"><c>+pad &lt;boolean&gt;</c></seecom>.
+ 
+ </item>
 </taglist>
 </desc>
 </func>
@@ -10861,6 +10950,8 @@ Metadata = #{ pid => pid(),
 </func>
 
 <func>
+ <name name="system_info" arity="1" clause_i="79"
+	 anchor="system_info_async_dist" since="OTP 25.3"/> 
 <name name="system_info" arity="1" clause_i="14"
 	 anchor="system_info_dist" since=""/> 
 <name name="system_info" arity="1" clause_i="16" since="OTP 18.0"/> 
@@ -10873,6 +10964,17 @@ Metadata = #{ pid => pid(),
 Returns information about Erlang Distribution in the
 current system as specified by <c><anno>Item</anno></c>:
 <taglist>
+ <tag><marker id="system_info_async_dist"/><c>async_dist</c></tag>
+ <item>
+ Since: OTP 25.3
+ 
+ Returns the value of the command line argument
+ <seecom marker="erl#+pad">+pad &lt;boolean&gt;</seecom>
+ which the runtime system use. This value determines the default
+ <seeerl marker="erlang#process_flag_async_dist">
+ <c>async_dist</c></seeerl> value for newly spawned processes.
+ 
+ </item>
 <tag><marker id="system_info_creation"/>
 <c>creation</c></tag>
 <item>
@@ -11011,7 +11113,7 @@ Metadata = #{ pid => pid(),
 
 <name name="system_info" arity="1" clause_i="77" since=""/> 
 <name name="system_info" arity="1" clause_i="78" since=""/> 
- 
+ 
 
 <fsummary>Information about the system.</fsummary>
 <desc>
@@ -11379,7 +11481,7 @@ Metadata = #{ pid => pid(),
 <c><anno>MonitorPid</anno></c>. <c>SusPid</c> is the pid
 that got suspended when sending to <c>Port</c>.
 </item>
- <tag><c>busy_dist_port</c></tag>
+ <tag><c>busy_dist_port</c><marker id="busy_dist_port"/></tag>
 <item>
 If a process in the system gets suspended because it
 sends to a process on a remote node whose inter-node
diff --git a/erts/emulator/beam/atom.names b/erts/emulator/beam/atom.names
index 313b646cdd..5a5bcbceea 100644
--- a/erts/emulator/beam/atom.names
+++ b/erts/emulator/beam/atom.names
@@ -105,6 +105,7 @@ atom arg0
 atom arity
 atom asn1
 atom async
+atom async_dist
 atom asynchronous
 atom atom
 atom atom_used
diff --git a/erts/emulator/beam/bif.c b/erts/emulator/beam/bif.c
index 4e6279704b..fdef0682e0 100644
--- a/erts/emulator/beam/bif.c
+++ b/erts/emulator/beam/bif.c
@@ -1777,7 +1777,21 @@ static Eterm process_flag_aux(Process *c_p, int *redsp, Eterm flag, Eterm val)
 BIF_RETTYPE process_flag_2(BIF_ALIST_2)
 {
 Eterm old_value;
- if (BIF_ARG_1 == am_error_handler) {
+
+ if (BIF_ARG_1 == am_async_dist) {
+ old_value = (BIF_P->flags & F_ASYNC_DIST) ? am_true : am_false;
+ if (BIF_ARG_2 == am_false) {
+ BIF_P->flags &= ~F_ASYNC_DIST;
+ }
+ else if (BIF_ARG_2 == am_true) {
+ BIF_P->flags |= F_ASYNC_DIST;
+ }
+ else {
+ goto error;
+ }
+ BIF_RET(old_value);
+ }
+ else if (BIF_ARG_1 == am_error_handler) {
 if (is_not_atom(BIF_ARG_2)) {
 	 goto error;
 }
diff --git a/erts/emulator/beam/dist.c b/erts/emulator/beam/dist.c
index 0789813a9c..5455b47f59 100644
--- a/erts/emulator/beam/dist.c
+++ b/erts/emulator/beam/dist.c
@@ -211,6 +211,37 @@ struct {
 ErlHeapFragment *bp;
 } nodedown;
 
+/*
+ * Dist entry queue flags are only modified while
+ * the dist entry queue lock is held...
+ */
+static ERTS_INLINE erts_aint32_t
+de_qflags_read(DistEntry *dep)
+{
+ return erts_atomic32_read_nob(&dep->qflgs);
+}
+
+static ERTS_INLINE erts_aint32_t
+de_qflags_read_set(DistEntry *dep, erts_aint32_t set)
+{
+ erts_aint32_t qflgs, new_qflgs;
+ ERTS_LC_ASSERT(erts_lc_mtx_is_locked(&dep->qlock));
+ new_qflgs = qflgs = erts_atomic32_read_nob(&dep->qflgs);
+ new_qflgs |= set;
+ erts_atomic32_set_nob(&dep->qflgs, new_qflgs);
+ return qflgs;
+}
+
+static ERTS_INLINE erts_aint32_t
+de_qflags_read_unset(DistEntry *dep, erts_aint32_t unset)
+{
+ erts_aint32_t qflgs, new_qflgs;
+ ERTS_LC_ASSERT(erts_lc_mtx_is_locked(&dep->qlock));
+ new_qflgs = qflgs = erts_atomic32_read_nob(&dep->qflgs);
+ new_qflgs &= ~unset;
+ erts_atomic32_set_nob(&dep->qflgs, new_qflgs);
+ return qflgs;
+}
 
 static void
 delete_cache(ErtsAtomCache *cache)
@@ -250,7 +281,7 @@ get_suspended_on_de(DistEntry *dep, erts_aint32_t unset_qflgs)
 {
 erts_aint32_t qflgs;
 ERTS_LC_ASSERT(erts_lc_mtx_is_locked(&dep->qlock));
- qflgs = erts_atomic32_read_band_acqb(&dep->qflgs, ~unset_qflgs);
+ qflgs = de_qflags_read_unset(dep, unset_qflgs);
 qflgs &= ~unset_qflgs;
 if (qflgs & ERTS_DE_QFLG_EXIT) {
 	/* No resume when exit has been scheduled */
@@ -1010,14 +1041,14 @@ int erts_do_net_exits(DistEntry *dep, Eterm reason)
 }
 
 	if (dep->state == ERTS_DE_STATE_EXITING) {
-	 ASSERT(erts_atomic32_read_nob(&dep->qflgs) & ERTS_DE_QFLG_EXIT);
+	 ASSERT(de_qflags_read(dep) & ERTS_DE_QFLG_EXIT);
 	}
 	else {
 ASSERT(dep->state == ERTS_DE_STATE_CONNECTED);
 	 dep->state = ERTS_DE_STATE_EXITING;
 	 erts_mtx_lock(&dep->qlock);
-	 ASSERT(!(erts_atomic32_read_nob(&dep->qflgs) & ERTS_DE_QFLG_EXIT));
-	 erts_atomic32_read_bor_relb(&dep->qflgs, ERTS_DE_QFLG_EXIT);
+	 ASSERT(!(de_qflags_read(dep) & ERTS_DE_QFLG_EXIT));
+	 de_qflags_read_set(dep, ERTS_DE_QFLG_EXIT);
 	 erts_mtx_unlock(&dep->qlock);
 	}
 
@@ -1043,6 +1074,8 @@ int erts_do_net_exits(DistEntry *dep, Eterm reason)
 suspendees = get_suspended_on_de(dep, ERTS_DE_QFLGS_ALL);
 
 erts_mtx_unlock(&dep->qlock);
+
+ erts_atomic32_set_relb(&dep->notify, 0);
 erts_atomic_set_nob(&dep->dist_cmd_scheduled, 0);
 dep->send = NULL;
 
@@ -1144,7 +1177,8 @@ void init_dist(void)
 
 static ERTS_INLINE ErtsDistOutputBuf *
 alloc_dist_obufs(byte **extp, TTBEncodeContext *ctx,
- Uint data_size, Uint fragments, Uint vlen)
+ Uint data_size, Uint fragments, Uint vlen,
+ int ignore_busy)
 {
 int ix;
 ErtsDistOutputBuf *obuf;
@@ -1175,6 +1209,7 @@ alloc_dist_obufs(byte **extp, TTBEncodeContext *ctx,
 erts_refc_add(&bin->intern.refc, fragments - 1, 1);
 
 for (ix = 0; ix < fragments; ix++) {
+ obuf[ix].ignore_busy = ignore_busy;
 obuf[ix].bin = bin;
 obuf[ix].eiov = &ctx->fragment_eiovs[ix];
 #ifdef DEBUG
@@ -1220,6 +1255,81 @@ size_obuf(ErtsDistOutputBuf *obuf)
 return sz;
 }
 
+static ERTS_INLINE void
+get_obuf_sizes(ErtsDistOutputBuf *obuf, Sint *size, Sint *ignore_size)
+{
+ Sint sz = size_obuf(obuf);
+ ASSERT(sz >= 0);
+ *size = sz;
+ *ignore_size = obuf->ignore_busy ? sz : 0;
+}
+
+static ERTS_INLINE void
+add_obuf_sizes(ErtsDistOutputBuf *obuf, Sint *size, Sint *ignore_size)
+{
+ Sint sz, isz;
+ get_obuf_sizes(obuf, &sz, &isz);
+ *size += sz;
+ *ignore_size += isz;
+}
+
+static ERTS_INLINE void
+subtract_obuf_sizes(ErtsDistOutputBuf *obuf, Sint *size, Sint *ignore_size)
+{
+ Sint sz, isz;
+ get_obuf_sizes(obuf, &sz, &isz);
+ *size -= sz;
+ *ignore_size -= isz;
+}
+
+static ERTS_INLINE void
+update_qsizes(DistEntry *dep, int *empty_fillp, Sint *qsizep,
+ Sint add_total_qsize, Sint ignore_qsize)
+{
+ /*
+ * All modifications of the 'total_qsize' and 'qsize' fields are
+ * made while holding the 'qlock', so read/modify/write of each
+ * field does not need to be atomic. Readers without the lock will
+ * still see consistent updates of each 'field'.
+ */
+ erts_aint_t qsize, add_qsize;
+
+ ERTS_LC_ASSERT(erts_lc_mtx_is_locked(&dep->qlock));
+
+ if (empty_fillp)
+ *empty_fillp = 0;
+
+ if (add_total_qsize) {
+ qsize = erts_atomic_read_nob(&dep->total_qsize);
+ qsize += (erts_aint_t) add_total_qsize;
+ if (empty_fillp && qsize == add_total_qsize)
+ *empty_fillp = !0;
+ erts_atomic_set_nob(&dep->total_qsize, (erts_aint_t) qsize);
+ }
+
+ add_qsize = (erts_aint_t) (add_total_qsize - ignore_qsize);
+ if (add_qsize) {
+ qsize = erts_atomic_read_nob(&dep->qsize);
+ qsize += add_qsize;
+ if (qsizep)
+ *qsizep = qsize;
+ erts_atomic_set_nob(&dep->qsize, (erts_aint_t) qsize);
+ }
+ else if (qsizep) {
+ *qsizep = erts_atomic_read_nob(&dep->qsize);
+ }
+
+#ifdef DEBUG
+ {
+ erts_aint_t tqsize = erts_atomic_read_nob(&dep->total_qsize);
+ qsize = erts_atomic_read_nob(&dep->qsize);
+ ASSERT(tqsize >= 0);
+ ASSERT(qsize >= 0);
+ ASSERT(tqsize >= qsize);
+ }
+#endif
+}
+
 static ErtsDistOutputBuf* clear_de_out_queues(DistEntry* dep)
 {
 ErtsDistOutputBuf *obuf;
@@ -1250,21 +1360,19 @@ static ErtsDistOutputBuf* clear_de_out_queues(DistEntry* dep)
 
 static void free_de_out_queues(DistEntry* dep, ErtsDistOutputBuf *obuf)
 {
- Sint obufsize = 0;
+ Sint obufsize = 0, ignore_obufsize = 0;
 
 while (obuf) {
 	ErtsDistOutputBuf *fobuf;
 	fobuf = obuf;
 	obuf = obuf->next;
-	obufsize += size_obuf(fobuf);
+ add_obuf_sizes(fobuf, &obufsize, &ignore_obufsize);
 	free_dist_obuf(fobuf, !0);
 }
 
 if (obufsize) {
 	erts_mtx_lock(&dep->qlock);
- ASSERT(erts_atomic_read_nob(&dep->qsize) >= obufsize);
- erts_atomic_add_nob(&dep->qsize,
- (erts_aint_t) -obufsize);
+ update_qsizes(dep, NULL, NULL, -obufsize, -ignore_obufsize);
 	erts_mtx_unlock(&dep->qlock);
 }
 }
@@ -3033,11 +3141,17 @@ retry:
 	goto fail;
 }
 
- if (no_suspend && proc) {
-	if (erts_atomic32_read_acqb(&dep->qflgs) & ERTS_DE_QFLG_BUSY) {
-	 res = ERTS_DSIG_PREP_WOULD_SUSPEND;
-	 goto fail;
-	}
+ if (!proc || (proc->flags & F_ASYNC_DIST)) {
+ ctx->ignore_busy = !0;
+ }
+ else {
+ ctx->ignore_busy = 0;
+ if (no_suspend) {
+ if (de_qflags_read(dep) & ERTS_DE_QFLG_BUSY) {
+ res = ERTS_DSIG_PREP_WOULD_SUSPEND;
+ goto fail;
+ }
+ }
 }
 
 ctx->c_p = proc;
@@ -3211,7 +3325,8 @@ erts_dsig_send(ErtsDSigSendContext *ctx)
 + ((ctx->fragments - 1)
 * ERTS_DIST_FRAGMENT_HEADER_SIZE),
 ctx->fragments,
- ctx->vlen);
+ ctx->vlen,
+ ctx->ignore_busy);
 ctx->alloced_fragments = ctx->fragments;
 	 /* Encode internal version of dist header */
 ctx->dhdrp = ctx->extp;
@@ -3362,20 +3477,23 @@ erts_dsig_send(ErtsDSigSendContext *ctx)
 ctx->fragments = 0;
 	 }
 	 else {
- Sint qsize = erts_atomic_read_nob(&dep->qsize);
+ Sint qsize = (Sint) erts_atomic_read_nob(&dep->qsize);
 erts_aint32_t qflgs;
 		ErtsProcList *plp = NULL;
 Eterm notify_proc = NIL;
 Sint obsz;
- int fragments;
+ int fragments, empty_fill;
 
 /* Calculate how many fragments to send. This depends on
 the available space in the distr queue and the amount
 of remaining reductions. */
 for (fragments = 0, obsz = 0;
- fragments < ctx->fragments &&
- ((ctx->reds > 0 && (qsize + obsz) < erts_dist_buf_busy_limit) ||
- ctx->no_trap || ctx->no_suspend);
+ (fragments < ctx->fragments
+ && ((ctx->reds > 0
+ && (ctx->ignore_busy
+ || (qsize + obsz < erts_dist_buf_busy_limit)))
+ || ctx->no_trap
+ || ctx->no_suspend));
 fragments++) {
 #ifdef DEBUG
 int reds = 100;
@@ -3391,33 +3509,27 @@ erts_dsig_send(ErtsDSigSendContext *ctx)
 (!ctx->no_trap && !ctx->no_suspend));
 
 		erts_mtx_lock(&dep->qlock);
-		qsize = erts_atomic_add_read_mb(&dep->qsize, (erts_aint_t) obsz);
- ASSERT(qsize >= obsz);
- qflgs = erts_atomic32_read_nob(&dep->qflgs);
-		if (!(qflgs & ERTS_DE_QFLG_BUSY) && qsize >= erts_dist_buf_busy_limit) {
-		 erts_atomic32_read_bor_relb(&dep->qflgs, ERTS_DE_QFLG_BUSY);
+ update_qsizes(dep, &empty_fill, &qsize, obsz,
+ ctx->ignore_busy ? obsz : 0);
+ qflgs = de_qflags_read(dep);
+		if (!(qflgs & ERTS_DE_QFLG_BUSY)
+ && qsize >= erts_dist_buf_busy_limit) {
+		 qflgs = de_qflags_read_set(dep, ERTS_DE_QFLG_BUSY);
 qflgs |= ERTS_DE_QFLG_BUSY;
 }
- if (qsize == obsz && (qflgs & ERTS_DE_QFLG_REQ_INFO)) {
- /* Previously empty queue and info requested... */
- qflgs = erts_atomic32_read_band_mb(&dep->qflgs,
- ~ERTS_DE_QFLG_REQ_INFO);
- if (qflgs & ERTS_DE_QFLG_REQ_INFO) {
+ if (empty_fill && is_internal_pid(dep->cid)) {
+ erts_aint32_t notify;
+ notify = erts_atomic32_xchg_mb(&dep->notify,
+ (erts_aint32_t) 0);
+ if (notify) {
+ /*
+ * Previously empty queue and notification
+ * requested...
+ */
 notify_proc = dep->cid;
 ASSERT(is_internal_pid(notify_proc));
 }
- /* else: requester will send itself the message... */
- qflgs &= ~ERTS_DE_QFLG_REQ_INFO;
 }
-		if (!ctx->no_suspend && (qflgs & ERTS_DE_QFLG_BUSY)) {
-		 erts_mtx_unlock(&dep->qlock);
-
-		 plp = erts_proclist_create(ctx->c_p);
-
-		 erts_suspend(ctx->c_p, ERTS_PROC_LOCK_MAIN, NULL);
-		 suspended = 1;
-		 erts_mtx_lock(&dep->qlock);
-		}
 
 ASSERT(fragments < 2
 || (get_int64(&((char*)ctx->obuf->eiov->iov[1].iov_base)[10])
@@ -3435,30 +3547,41 @@ erts_dsig_send(ErtsDSigSendContext *ctx)
 ctx->obuf = &ctx->obuf[fragments];
 }
 
-		if (!ctx->no_suspend) {
- qflgs = erts_atomic32_read_nob(&dep->qflgs);
-		 if (!(qflgs & ERTS_DE_QFLG_BUSY)) {
-			if (suspended)
-			 resume = 1; /* was busy when we started, but isn't now */
+		if ((qflgs & ERTS_DE_QFLG_BUSY)
+ && !ctx->ignore_busy
+ && !ctx->no_suspend) {
+
+ erts_mtx_unlock(&dep->qlock);
+
+ plp = erts_proclist_create(ctx->c_p);
+
+ erts_suspend(ctx->c_p, ERTS_PROC_LOCK_MAIN, NULL);
+ suspended = 1;
+
+ erts_mtx_lock(&dep->qlock);
+
+ qflgs = de_qflags_read(dep);
+ if (qflgs & ERTS_DE_QFLG_BUSY) {
+ /* Enqueue suspended process on dist entry */
+ ASSERT(plp);
+ erts_proclist_store_last(&dep->suspended, plp);
+ }
+ else {
+ resume = 1; /* was busy, but isn't now */
 #ifdef USE_VM_PROBES
-			if (resume && DTRACE_ENABLED(dist_port_not_busy)) {
-			 DTRACE_CHARBUF(port_str, 64);
-			 DTRACE_CHARBUF(remote_str, 64);
-
-			 erts_snprintf(port_str, sizeof(DTRACE_CHARBUF_NAME(port_str)),
-					 "%T", cid);
-			 erts_snprintf(remote_str, sizeof(DTRACE_CHARBUF_NAME(remote_str)),
-					 "%T", dep->sysname);
-			 DTRACE3(dist_port_not_busy, erts_this_node_sysname,
-				 port_str, remote_str);
-			}
+ if (resume && DTRACE_ENABLED(dist_port_not_busy)) {
+ DTRACE_CHARBUF(port_str, 64);
+ DTRACE_CHARBUF(remote_str, 64);
+
+ erts_snprintf(port_str, sizeof(DTRACE_CHARBUF_NAME(port_str)),
+ "%T", cid);
+ erts_snprintf(remote_str, sizeof(DTRACE_CHARBUF_NAME(remote_str)),
+ "%T", dep->sysname);
+ DTRACE3(dist_port_not_busy, erts_this_node_sysname,
+ port_str, remote_str);
+ }
 #endif
-		 }
-		 else {
-			/* Enqueue suspended process on dist entry */
-			ASSERT(plp);
-			erts_proclist_store_last(&dep->suspended, plp);
-		 }
+ }
 		}
 
 		erts_mtx_unlock(&dep->qlock);
@@ -3676,13 +3799,64 @@ dist_port_commandv(Port *prt, ErtsDistOutputBuf *obuf)
 ? ((Sint) 1) \
 : ((((Sint) (SZ)) >> 10) & ((Sint) ERTS_PORT_REDS_MASK__)))
 
+#ifndef DEBUG
+#define ERTS_DBG_CHK_DIST_QSIZE(DEP, PRT)
+#else
+#define ERTS_DBG_CHK_DIST_QSIZE(DEP, PRT) \
+ dbg_check_dist_qsize((DEP), (PRT))
+
+static void
+dbg_check_dist_qsize(DistEntry *dep, Port *prt)
+{
+ int ix;
+ Sint sz = 0, isz = 0, tqsz, qsz;
+ ErtsDistOutputBuf *qs[2];
+
+ ERTS_LC_ASSERT(dep && erts_lc_mtx_is_locked(&dep->qlock));
+ ASSERT(prt && erts_lc_is_port_locked(prt));
+ ERTS_LC_ASSERT((erts_atomic32_read_nob(&prt->sched.flags)
+ & ERTS_PTS_FLG_EXIT)
+ || prt->common.id == dep->cid);
+
+ tqsz = erts_atomic_read_nob(&dep->total_qsize);
+ qsz = erts_atomic_read_nob(&dep->qsize);
+
+ ASSERT(tqsz >= 0);
+ ASSERT(qsz >= 0);
+ ASSERT(tqsz >= qsz);
+
+ qs[0] = dep->out_queue.first;
+ qs[1] = dep->finalized_out_queue.first;
+
+ for (ix = 0; ix < sizeof(qs)/sizeof(qs[0]); ix++) {
+ ErtsDistOutputBuf *obuf = qs[ix];
+ while (obuf) {
+ add_obuf_sizes(obuf, &sz, &isz);
+ obuf = obuf->next;
+ }
+ }
+
+ ASSERT(tqsz == sz);
+ ASSERT(qsz == sz - isz);
+}
+
+#endif
+
 int
 erts_dist_command(Port *prt, int initial_reds)
 {
 Sint reds = initial_reds - ERTS_PORT_REDS_DIST_CMD_START;
 enum dist_entry_state state;
 Uint64 flags;
- Sint qsize, obufsize = 0;
+ /*
+ * 'obufsize' and 'ignore_obufsize' contains the number of bytes removed
+ * from the queue which will be updated (in dep->total_qsize and
+ * dep->qsize) before we return from this function. Note that
+ * 'obufsize' and 'ignore_obufsize' may be negative if we added to the
+ * queue size. This may occur since finalization of a buffer may increase
+ * buffer size.
+ */
+ Sint qsize, obufsize = 0, ignore_obufsize = 0;
 ErtsDistOutputQueue oq, foq;
 DistEntry *dep = (DistEntry*) erts_prtsd_get(prt, ERTS_PRTSD_DIST_ENTRY);
 Uint (*send)(Port *prt, ErtsDistOutputBuf *obuf);
@@ -3718,6 +3892,7 @@ erts_dist_command(Port *prt, int initial_reds)
 */
 
 erts_mtx_lock(&dep->qlock);
+ ERTS_DBG_CHK_DIST_QSIZE(dep, prt);
 oq.first = dep->out_queue.first;
 oq.last = dep->out_queue.last;
 dep->out_queue.first = NULL;
@@ -3729,23 +3904,6 @@ erts_dist_command(Port *prt, int initial_reds)
 dep->finalized_out_queue.first = NULL;
 dep->finalized_out_queue.last = NULL;
 
-#ifdef DEBUG
- {
- Uint sz = 0;
- ErtsDistOutputBuf *curr = oq.first;
- while (curr) {
- sz += size_obuf(curr);
- curr = curr->next;
- }
- curr = foq.first;
- while (curr) {
- sz += size_obuf(curr);
- curr = curr->next;
- }
- ASSERT(sz <= erts_atomic_read_nob(&dep->qsize));
- }
-#endif
-
 sched_flags = erts_atomic32_read_nob(&prt->sched.flags);
 
 if (reds < 0)
@@ -3756,7 +3914,7 @@ erts_dist_command(Port *prt, int initial_reds)
 	do {
 Uint size;
 ErtsDistOutputBuf *fob;
-	 obufsize += size_obuf(foq.first);
+ add_obuf_sizes(foq.first, &obufsize, &ignore_obufsize);
 size = (*send)(prt, foq.first);
 erts_atomic64_inc_nob(&dep->out);
 esdp->io.out += (Uint64) size;
@@ -3784,9 +3942,9 @@ erts_dist_command(Port *prt, int initial_reds)
 	 ob = oq.first;
 	 ASSERT(ob);
 	 do {
- obufsize += size_obuf(ob);
+ add_obuf_sizes(ob, &obufsize, &ignore_obufsize);
 		reds = erts_encode_ext_dist_header_finalize(ob, dep, flags, reds);
- obufsize -= size_obuf(ob);
+ subtract_obuf_sizes(ob, &obufsize, &ignore_obufsize);
 if (reds < 0)
 break; /* finalize needs to be restarted... */
 last_finalized = ob;
@@ -3824,12 +3982,11 @@ erts_dist_command(Port *prt, int initial_reds)
 	int preempt = 0;
 	while (oq.first && !preempt) {
 	 ErtsDistOutputBuf *fob;
-	 Uint size, obsz;
- obufsize += size_obuf(oq.first);
+	 Uint size;
+ add_obuf_sizes(oq.first, &obufsize, &ignore_obufsize);
 reds = erts_encode_ext_dist_header_finalize(oq.first, dep, flags, reds);
- obsz = size_obuf(oq.first);
- obufsize -= obsz;
 if (reds < 0) { /* finalize needs to be restarted... */
+ subtract_obuf_sizes(oq.first, &obufsize, &ignore_obufsize);
 preempt = 1;
 break;
 }
@@ -3838,7 +3995,6 @@ erts_dist_command(Port *prt, int initial_reds)
 	 esdp->io.out += (Uint64) size;
 	 reds -= ERTS_PORT_REDS_DIST_CMD_DATA(size);
 	 fob = oq.first;
-	 obufsize += obsz;
 	 oq.first = oq.first->next;
 	 free_dist_obuf(fob, !0);
 	 sched_flags = erts_atomic32_read_nob(&prt->sched.flags);
@@ -3869,13 +4025,12 @@ erts_dist_command(Port *prt, int initial_reds)
 	 * processes.
 	 */
 	erts_mtx_lock(&dep->qlock);
- de_busy = !!(erts_atomic32_read_nob(&dep->qflgs) & ERTS_DE_QFLG_BUSY);
- qsize = (Sint) erts_atomic_add_read_nob(&dep->qsize,
- (erts_aint_t) -obufsize);
-	ASSERT(qsize >= 0);
-	obufsize = 0;
+ de_busy = !!(de_qflags_read(dep) & ERTS_DE_QFLG_BUSY);
+ update_qsizes(dep, NULL, &qsize, -obufsize, -ignore_obufsize);
+	obufsize = ignore_obufsize = 0;
 	if (!(sched_flags & ERTS_PTS_FLG_BUSY_PORT)
-	 && de_busy && qsize < erts_dist_buf_busy_limit) {
+	 && de_busy
+ && qsize < erts_dist_buf_busy_limit) {
 	 int resumed;
 	 ErtsProcList *suspendees = get_suspended_on_de(dep, ERTS_DE_QFLG_BUSY);
 	 erts_mtx_unlock(&dep->qlock);
@@ -3891,17 +4046,7 @@ erts_dist_command(Port *prt, int initial_reds)
 
 done:
 
- if (obufsize != 0) {
-	erts_mtx_lock(&dep->qlock);
-#ifdef DEBUG
- qsize = (Sint) erts_atomic_add_read_nob(&dep->qsize,
- (erts_aint_t) -obufsize);
-	ASSERT(qsize >= 0);
-#else
- erts_atomic_add_nob(&dep->qsize, (erts_aint_t) -obufsize);
-#endif
-	erts_mtx_unlock(&dep->qlock);
- }
+ ASSERT(!ignore_obufsize || obufsize);
 
 ASSERT(!!foq.first == !!foq.last);
 ASSERT(!dep->finalized_out_queue.first);
@@ -3912,7 +4057,21 @@ erts_dist_command(Port *prt, int initial_reds)
 	dep->finalized_out_queue.last = foq.last;
 }
 
- /* Avoid wrapping reduction counter... */
+ if (obufsize != 0) {
+	erts_mtx_lock(&dep->qlock);
+ update_qsizes(dep, NULL, NULL, -obufsize, -ignore_obufsize);
+ ERTS_DBG_CHK_DIST_QSIZE(dep, prt);
+	erts_mtx_unlock(&dep->qlock);
+ }
+#ifdef DEBUG
+ else {
+ erts_mtx_lock(&dep->qlock);
+ ERTS_DBG_CHK_DIST_QSIZE(dep, prt);
+	erts_mtx_unlock(&dep->qlock);
+ }
+#endif
+
+ /* Avoid wrapping reduction counter... */
 if (reds < INT_MIN/2)
 	reds = INT_MIN/2;
 
@@ -3942,7 +4101,7 @@ erts_dist_command(Port *prt, int initial_reds)
 	while (oq.first) {
 	 ErtsDistOutputBuf *fob = oq.first;
 	 oq.first = oq.first->next;
-	 obufsize += size_obuf(fob);
+ add_obuf_sizes(fob, &obufsize, &ignore_obufsize);
 	 free_dist_obuf(fob, !0);
 	}
 
@@ -3951,14 +4110,15 @@ erts_dist_command(Port *prt, int initial_reds)
 }
 else {
 	if (oq.first) {
+	 erts_mtx_lock(&dep->qlock);
+ update_qsizes(dep, NULL, NULL, -obufsize, -ignore_obufsize);
+	 obufsize = ignore_obufsize = 0;
+
 	 /*
-	 * Unhandle buffers need to be put back first
+	 * Unhandled buffers need to be put back first
 	 * in out_queue.
 	 */
-	 erts_mtx_lock(&dep->qlock);
-	 erts_atomic_add_nob(&dep->qsize, -obufsize);
-	 obufsize = 0;
-	 oq.last->next = dep->out_queue.first;
+ oq.last->next = dep->out_queue.first;
 	 dep->out_queue.first = oq.first;
 	 if (!dep->out_queue.last)
 		dep->out_queue.last = oq.last;
@@ -3974,7 +4134,7 @@ BIF_RETTYPE
 dist_ctrl_get_data_notification_1(BIF_ALIST_1)
 {
 DistEntry *dep = ERTS_PROC_GET_DIST_ENTRY(BIF_P);
- erts_aint32_t qflgs;
+ erts_aint32_t notify;
 erts_aint_t qsize;
 Eterm receiver = NIL;
 Uint32 conn_id;
@@ -3987,7 +4147,7 @@ dist_ctrl_get_data_notification_1(BIF_ALIST_1)
 
 /*
 * Caller is the only one that can consume from this queue
- * and the only one that can set the req-info flag...
+ * and the only one that can set the notify field...
 */
 
 erts_de_rlock(dep);
@@ -3999,23 +4159,21 @@ dist_ctrl_get_data_notification_1(BIF_ALIST_1)
 
 ASSERT(dep->cid == BIF_P->common.id);
 
- qflgs = erts_atomic32_read_nob(&dep->qflgs);
+ notify = erts_atomic32_read_nob(&dep->notify);
 
- if (!(qflgs & ERTS_DE_QFLG_REQ_INFO)) {
+ if (!notify) {
 ERTS_THR_READ_MEMORY_BARRIER;
- qsize = erts_atomic_read_nob(&dep->qsize);
+ qsize = erts_atomic_read_nob(&dep->total_qsize);
 ASSERT(qsize >= 0);
 if (qsize > 0)
 receiver = BIF_P->common.id; /* Notify ourselves... */
- else { /* Empty queue; set req-info flag... */
- qflgs = erts_atomic32_read_bor_mb(&dep->qflgs,
- ERTS_DE_QFLG_REQ_INFO);
- qsize = erts_atomic_read_nob(&dep->qsize);
+ else { /* Empty queue; set the notify field... */
+ notify = erts_atomic32_xchg_mb(&dep->notify, (erts_aint32_t) !0);
+ qsize = erts_atomic_read_nob(&dep->total_qsize);
 ASSERT(qsize >= 0);
 if (qsize > 0) {
- qflgs = erts_atomic32_read_band_mb(&dep->qflgs,
- ~ERTS_DE_QFLG_REQ_INFO);
- if (qflgs & ERTS_DE_QFLG_REQ_INFO)
+ notify = erts_atomic32_xchg_mb(&dep->notify, (erts_aint32_t) 0);
+ if (notify)
 receiver = BIF_P->common.id; /* Notify ourselves... */
 /* else: someone else will notify us... */
 }
@@ -4203,7 +4361,7 @@ dist_get_stat_1(BIF_ALIST_1)
 }
 read = (Sint64) erts_atomic64_read_nob(&dep->in);
 write = (Sint64) erts_atomic64_read_nob(&dep->out);
- pend = (Sint64) erts_atomic_read_nob(&dep->qsize);
+ pend = (Sint64) erts_atomic_read_nob(&dep->total_qsize);
 
 erts_de_runlock(dep);
 
@@ -4258,10 +4416,19 @@ dist_ctrl_get_data_1(BIF_ALIST_1)
 {
 DistEntry *dep = ERTS_PROC_GET_DIST_ENTRY(BIF_P);
 const Sint initial_reds = ERTS_BIF_REDS_LEFT(BIF_P);
- Sint reds = initial_reds, obufsize = 0, ix, vlen;
+ Sint reds = initial_reds, ix, vlen;
+ /*
+ * 'obufsize' and 'ignore_obufsize' contains the number of bytes removed
+ * from the queue which will be updated (in dep->total_qsize and
+ * dep->qsize) before we return from this function. Note that
+ * 'obufsize' and 'ignore_obufsize' may be negative if we added to the
+ * queue size. This may occur since finalization of a buffer may increase
+ * buffer size.
+ */
+ Sint obufsize = 0, ignore_obufsize = 0;
 ErtsDistOutputBuf *obuf;
 Eterm *hp, res;
- erts_aint_t qsize;
+ Sint qsize;
 Uint32 conn_id, get_size;
 Uint hsz = 0, data_sz;
 SysIOVec *iov;
@@ -4300,7 +4467,7 @@ dist_ctrl_get_data_1(BIF_ALIST_1)
 {
 if (!dep->tmp_out_queue.first) {
 ASSERT(!dep->tmp_out_queue.last);
- qsize = erts_atomic_read_acqb(&dep->qsize);
+ qsize = (Sint) erts_atomic_read_acqb(&dep->total_qsize);
 if (qsize > 0) {
 erts_mtx_lock(&dep->qlock);
 dep->tmp_out_queue.first = dep->out_queue.first;
@@ -4319,13 +4486,16 @@ dist_ctrl_get_data_1(BIF_ALIST_1)
 }
 
 obuf = dep->tmp_out_queue.first;
- obufsize += size_obuf(obuf);
+ add_obuf_sizes(obuf, &obufsize, &ignore_obufsize);
 reds = erts_encode_ext_dist_header_finalize(obuf, dep, dep->dflags, reds);
- obufsize -= size_obuf(obuf);
+ subtract_obuf_sizes(obuf, &obufsize, &ignore_obufsize);
 if (reds < 0) { /* finalize needs to be restarted... */
 erts_de_runlock(dep);
- if (obufsize)
- erts_atomic_add_nob(&dep->qsize, (erts_aint_t) -obufsize);
+ if (obufsize) {
+ erts_mtx_lock(&dep->qlock);
+ update_qsizes(dep, NULL, NULL, -obufsize, -ignore_obufsize);
+ erts_mtx_unlock(&dep->qlock);
+ }
 ERTS_BIF_YIELD1(BIF_TRAP_EXPORT(BIF_dist_ctrl_get_data_1),
 BIF_P, BIF_ARG_1);
 }
@@ -4406,16 +4576,18 @@ dist_ctrl_get_data_1(BIF_ALIST_1)
 hp += 2;
 }
 
- obufsize += size_obuf(obuf);
+ add_obuf_sizes(obuf, &obufsize, &ignore_obufsize);
 
- qsize = erts_atomic_add_read_nob(&dep->qsize, (erts_aint_t) -obufsize);
+ erts_mtx_lock(&dep->qlock);
 
- ASSERT(qsize >= 0);
+ update_qsizes(dep, NULL, &qsize, -obufsize, -ignore_obufsize);
 
- if (qsize < erts_dist_buf_busy_limit/2
- && (erts_atomic32_read_acqb(&dep->qflgs) & ERTS_DE_QFLG_BUSY)) {
+ if (qsize >= erts_dist_buf_busy_limit/2
+ || !(de_qflags_read(dep) & ERTS_DE_QFLG_BUSY)) {
+ erts_mtx_unlock(&dep->qlock);
+ }
+ else {
 ErtsProcList *resume_procs = NULL;
- erts_mtx_lock(&dep->qlock);
 resume_procs = get_suspended_on_de(dep, ERTS_DE_QFLG_BUSY);
 erts_mtx_unlock(&dep->qlock);
 if (resume_procs) {
@@ -4470,8 +4642,8 @@ static void kill_connection(DistEntry *dep)
 
 dep->state = ERTS_DE_STATE_EXITING;
 erts_mtx_lock(&dep->qlock);
- ASSERT(!(erts_atomic32_read_nob(&dep->qflgs) & ERTS_DE_QFLG_EXIT));
- erts_atomic32_read_bor_nob(&dep->qflgs, ERTS_DE_QFLG_EXIT);
+ ASSERT(!(de_qflags_read(dep) & ERTS_DE_QFLG_EXIT));
+ de_qflags_read_set(dep, ERTS_DE_QFLG_EXIT);
 erts_mtx_unlock(&dep->qlock);
 
 if (is_internal_port(dep->cid))
@@ -5071,7 +5243,6 @@ setup_connection_epiloge_rwunlock(Process *c_p, DistEntry *dep,
 Process *net_kernel)
 {
 Eterm notify_proc = NIL;
- erts_aint32_t qflgs;
 ErtsProcLocks nk_locks;
 int success = 0;
 
@@ -5107,17 +5278,18 @@ setup_connection_epiloge_rwunlock(Process *c_p, DistEntry *dep,
 erts_set_dist_entry_connected(dep, ctrlr, flags);
 
 notify_proc = NIL;
- if (erts_atomic_read_nob(&dep->qsize)) {
+ if (erts_atomic_read_nob(&dep->total_qsize)) {
 if (is_internal_port(dep->cid)) {
 erts_schedule_dist_command(NULL, dep);
 }
 else {
+ erts_aint32_t notify;
 ERTS_THR_READ_MEMORY_BARRIER;
- qflgs = erts_atomic32_read_nob(&dep->qflgs);
- if (qflgs & ERTS_DE_QFLG_REQ_INFO) {
- qflgs = erts_atomic32_read_band_mb(&dep->qflgs,
- ~ERTS_DE_QFLG_REQ_INFO);
- if (qflgs & ERTS_DE_QFLG_REQ_INFO) {
+ notify = erts_atomic32_read_nob(&dep->notify);
+ if (notify) {
+ notify = erts_atomic32_xchg_mb(&dep->notify,
+ (erts_aint32_t) 0);
+ if (notify) {
 notify_proc = dep->cid;
 ASSERT(is_internal_pid(notify_proc));
 }
@@ -5337,6 +5509,7 @@ Sint erts_abort_pending_connection_rwunlock(DistEntry* dep,
 ASSERT(!dep->finalized_out_queue.first);
 resume_procs = get_suspended_on_de(dep, ERTS_DE_QFLGS_ALL);
 	erts_mtx_unlock(&dep->qlock);
+ erts_atomic32_set_relb(&dep->notify, 0);
 	erts_atomic_set_nob(&dep->dist_cmd_scheduled, 0);
 	dep->send = NULL;
 
diff --git a/erts/emulator/beam/dist.h b/erts/emulator/beam/dist.h
index 9d10bbcd36..7dabdd4961 100644
--- a/erts/emulator/beam/dist.h
+++ b/erts/emulator/beam/dist.h
@@ -342,6 +342,7 @@ typedef struct erts_dsig_send_context {
 int connect;
 int no_suspend;
 int no_trap;
+ int ignore_busy;
 
 Eterm ctl;
 Eterm msg;
diff --git a/erts/emulator/beam/erl_bif_info.c b/erts/emulator/beam/erl_bif_info.c
index 6ae5f3ce57..8f8ff989d5 100644
--- a/erts/emulator/beam/erl_bif_info.c
+++ b/erts/emulator/beam/erl_bif_info.c
@@ -777,6 +777,7 @@ collect_one_suspend_monitor(ErtsMonitor *mon, void *vsmicp, Sint reds)
 #define ERTS_PI_IX_MAGIC_REF 34
 #define ERTS_PI_IX_FULLSWEEP_AFTER 35
 #define ERTS_PI_IX_PARENT 36
+#define ERTS_PI_IX_ASYNC_DIST 37
 
 #define ERTS_PI_FLAG_SINGELTON (1 << 0)
 #define ERTS_PI_FLAG_ALWAYS_WRAP (1 << 1)
@@ -833,7 +834,8 @@ static ErtsProcessInfoArgs pi_args[] = {
 {am_garbage_collection_info, ERTS_PROCESS_GC_INFO_MAX_SIZE, 0, ERTS_PROC_LOCK_MAIN},
 {am_magic_ref, 0, ERTS_PI_FLAG_FORCE_SIG_SEND, ERTS_PROC_LOCK_MAIN},
 {am_fullsweep_after, 0, 0, ERTS_PROC_LOCK_MAIN},
- {am_parent, 0, 0, ERTS_PROC_LOCK_MAIN}
+ {am_parent, 0, 0, ERTS_PROC_LOCK_MAIN},
+ {am_async_dist, 0, 0, ERTS_PROC_LOCK_MAIN}
 };
 
 #define ERTS_PI_ARGS ((int) (sizeof(pi_args)/sizeof(pi_args[0])))
@@ -954,6 +956,8 @@ pi_arg2ix(Eterm arg)
 return ERTS_PI_IX_FULLSWEEP_AFTER;
 case am_parent:
 return ERTS_PI_IX_PARENT;
+ case am_async_dist:
+ return ERTS_PI_IX_ASYNC_DIST;
 default:
 return -1;
 }
@@ -2125,6 +2129,10 @@ process_info_aux(Process *c_p,
 }
 break;
 
+ case ERTS_PI_IX_ASYNC_DIST:
+ res = (rp->flags & F_ASYNC_DIST) ? am_true : am_false;
+ break;
+
 case ERTS_PI_IX_MAGIC_REF: {
 	Uint sz = 0;
 	(void) bld_magic_ref_bin_list(NULL, &sz, &MSO(rp));
@@ -2781,6 +2789,10 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	res = new_binary(BIF_P, (byte *) dsbufp->str, dsbufp->str_len);
 	erts_destroy_info_dsbuf(dsbufp);
 	BIF_RET(res);
+ } else if (am_async_dist == BIF_ARG_1) {
+ BIF_RET((erts_default_spo_flags & SPO_ASYNC_DIST)
+ ? am_true
+ : am_false);
 } else if (ERTS_IS_ATOM_STR("dist_ctrl", BIF_ARG_1)) {
 	DistEntry *dep;
 	i = 0;
diff --git a/erts/emulator/beam/erl_init.c b/erts/emulator/beam/erl_init.c
index 95e41226fc..ab3823d51a 100644
--- a/erts/emulator/beam/erl_init.c
+++ b/erts/emulator/beam/erl_init.c
@@ -663,6 +663,7 @@ void erts_usage(void)
 erts_fprintf(stderr, "\n");
 
 erts_fprintf(stderr, "-pc <set> control what characters are considered printable (default latin1)\n");
+ erts_fprintf(stderr, "-pad bool set default process async data (default false)\n");
 erts_fprintf(stderr, "-P number set maximum number of processes on this node;\n");
 erts_fprintf(stderr, " valid range is [%d-%d]\n",
 		 ERTS_MIN_PROCESSES, ERTS_MAX_PROCESSES);
@@ -1417,11 +1418,28 @@ erl_start(int argc, char **argv)
 		 erts_usage();
 		}
 		erts_set_printable_characters(printable_chars);
-		break;
-	 } else {
-		erts_fprintf(stderr, "%s unknown flag %s\n", argv[0], argv[i]);
-		erts_usage();
 	 }
+ else {
+ char *sub_param = argv[i]+2;
+ if (has_prefix("ad", sub_param)) {
+ arg = get_arg(sub_param+2, argv[i+1], &i);
+ if (sys_strcmp("true", arg) == 0) {
+ erts_default_spo_flags |= SPO_ASYNC_DIST;
+ }
+ else if (sys_strcmp("false", arg) == 0) {
+ erts_default_spo_flags &= ~SPO_ASYNC_DIST;
+ }
+ else {
+ erts_fprintf(stderr, "bad async dist value %s\n", arg);
+ erts_usage();
+ }
+ }
+ else {
+ erts_fprintf(stderr, "%s unknown flag %s\n", argv[0], argv[i]);
+ erts_usage();
+ }
+ }
+ break;
 	case 'f':
 	 if (!sys_strncmp(argv[i],"-fn",3)) {
 		int warning_type = ERL_FILENAME_WARNING_WARNING;
diff --git a/erts/emulator/beam/erl_node_tables.c b/erts/emulator/beam/erl_node_tables.c
index 23a19064b1..2bec8ff20e 100644
--- a/erts/emulator/beam/erl_node_tables.c
+++ b/erts/emulator/beam/erl_node_tables.c
@@ -191,6 +191,8 @@ dist_table_alloc(void *dep_tmpl)
 erts_mtx_init(&dep->qlock, "dist_entry_out_queue", sysname,
 ERTS_LOCK_FLAGS_CATEGORY_DISTRIBUTION);
 erts_atomic32_init_nob(&dep->qflgs, 0);
+ erts_atomic32_init_nob(&dep->notify, 0);
+ erts_atomic_init_nob(&dep->total_qsize, 0);
 erts_atomic_init_nob(&dep->qsize, 0);
 erts_atomic64_init_nob(&dep->in, 0);
 erts_atomic64_init_nob(&dep->out, 0);
@@ -729,8 +731,6 @@ erts_set_dist_entry_pending(DistEntry *dep)
 void
 erts_set_dist_entry_connected(DistEntry *dep, Eterm cid, Uint64 flags)
 {
- erts_aint32_t set_qflgs;
-
 ASSERT(dep->mld);
 
 ERTS_LC_ASSERT(erts_lc_is_de_rwlocked(dep));
@@ -767,9 +767,6 @@ erts_set_dist_entry_connected(DistEntry *dep, Eterm cid, Uint64 flags)
 
 erts_atomic64_set_nob(&dep->in, 0);
 erts_atomic64_set_nob(&dep->out, 0);
- set_qflgs = (is_internal_port(cid) ?
- ERTS_DE_QFLG_PORT_CTRL : ERTS_DE_QFLG_PROC_CTRL);
- erts_atomic32_read_bor_nob(&dep->qflgs, set_qflgs);
 
 if(flags & DFLAG_PUBLISHED) {
 	dep->next = erts_visible_dist_entries;
diff --git a/erts/emulator/beam/erl_node_tables.h b/erts/emulator/beam/erl_node_tables.h
index f8c448de3c..9109c50e1d 100644
--- a/erts/emulator/beam/erl_node_tables.h
+++ b/erts/emulator/beam/erl_node_tables.h
@@ -77,15 +77,9 @@ enum dist_entry_state {
 
 #define ERTS_DE_QFLG_BUSY			(((erts_aint32_t) 1) << 0)
 #define ERTS_DE_QFLG_EXIT			(((erts_aint32_t) 1) << 1)
-#define ERTS_DE_QFLG_REQ_INFO			(((erts_aint32_t) 1) << 2)
-#define ERTS_DE_QFLG_PORT_CTRL (((erts_aint32_t) 1) << 3)
-#define ERTS_DE_QFLG_PROC_CTRL (((erts_aint32_t) 1) << 4)
 
 #define ERTS_DE_QFLGS_ALL			(ERTS_DE_QFLG_BUSY \
-						 | ERTS_DE_QFLG_EXIT \
- | ERTS_DE_QFLG_REQ_INFO \
- | ERTS_DE_QFLG_PORT_CTRL \
- | ERTS_DE_QFLG_PROC_CTRL)
+						 | ERTS_DE_QFLG_EXIT)
 
 #if defined(ARCH_64)
 #define ERTS_DIST_OUTPUT_BUF_DBG_PATTERN ((Uint) 0xf713f713f713f713UL)
@@ -105,6 +99,7 @@ struct ErtsDistOutputBuf_ {
 * iov[2 ... vsize-1] data
 */
 ErlIOVec *eiov;
+ int ignore_busy;
 };
 
 struct ErtsDistOutputBufsContainer_ {
@@ -153,7 +148,9 @@ struct dist_entry_ {
 
 erts_mtx_t qlock; /* Protects qflgs and out_queue */
 erts_atomic32_t qflgs;
- erts_atomic_t qsize;
+ erts_atomic32_t notify; /* User wants queue notification? */
+ erts_atomic_t qsize; /* Size of data in queue respecting busy dist */
+ erts_atomic_t total_qsize; /* Total size of data in queue */
 erts_atomic64_t in;
 erts_atomic64_t out;
 ErtsDistOutputQueue out_queue;
diff --git a/erts/emulator/beam/erl_process.c b/erts/emulator/beam/erl_process.c
index 3e56d6c62b..15d4be3755 100644
--- a/erts/emulator/beam/erl_process.c
+++ b/erts/emulator/beam/erl_process.c
@@ -12030,6 +12030,22 @@ erts_parse_spawn_opts(ErlSpawnOpts *sop, Eterm opts_list, Eterm *tag,
 		 sop->priority = PRIORITY_LOW;
 		else
 result = -1;
+ } else if (arg == am_async_dist) {
+ if (val == am_true) {
+ if (sop->flags & SPO_ASYNC_DIST)
+ sop->multi_set = !0;
+ else
+ sop->flags |= SPO_ASYNC_DIST;
+ }
+ else if (val == am_false) {
+ if (!(sop->flags & SPO_ASYNC_DIST))
+ sop->multi_set = !0;
+ else
+ sop->flags &= ~SPO_ASYNC_DIST;
+ }
+ else {
+ result = -1;
+ }
 	 } else if (arg == am_message_queue_data) {
 if (sop->flags & (SPO_OFF_HEAP_MSGQ|SPO_ON_HEAP_MSGQ))
 sop->multi_set = !0;
@@ -12267,6 +12283,9 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
 /* Reserve place for continuation pointer, redzone, etc */
 heap_need = arg_size + S_RESERVED;
 
+ if (so->flags & SPO_ASYNC_DIST)
+ flags |= F_ASYNC_DIST;
+
 p->flags = flags;
 p->sig_qs.flags = qs_flags;
 
diff --git a/erts/emulator/beam/erl_process.h b/erts/emulator/beam/erl_process.h
index 949727d950..ac70260d0d 100644
--- a/erts/emulator/beam/erl_process.h
+++ b/erts/emulator/beam/erl_process.h
@@ -1410,8 +1410,9 @@ void erts_check_for_holes(Process* p);
 #define SPO_IX_ASYNC 11
 #define SPO_IX_NO_SMSG 12
 #define SPO_IX_NO_EMSG 13
+#define SPO_IX_ASYNC_DIST 14
 
-#define SPO_NO_INDICES (SPO_IX_ASYNC+1)
+#define SPO_NO_INDICES (SPO_IX_ASYNC_DIST+1)
 
 #define SPO_LINK (1 << SPO_IX_LINK)
 #define SPO_MONITOR (1 << SPO_IX_MONITOR)
@@ -1427,8 +1428,9 @@ void erts_check_for_holes(Process* p);
 #define SPO_ASYNC (1 << SPO_IX_ASYNC)
 #define SPO_NO_SMSG (1 << SPO_IX_NO_SMSG)
 #define SPO_NO_EMSG (1 << SPO_IX_NO_EMSG)
+#define SPO_ASYNC_DIST (1 << SPO_IX_ASYNC_DIST)
 
-#define SPO_MAX_FLAG SPO_NO_EMSG
+#define SPO_MAX_FLAG SPO_ASYNC_DIST
 
 #define SPO_USE_ARGS \
 (SPO_MIN_HEAP_SIZE \
@@ -1570,6 +1572,7 @@ extern int erts_system_profile_ts_type;
 #define F_FRAGMENTED_SEND (1 << 23) /* Process is doing a distributed fragmented send */
 #define F_DBG_FORCED_TRAP (1 << 24) /* DEBUG: Last BIF call was a forced trap */
 #define F_DIRTY_CHECK_CLA (1 << 25) /* Check if copy literal area GC scheduled */
+#define F_ASYNC_DIST (1 << 26) /* Truly asynchronous distribution */
 
 /* Signal queue flags */
 #define FS_OFF_HEAP_MSGQ (1 << 0) /* Off heap msg queue */
diff --git a/erts/emulator/test/distribution_SUITE.erl b/erts/emulator/test/distribution_SUITE.erl
index edab0be20b..c1a31c6abf 100644
--- a/erts/emulator/test/distribution_SUITE.erl
+++ b/erts/emulator/test/distribution_SUITE.erl
@@ -81,6 +81,9 @@
 hopefull_data_encoding/1,
 hopefull_export_fun_bug/1,
 huge_iovec/1,
+ async_dist_flag/1,
+ async_dist_port_dctrlr/1,
+ async_dist_proc_dctrlr/1,
 creation_selection/1,
 creation_selection_test/1]).
 
@@ -114,7 +117,8 @@ all() ->
 dist_entry_refc_race,
 start_epmd_false, no_epmd, epmd_module, system_limit,
 hopefull_data_encoding, hopefull_export_fun_bug,
- huge_iovec, creation_selection].
+ huge_iovec,
+ {group, async_dist}, creation_selection].
 
 groups() ->
 [{bulk_send, [], [bulk_send_small, bulk_send_big, bulk_send_bigbig]},
@@ -133,7 +137,11 @@ groups() ->
 [message_latency_large_message,
 message_latency_large_link_exit,
 message_latency_large_monitor_exit,
- message_latency_large_exit2]}
+ message_latency_large_exit2]},
+ {async_dist, [],
+ [async_dist_flag,
+ async_dist_port_dctrlr,
+ async_dist_proc_dctrlr]}
 ].
 
 init_per_suite(Config) ->
@@ -3088,6 +3096,271 @@ derr_sender(Main, Nodes) ->
 Main ! count,
 derr_sender(Main, Nodes).
 
+async_dist_flag(Config) when is_list(Config) ->
+ {ok, Peer1, Node1} = ?CT_PEER(),
+ async_dist_flag_test(Node1, false),
+ peer:stop(Peer1),
+ {ok, Peer2, Node2} = ?CT_PEER(["+pad", "false"]),
+ async_dist_flag_test(Node2, false),
+ peer:stop(Peer2),
+ {ok, Peer3, Node3} = ?CT_PEER(["+pad", "true", "+pad", "false"]),
+ async_dist_flag_test(Node3, false),
+ peer:stop(Peer3),
+
+ {ok, Peer4, Node4} = ?CT_PEER(["+pad", "true"]),
+ async_dist_flag_test(Node4, true),
+ peer:stop(Peer4),
+ {ok, Peer5, Node5} = ?CT_PEER(["+pad", "false", "+pad", "true"]),
+ async_dist_flag_test(Node5, true),
+ peer:stop(Peer5),
+
+ ok.
+
+async_dist_flag_test(Node, Default) when is_atom(Node), is_boolean(Default) ->
+ Tester = self(),
+ NotDefault = not Default,
+
+ Default = erpc:call(Node, erlang, system_info, [async_dist]),
+
+ {P1, M1} = spawn_opt(Node, fun () ->
+ receive after infinity -> ok end
+ end, [link, monitor]),
+ {P2, M2} = spawn_opt(Node, fun () ->
+ receive after infinity -> ok end
+ end, [link, monitor, {async_dist, false}]),
+ {P3, M3} = spawn_opt(Node, fun () ->
+ receive after infinity -> ok end
+ end, [link, monitor, {async_dist, true}]),
+ {async_dist, Default} = erpc:call(Node, erlang, process_info, [P1, async_dist]),
+ {async_dist, false} = erpc:call(Node, erlang, process_info, [P2, async_dist]),
+ {async_dist, true} = erpc:call(Node, erlang, process_info, [P3, async_dist]),
+
+ R4 = make_ref(),
+ {P4, M4} = spawn_opt(Node, fun () ->
+ Default = process_flag(async_dist, NotDefault),
+ Tester ! R4,
+ receive after infinity -> ok end
+ end, [link, monitor]),
+
+ R5 = make_ref(),
+ {P5, M5} = spawn_opt(Node, fun () ->
+ false = process_flag(async_dist, true),
+ Tester ! R5,
+ receive after infinity -> ok end
+ end, [link, monitor, {async_dist, false}]),
+ R6 = make_ref(),
+ {P6, M6} = spawn_opt(Node, fun () ->
+ true = process_flag(async_dist, false),
+ Tester ! R6,
+ receive after infinity -> ok end
+ end, [link, monitor, {async_dist, true}]),
+ receive R4 -> ok end,
+ {async_dist, NotDefault} = erpc:call(Node, erlang, process_info, [P4, async_dist]),
+ receive R5 -> ok end,
+ {async_dist, true} = erpc:call(Node, erlang, process_info, [P5, async_dist]),
+ receive R6 -> ok end,
+ {async_dist, false} = erpc:call(Node, erlang, process_info, [P6, async_dist]),
+
+
+ R7 = make_ref(),
+ {P7, M7} = spawn_opt(Node, fun () ->
+ Default = process_flag(async_dist, NotDefault),
+ NotDefault = process_flag(async_dist, NotDefault),
+ NotDefault = process_flag(async_dist, NotDefault),
+ NotDefault = process_flag(async_dist, Default),
+ Default = process_flag(async_dist, Default),
+ Default = process_flag(async_dist, Default),
+ Tester ! R7,
+ receive after infinity -> ok end
+ end, [link, monitor]),
+ receive R7 -> ok end,
+
+ unlink(P1),
+ exit(P1, bang),
+ unlink(P2),
+ exit(P2, bang),
+ unlink(P3),
+ exit(P3, bang),
+ unlink(P4),
+ exit(P4, bang),
+ unlink(P5),
+ exit(P5, bang),
+ unlink(P6),
+ exit(P6, bang),
+ unlink(P7),
+ exit(P7, bang),
+
+ receive {'DOWN', M1, process, P1, bang} -> ok end,
+ receive {'DOWN', M2, process, P2, bang} -> ok end,
+ receive {'DOWN', M3, process, P3, bang} -> ok end,
+ receive {'DOWN', M4, process, P4, bang} -> ok end,
+ receive {'DOWN', M5, process, P5, bang} -> ok end,
+ receive {'DOWN', M6, process, P6, bang} -> ok end,
+ receive {'DOWN', M7, process, P7, bang} -> ok end,
+
+ ok.
+
+async_dist_port_dctrlr(Config) when is_list(Config) ->
+ {ok, RecvPeer, RecvNode} = ?CT_PEER(),
+ ok = async_dist_test(RecvNode),
+ peer:stop(RecvPeer),
+ ok.
+
+async_dist_proc_dctrlr(Config) when is_list(Config) ->
+ {ok, SendPeer, SendNode} = ?CT_PEER(["-proto_dist", "gen_tcp"]),
+ {ok, RecvPeer, RecvNode} = ?CT_PEER(["-proto_dist", "gen_tcp"]),
+ {Pid, Mon} = spawn_monitor(SendNode,
+ fun () ->
+ ok = async_dist_test(RecvNode),
+ exit(test_success)
+ end),
+ receive
+ {'DOWN', Mon, process, Pid, Reason} ->
+ test_success = Reason
+ end,
+ peer:stop(SendPeer),
+ peer:stop(RecvPeer),
+ ok.
+
+async_dist_test(Node) ->
+ Scale = case round(test_server:timetrap_scale_factor()/3) of
+ S when S < 1 -> 1;
+ S -> S
+ end,
+ _ = process_flag(async_dist, false),
+ Tester = self(),
+ AliveReceiver1 = spawn_link(Node, fun () ->
+ register(alive_receiver_1, self()),
+ Tester ! {registered, self()},
+ receive after infinity -> ok end
+ end),
+ receive {registered, AliveReceiver1} -> ok end,
+ AliveReceiver2 = spawn(Node, fun () -> receive after infinity -> ok end end),
+ {AliveReceiver3, AR3Mon} = spawn_monitor(Node,
+ fun () ->
+ receive after infinity -> ok end
+ end),
+ {DeadReceiver, DRMon} = spawn_monitor(Node, fun () -> ok end),
+ receive
+ {'DOWN', DRMon, process, DeadReceiver, DRReason} ->
+ normal = DRReason
+ end,
+ Data = lists:duplicate($x, 256),
+ GoNuts = fun GN () ->
+ DeadReceiver ! hello,
+ GN()
+ end,
+ erpc:call(Node, erts_debug, set_internal_state, [available_internal_state, true]),
+ erpc:cast(Node, erts_debug, set_internal_state, [block, 4000*Scale]),
+ DistBufFiller = spawn_link(fun () ->
+ process_flag(async_dist, false),
+ receive go_nuts -> ok end,
+ GoNuts()
+ end),
+ BDMon = spawn_link(fun SysMon () ->
+ receive
+ {monitor, Pid, busy_dist_port, _} ->
+ Tester ! {busy_dist_port, Pid}
+ end,
+ SysMon()
+ end),
+ _ = erlang:system_monitor(BDMon, [busy_dist_port]),
+ DistBufFiller ! go_nuts,
+
+ %% Busy dist entry may release after it has triggered even
+ %% though noone is consuming anything at the receiving end.
+ %% Continue banging until we stop getting new busy_dist_port...
+ WaitFilled = fun WF (Tmo) ->
+ receive
+ {busy_dist_port, DistBufFiller} ->
+ WF(1000*Scale)
+ after
+ Tmo ->
+ ok
+ end
+ end,
+ WaitFilled(infinity),
+
+ BusyDistChecker = spawn_link(fun () ->
+ process_flag(async_dist, false),
+ DeadReceiver ! hello,
+ exit(unexpected_return_from_bang)
+ end),
+ receive {busy_dist_port, BusyDistChecker} -> ok end,
+ {async_dist, false} = process_info(self(), async_dist),
+ {async_dist, false} = process_info(BusyDistChecker, async_dist),
+ false = process_flag(async_dist, true),
+ {async_dist, true} = process_info(self(), async_dist),
+
+ Start = erlang:monotonic_time(millisecond),
+ M1 = erlang:monitor(process, AliveReceiver1),
+ true = is_reference(M1),
+ M2 = erlang:monitor(process, AliveReceiver2),
+ true = is_reference(M2),
+ {pid, Data} = AliveReceiver1 ! {pid, Data},
+ {reg_name, Data} = {alive_receiver_1, Node} ! {reg_name, Data},
+ true = erlang:demonitor(M1),
+ true = link(AliveReceiver2),
+ true = unlink(AliveReceiver1),
+ RId = spawn_request(Node, fun () -> receive bye -> ok end end, [link, monitor]),
+ true = is_reference(RId),
+ erlang:group_leader(self(), AliveReceiver2),
+ AR3XReason = make_ref(),
+ true = exit(AliveReceiver3, AR3XReason),
+ End = erlang:monotonic_time(millisecond),
+
+ %% These signals should have been buffered immediately. Make sure
+ %% it did not take a long time...
+ true = 500*Scale >= End - Start,
+
+ receive after 500*Scale -> ok end,
+
+ unlink(BusyDistChecker),
+ exit(BusyDistChecker, bang),
+ false = is_process_alive(BusyDistChecker),
+
+ unlink(DistBufFiller),
+ exit(DistBufFiller, bang),
+ false = is_process_alive(DistBufFiller),
+
+ %% Verify that the signals eventually get trough when the other
+ %% node continue to work...
+ {links, []}
+ = erpc:call(Node, erlang, process_info, [AliveReceiver1, links]),
+ {links, [Tester]}
+ = erpc:call(Node, erlang, process_info, [AliveReceiver2, links]),
+ {monitored_by, []}
+ = erpc:call(Node, erlang, process_info, [AliveReceiver1, monitored_by]),
+ {monitored_by, [Tester]}
+ = erpc:call(Node, erlang, process_info, [AliveReceiver2, monitored_by]),
+ {messages, [{pid, Data}, {reg_name, Data}]}
+ = erpc:call(Node, erlang, process_info, [AliveReceiver1, messages]),
+ {group_leader, Tester}
+ = erpc:call(Node, erlang, process_info, [AliveReceiver2, group_leader]),
+
+ Spawned = receive
+ {spawn_reply, RId, SpawnRes, Pid} ->
+ ok = SpawnRes,
+ true = is_pid(Pid),
+ Pid
+ end,
+ {links, [Tester]}
+ = erpc:call(Node, erlang, process_info, [Spawned, links]),
+ {monitored_by, [Tester]}
+ = erpc:call(Node, erlang, process_info, [Spawned, monitored_by]),
+
+ receive
+ {'DOWN', AR3Mon, process, AliveReceiver3, ActualAR3XReason} ->
+ AR3XReason = ActualAR3XReason
+ end,
+
+ unlink(AliveReceiver2),
+ unlink(Spawned),
+
+ true = process_flag(async_dist, false),
+ {async_dist, false} = process_info(self(), async_dist),
+
+ ok.
 
 %%% Utilities
 
diff --git a/erts/etc/common/erlexec.c b/erts/etc/common/erlexec.c
index 3e44ad2b77..fa951ae770 100644
--- a/erts/etc/common/erlexec.c
+++ b/erts/etc/common/erlexec.c
@@ -1000,8 +1000,10 @@ int main(int argc, char **argv)
 		 }
 		 break;
 		 case 'p':
-		 if (argv[i][2] != 'c' || argv[i][3] != '\0')
+		 if (!(argv[i][2] == 'c' && argv[i][3] == '\0')
+ && !(argv[i][2] == 'a' && argv[i][3] == 'd' && argv[i][4] == '\0')) {
 			 goto the_default;
+ }
 NEXT_ARG_CHECK();
 		 argv[i][0] = '-';
 		 add_Eargs(argv[i]);
diff --git a/erts/preloaded/src/erlang.erl b/erts/preloaded/src/erlang.erl
index 470f861e08..41471e51e2 100644
--- a/erts/preloaded/src/erlang.erl
+++ b/erts/preloaded/src/erlang.erl
@@ -2425,7 +2425,10 @@ open_port(PortName, PortSettings) ->
 -type message_queue_data() ::
 	off_heap | on_heap.
 
--spec process_flag(trap_exit, Boolean) -> OldBoolean when
+-spec process_flag(async_dist, Boolean) -> OldBoolean when
+ Boolean :: boolean(),
+ OldBoolean :: boolean();
+ (trap_exit, Boolean) -> OldBoolean when
 Boolean :: boolean(),
 OldBoolean :: boolean();
 (error_handler, Module) -> OldModule when
@@ -2463,6 +2466,7 @@ process_flag(_Flag, _Value) ->
 erlang:nif_error(undefined).
 
 -type process_info_item() ::
+ async_dist |
 backtrace |
 binary |
 catchlevel |
@@ -2499,6 +2503,7 @@ process_flag(_Flag, _Value) ->
 trap_exit.
 
 -type process_info_result_item() ::
+ {async_dist, Enabled :: boolean()} |
 {backtrace, Bin :: binary()} |
 {binary, BinInfo :: [{non_neg_integer(),
 non_neg_integer(),
@@ -2949,6 +2954,7 @@ tuple_to_list(_Tuple) ->
 (update_cpu_info) -> changed | unchanged;
 (version) -> string();
 (wordsize | {wordsize, internal} | {wordsize, external}) -> 4 | 8;
+ (async_dist) -> boolean();
 (overview) -> boolean();
 %% Deliberately left undocumented
 (sequential_tracer) -> {sequential_tracer, pid() | port() | {module(),term()} | false}.
@@ -3078,7 +3084,8 @@ spawn_monitor(M, F, A) ->
 | {min_heap_size, Size :: non_neg_integer()}
 | {min_bin_vheap_size, VSize :: non_neg_integer()}
 | {max_heap_size, Size :: max_heap_size()}
- | {message_queue_data, MQD :: message_queue_data()}.
+ | {message_queue_data, MQD :: message_queue_data()}
+ | {async_dist, Enabled :: boolean()}.
 
 -spec spawn_opt(Fun, Options) -> pid() | {pid(), reference()} when
 Fun :: function(),
-- 
2.35.3

Places

File 6401-erts-Support-for-truly-asynchronous-distributed-sign.patch of Package erlang (Revision 33fb1bfcfb3bce757d0f7cfc6a9a3eaf)

Places