svn commit: r270995 - stable/10/sys/vm

[Alan Cox]

Author: alc
Date: Wed Sep  3 06:47:05 2014
New Revision: 270995
URL: http://svnweb.freebsd.org/changeset/base/270995

Log:
  MFC r270666
    Back in the days when the kernel was single threaded, testing
    "vm_paging_target() > 0" was a reasonable way of determining if the
    inactive queue scan met its target.  However, now that other threads
    can be allocating pages while the inactive queue scan is running, it's
    an unreliable method.  The effect of it being unreliable is that we
    can start swapping out processes when we didn't intend to.
  
    This issue has existed since the kernel was multithreaded, but the
    changes to the inactive queue target in 10.0-RELEASE have made its
    effects visible.
  
    This change introduces a more direct method for determining if the
    inactive queue scan met its target that is not affected by the actions
    of other threads.

Modified:
  stable/10/sys/vm/vm_pageout.c
Directory Properties:
  stable/10/   (props changed)

Modified: stable/10/sys/vm/vm_pageout.c
==============================================================================
--- stable/10/sys/vm/vm_pageout.c	Wed Sep  3 06:34:16 2014	(r270994)
+++ stable/10/sys/vm/vm_pageout.c	Wed Sep  3 06:47:05 2014	(r270995)
@@ -1308,6 +1308,23 @@ relock_queues:
 	}
 	vm_pagequeue_unlock(pq);
 
+#if !defined(NO_SWAPPING)
+	/*
+	 * Wakeup the swapout daemon if we didn't cache or free the targeted
+	 * number of pages. 
+	 */
+	if (vm_swap_enabled && page_shortage > 0)
+		vm_req_vmdaemon(VM_SWAP_NORMAL);
+#endif
+
+	/*
+	 * Wakeup the sync daemon if we skipped a vnode in a writeable object
+	 * and we didn't cache or free enough pages.
+	 */
+	if (vnodes_skipped > 0 && page_shortage > vm_cnt.v_free_target -
+	    vm_cnt.v_free_min)
+		(void)speedup_syncer();
+
 	/*
 	 * Compute the number of pages we want to try to move from the
 	 * active queue to the inactive queue.
@@ -1418,20 +1435,6 @@ relock_queues:
 		}
 	}
 #endif
-		
-	/*
-	 * If we didn't get enough free pages, and we have skipped a vnode
-	 * in a writeable object, wakeup the sync daemon.  And kick swapout
-	 * if we did not get enough free pages.
-	 */
-	if (vm_paging_target() > 0) {
-		if (vnodes_skipped && vm_page_count_min())
-			(void) speedup_syncer();
-#if !defined(NO_SWAPPING)
-		if (vm_swap_enabled && vm_page_count_target())
-			vm_req_vmdaemon(VM_SWAP_NORMAL);
-#endif
-	}
 
 	/*
 	 * If we are critically low on one of RAM or swap and low on