From 938cd607f31559d5497b620b0c5ed8b92972bb27 Mon Sep 17 00:00:00 2001
From: Rickard Green <rickard@erlang.org>
Date: Wed, 20 Mar 2019 17:43:24 +0100
Subject: Fix reception of resume signal on process executing dirty

If a suspend/resume signal pair was sent to a process while it
was executing dirty the resume counter on the process got into
an inconsistent state. This in turn could cause the process
to enter a suspended state indefinitely.
---
 erts/emulator/beam/erl_proc_sig_queue.c | 10 ++++--
 erts/emulator/test/dirty_bif_SUITE.erl  | 54 +++++++++++++++++++++++++++++++--
 2 files changed, 60 insertions(+), 4 deletions(-)

diff --git a/erts/emulator/beam/erl_proc_sig_queue.c b/erts/emulator/beam/erl_proc_sig_queue.c
index 18418a76e1..c95b7057fd 100644
--- a/erts/emulator/beam/erl_proc_sig_queue.c
+++ b/erts/emulator/beam/erl_proc_sig_queue.c
@@ -3247,9 +3247,15 @@ erts_proc_sig_handle_incoming(Process *c_p, erts_aint32_t *statep,
                             erts_nif_demonitored((ErtsResource *) tmon->other.ptr);
                             cnt++;
                             break;
-                        case ERTS_MON_TYPE_SUSPEND:
-                            erts_resume(c_p, ERTS_PROC_LOCK_MAIN);
+                        case ERTS_MON_TYPE_SUSPEND: {
+                            ErtsMonitorSuspend *msp;
+                            erts_aint_t mstate;
+                            msp = (ErtsMonitorSuspend *) erts_monitor_to_data(tmon);
+                            mstate = erts_atomic_read_acqb(&msp->state);
+                            if (mstate & ERTS_MSUSPEND_STATE_FLG_ACTIVE)
+                                erts_resume(c_p, ERTS_PROC_LOCK_MAIN);
                             break;
+                        }
                         default:
                             break;
                         }
diff --git a/erts/emulator/test/dirty_bif_SUITE.erl b/erts/emulator/test/dirty_bif_SUITE.erl
index 46eb0cba58..4f5ad0295a 100644
--- a/erts/emulator/test/dirty_bif_SUITE.erl
+++ b/erts/emulator/test/dirty_bif_SUITE.erl
@@ -38,7 +38,8 @@
 	 dirty_process_info/1,
 	 dirty_process_register/1,
 	 dirty_process_trace/1,
-	 code_purge/1]).
+	 code_purge/1,
+         otp_15688/1]).
 
 suite() -> [{ct_hooks,[ts_install_cth]}].
 
@@ -64,7 +65,8 @@ all() ->
      dirty_process_info,
      dirty_process_register,
      dirty_process_trace,
-     code_purge].
+     code_purge,
+     otp_15688].
 
 init_per_suite(Config) ->
     case erlang:system_info(dirty_cpu_schedulers) of
@@ -498,10 +500,58 @@ code_purge(Config) when is_list(Config) ->
     true = Time =< 1000,
     ok.
 
+otp_15688(Config) when is_list(Config) ->
+    ImBack = make_ref(),
+    {See, SeeMon} = spawn_monitor(fun () ->
+                                          erts_debug:dirty_io(wait, 2000),
+                                          exit(ImBack)
+                                  end),
+    wait_until(fun () ->
+                       [{current_function, {erts_debug, dirty_io, 2}},
+                        {status, running}]
+                           == process_info(See,
+                                           [current_function, status])
+               end),
+    {Ser1, Ser1Mon} = spawn_monitor(fun () ->
+                                            erlang:suspend_process(See,
+                                                                   [asynchronous])
+                                    end),
+    erlang:suspend_process(See, [asynchronous]),
+    receive {'DOWN', Ser1Mon, process, Ser1, normal} -> ok end,
+
+    %% Verify that we sent the suspend request while it was executing dirty...
+    [{current_function, {erts_debug, dirty_io, 2}},
+     {status, running}] = process_info(See, [current_function, status]),
+
+    wait_until(fun () ->
+                       {status, suspended} == process_info(See, status)
+               end),
+    erlang:resume_process(See),
+
+    receive
+        {'DOWN', SeeMon, process, See, Reason} ->
+            ImBack = Reason
+    after 4000 ->
+            %% Resume bug seems to have hit us...
+            PI = process_info(See),
+            exit(See, kill),
+            ct:fail({suspendee_stuck, PI})
+    end.
+    
+
 %%
 %% Internal...
 %%
 
+wait_until(Fun) ->
+    case Fun() of
+        true ->
+            ok;
+        _ ->
+            receive after 100 -> ok end,
+            wait_until(Fun)
+    end.
+
 access_dirty_process(Config, Start, Test, Finish) ->
     {ok, Node} = start_node(Config, ""),
     [ok] = mcall(Node,
-- 
cgit v1.2.3