Skip site navigation (1)Skip section navigation (2) 
Date:      Thu, 7 Jan 2010 00:47:50 +0000 (UTC)
From:      Attilio Rao <attilio@FreeBSD.org>
To:        src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org
Subject:   svn commit: r201703 - head/sys/kern
Message-ID:  <201001070047.o070looW091014@svn.freebsd.org>
next in thread | raw e-mail | index | archive | help 
Author: attilio
Date: Thu Jan  7 00:47:50 2010
New Revision: 201703
URL: http://svn.freebsd.org/changeset/base/201703

Log:
  Exclusive waiters sleeping with LK_SLEEPFAIL on and using interruptible
  sleeps/timeout may have left spourious lk_exslpfail counts on, so clean
  it up even when accessing a shared queue acquisition, giving to
  lk_exslpfail the value of 'upper limit'.
  In the worst case scenario, infact (mixed
  interruptible sleep / LK_SLEEPFAIL waiters) what may happen is that both
  queues are awaken even if that's not necessary, but still no harm.
  
  Reported by:	Lucius Windschuh <lwindschuh at googlemail dot com>
  Reviewed by:	kib
  Tested by:	pho, Lucius Windschuh <lwindschuh at googlemail dot com>

Modified:
  head/sys/kern/kern_lock.c

Modified: head/sys/kern/kern_lock.c
==============================================================================
--- head/sys/kern/kern_lock.c	Thu Jan  7 00:44:54 2010	(r201702)
+++ head/sys/kern/kern_lock.c	Thu Jan  7 00:47:50 2010	(r201703)
@@ -300,7 +300,14 @@ wakeupshlk(struct lock *lk, const char *
 			}
 				
 		} else {
-			MPASS(lk->lk_exslpfail == 0);
+
+			/*
+			 * Exclusive waiters sleeping with LK_SLEEPFAIL on
+			 * and using interruptible sleeps/timeout may have
+			 * left spourious lk_exslpfail counts on, so clean
+			 * it up anyway.
+			 */
+			lk->lk_exslpfail = 0;
 			queue = SQ_SHARED_QUEUE;
 		}
 
@@ -959,7 +966,14 @@ __lockmgr_args(struct lock *lk, u_int fl
 					queue = SQ_SHARED_QUEUE;
 				}
 			} else {
-				MPASS(lk->lk_exslpfail == 0);
+
+				/*
+				 * Exclusive waiters sleeping with LK_SLEEPFAIL
+				 * on and using interruptible sleeps/timeout
+				 * may have left spourious lk_exslpfail counts
+				 * on, so clean it up anyway. 
+				 */
+				lk->lk_exslpfail = 0;
 				queue = SQ_SHARED_QUEUE;
 			}
 
@@ -1037,8 +1051,16 @@ __lockmgr_args(struct lock *lk, u_int fl
 					queue = SQ_EXCLUSIVE_QUEUE;
 					v &= ~LK_EXCLUSIVE_WAITERS;
 				} else {
+
+					/*
+					 * Exclusive waiters sleeping with
+					 * LK_SLEEPFAIL on and using
+					 * interruptible sleeps/timeout may
+					 * have left spourious lk_exslpfail
+					 * counts on, so clean it up anyway.
+					 */
 					MPASS(v & LK_SHARED_WAITERS);
-					MPASS(lk->lk_exslpfail == 0);
+					lk->lk_exslpfail = 0;
 					queue = SQ_SHARED_QUEUE;
 					v &= ~LK_SHARED_WAITERS;
 				}

Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201001070047.o070looW091014>