OS-6363 system went to dark side of moon for ~467 seconds OS-6404 ARC reclaim should throttle its calls to arc_kmem_reap_now() Reviewed by: Bryan Cantrill <bryan@joyent.com> Reviewed by: Dan McDonald <danmcd@joyent.com>

@@ -20,10 +20,11 @@
  */
 /*
  * Copyright (c) 1994, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
  * Copyright 2015 Nexenta Systems, Inc.  All rights reserved.
+ * Copyright (c) 2017, Joyent, Inc.
  */
 
 /*
  * Kernel memory allocator, as described in the following two papers and a
  * statement about the consolidator:

@@ -3247,22 +3248,38 @@
         }
 
 }
 
 /*
- * Reap (almost) everything right now.
+ * Allow our caller to determine if there are running reaps.
+ *
+ * This call is very conservative and may return B_TRUE even when
+ * reaping activity isn't active. If it returns B_FALSE, then reaping
+ * activity is definitely inactive.
  */
+boolean_t
+kmem_cache_reap_active(void)
+{
+        return (!taskq_empty(kmem_taskq));
+}
+
+/*
+ * Reap (almost) everything soon.
+ *
+ * Note: this does not wait for the reap-tasks to complete. Caller
+ * should use kmem_cache_reap_active() (above) and/or moderation to
+ * avoid scheduling too many reap-tasks.
+ */
 void
-kmem_cache_reap_now(kmem_cache_t *cp)
+kmem_cache_reap_soon(kmem_cache_t *cp)
 {
         ASSERT(list_link_active(&cp->cache_link));
 
         kmem_depot_ws_zero(cp);
 
         (void) taskq_dispatch(kmem_taskq,
             (task_func_t *)kmem_depot_ws_reap, cp, TQ_SLEEP);
-        taskq_wait(kmem_taskq);
 }
 
 /*
  * Recompute a cache's magazine size.  The trade-off is that larger magazines
  * provide a higher transfer rate with the depot, while smaller magazines