replay: fix replay of the interrupts
authorPavel Dovgalyuk <pavel.dovgalyuk@ispras.ru>
Mon, 1 Feb 2021 07:05:27 +0000 (10:05 +0300)
committerPaolo Bonzini <pbonzini@redhat.com>
Mon, 8 Feb 2021 13:43:55 +0000 (14:43 +0100)
Sometimes interrupt event comes at the same time with
the virtual timers. In this case replay tries to proceed
the timers, because deadline for them is zero.
This patch allows processing interrupts and exceptions
by entering the vCPU execution loop, when deadline is zero,
but checkpoint associated with virtual timers is not ready
to be replayed.

Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Message-Id: <161216312794.2030770.1709657858900983160.stgit@pasha-ThinkPad-X280>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
accel/tcg/tcg-accel-ops-icount.c

index 87762b469cbfb3eebd234d9476c5d2fe69e3f5fe..13b8fbeb699df0810914e777f3b1c8e2555cc8fa 100644 (file)
@@ -81,7 +81,13 @@ void icount_handle_deadline(void)
     int64_t deadline = qemu_clock_deadline_ns_all(QEMU_CLOCK_VIRTUAL,
                                                   QEMU_TIMER_ATTR_ALL);
 
-    if (deadline == 0) {
+    /*
+     * Instructions, interrupts, and exceptions are processed in cpu-exec.
+     * Don't interrupt cpu thread, when these events are waiting
+     * (i.e., there is no checkpoint)
+     */
+    if (deadline == 0
+        && (replay_mode != REPLAY_MODE_PLAY || replay_has_checkpoint())) {
         icount_notify_aio_contexts();
     }
 }