>Number:         26108
>Category:       kern
>Synopsis:       losing ticks with MULTIPROCESSOR
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Jun 29 10:39:00 UTC 2004
>Closed-Date:
>Last-Modified:
>Originator:     YAMAMOTO Takashi <yamt@mwd.biglobe.ne.jp>
>Release:        NetBSD 2.0F
>Organization:

>Environment:
	
	
System: NetBSD kaeru 2.0F NetBSD 2.0F (build.kaeru) #1461: Tue Jun 29 16:56:46 JST 2004 takashi@kaeru:/home/takashi/work/kernel/build.kaeru i386
Architecture: i386
Machine: i386
>Description:
	as mi_switch() attempt to reacquire kernel_lock with higher IPL
	than IPL_CLOCK, ticks are frequently lost.
>How-To-Repeat:
	run on a multiprocessor box without ntpd.
>Fix:
	the following is a workaround.

Index: kern_synch.c
===================================================================
--- kern_synch.c	(revision 717)
+++ kern_synch.c	(working copy)
@@ -938,6 +938,7 @@ mi_switch(struct lwp *l, struct lwp *new
 		cpu_switchto(l, newl);
 		retval = 0;
 	}
+	spl0(); /* XXX */
 
 	/*
 	 * If we are using h/w performance counters, restore context.
@@ -968,6 +969,7 @@ mi_switch(struct lwp *l, struct lwp *new
 	 * we reacquire the interlock.
 	 */
 	KERNEL_LOCK_ACQUIRE_COUNT(hold_count);
+	(void)splsched(); /* XXX */
 
 	return retval;
 }
>Release-Note:
>Audit-Trail:
>Unformatted: