Module Name:    src
Committed By:   bouyer
Date:           Sun Sep 13 14:38:20 UTC 2009

Modified Files:
        src/sys/kern: kern_turnstile.c

Log Message:
PR kern/41923: assertion "cur != owner" failed
In the for(;;) loop of turnstile_block(), the lock owner can change while
cur's lock is released (cur's lock is also the tschain_t's mutex).
Remove the KASSERT about owner being invariant and try to deal with the
fact that the owner can change instead.
http://mail-index.netbsd.org/tech-kern/2009/08/24/msg005957.html
and followups.


To generate a diff of this commit:
cvs rdiff -u -r1.24 -r1.25 src/sys/kern/kern_turnstile.c

Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.

Modified files:

Index: src/sys/kern/kern_turnstile.c
diff -u src/sys/kern/kern_turnstile.c:1.24 src/sys/kern/kern_turnstile.c:1.25
--- src/sys/kern/kern_turnstile.c:1.24	Sat Mar 21 13:11:14 2009
+++ src/sys/kern/kern_turnstile.c	Sun Sep 13 14:38:20 2009
@@ -1,4 +1,4 @@
-/*	$NetBSD: kern_turnstile.c,v 1.24 2009/03/21 13:11:14 ad Exp $	*/
+/*	$NetBSD: kern_turnstile.c,v 1.25 2009/09/13 14:38:20 bouyer Exp $	*/
 
 /*-
  * Copyright (c) 2002, 2006, 2007, 2009 The NetBSD Foundation, Inc.
@@ -60,7 +60,7 @@
  */
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: kern_turnstile.c,v 1.24 2009/03/21 13:11:14 ad Exp $");
+__KERNEL_RCSID(0, "$NetBSD: kern_turnstile.c,v 1.25 2009/09/13 14:38:20 bouyer Exp $");
 
 #include <sys/param.h>
 #include <sys/lockdebug.h>
@@ -253,6 +253,7 @@
 	sq = &ts->ts_sleepq[q];
 	ts->ts_waiters[q]++;
 	sleepq_enter(sq, l, tc->tc_mutex);
+	/* now tc->tc_mutex is also cur->l_mutex and l->l_mutex */
 	LOCKDEBUG_BARRIER(tc->tc_mutex, 1);
 	l->l_kpriority = true;
 	obase = l->l_kpribase;
@@ -275,6 +276,7 @@
 	 * compiling a kernel with LOCKDEBUG to pinpoint the problem.
 	 */
 	prio = lwp_eprio(l);
+
 	for (;;) {
 		bool dolock;
 
@@ -285,9 +287,24 @@
 		if (owner == NULL)
 			break;
 
-		KASSERT(l != owner);
-		KASSERT(cur != owner);
+		/* The owner may have changed as we have dropped the tc lock */
+		if (cur == owner) {
+			/*
+			 * we own the lock: stop here, sleepq_block()
+			 * should wake up immediatly
+			 */
+			break;
+		}
 
+		if (l == owner) {
+			/* owner has changed, restart from curlwp */
+			lwp_unlock(l);
+			l = cur;
+			lwp_lock(l);
+			prio = lwp_eprio(l);
+			continue;
+		}
+			
 		if (l->l_mutex != owner->l_mutex)
 			dolock = true;
 		else
@@ -295,6 +312,10 @@
 		if (dolock && !lwp_trylock(owner)) {
 			/*
 			 * restart from curlwp.
+			 * Note that there may be a livelock here:
+			 * the owner may try grabing cur's lock (which is
+			 * the tc lock) while we're trying to grab
+			 * the owner's lock.
 			 */
 			lwp_unlock(l);
 			l = cur;

Reply via email to