diff --git a/src/racket/gc2/newgc.c b/src/racket/gc2/newgc.c
index 2b89eb4a7d..7460b66294 100644
--- a/src/racket/gc2/newgc.c
+++ b/src/racket/gc2/newgc.c
@@ -2716,7 +2716,11 @@ void GC_switch_back_from_master(void *gc) {
 void GC_gcollect(void)
 {
   NewGC *gc = GC_get_GC();
-  garbage_collect(gc, 1, 0, NULL);
+
+  if (postmaster_and_master_gc(gc))
+    master_collect_initiate(gc);
+  else
+    garbage_collect(gc, 1, 0, NULL);
 }
 
 static inline int atomic_mark(void *p) { return 0; }
diff --git a/src/racket/src/place.c b/src/racket/src/place.c
index 348d0bc574..118ae2d498 100644
--- a/src/racket/src/place.c
+++ b/src/racket/src/place.c
@@ -1961,7 +1961,7 @@ static void* GC_master_malloc_tagged(size_t size) {
   return ptr;
 }
 
-static void async_channel_finialize(void *p, void* data) {
+static void async_channel_finalize(void *p, void* data) {
   Scheme_Place_Async_Channel *ch;
   int i;
   ch = (Scheme_Place_Async_Channel*)p;
@@ -1984,6 +1984,9 @@ Scheme_Place_Async_Channel *scheme_place_async_channel_create() {
   Scheme_Object **msgs;
   Scheme_Place_Async_Channel *ch;
   void **msg_memory;
+#ifdef MZ_PRECISE_GC
+  void *original_gc;
+#endif
 
   ch = GC_master_malloc_tagged(sizeof(Scheme_Place_Async_Channel));
   msgs = GC_master_malloc(sizeof(Scheme_Object*) * 8);
@@ -1998,6 +2001,15 @@ Scheme_Place_Async_Channel *scheme_place_async_channel_create() {
   ch->msgs = msgs;
   ch->msg_memory = msg_memory;
   ch->wakeup_signal = NULL;
+
+#ifdef MZ_PRECISE_GC
+  original_gc = GC_switch_to_master_gc();
+  GC_set_finalizer(ch, 1, 1, async_channel_finalize, NULL, NULL, NULL);
+  GC_switch_back_from_master(original_gc);
+#endif
+  /* FIXME? Need finalizer for non-precise GC if places become supported 
+     in that mode. */
+
   return ch;
 }