On Jul 9, 2010, at 12:00 PM, Jan Beulich wrote:
>>>> On 09.07.10 at 16:57, Jed Smith <jed@xxxxxxxxxx> wrote:
>> On Jul 9, 2010, at 4:46 AM, Jan Beulich wrote:
>>> Is this perhaps also dependent on the CPU make/model?
>> - Between 2.6.29 and 2.6.30, /proc/uptime behaves much differently, and the
>> bug
>> then exposes itself. Something changed there.
>
> Others as well as me had seen similar misbehavior (outside of pv-ops),
> but only on reasonably new Intel systems. sched_clock_stable getting
> set to one in arch/x86/kernel/cpu/intel.c turned out to be the problem.
Like so? This change seems to fix the issue, but I'm going to continue to
hammer on it for the remainder of the day.
diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c
index 40e1835..e362517 100644
--- a/arch/x86/kernel/cpu/intel.c
+++ b/arch/x86/kernel/cpu/intel.c
@@ -71,7 +71,14 @@ static void __cpuinit early_init_intel(struct cpuinfo_x86 *c)
set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC);
set_cpu_cap(c, X86_FEATURE_TSC_RELIABLE);
+#ifndef CONFIG_XEN
+ /*
+ * Considering the clock stable in paravirt domUs under Xen
+ * causes timing instability on certain Intel CPUs.
+ */
sched_clock_stable = 1;
+#endif /* CONFIG_XEN */
}
/*
Regards,
Jed Smith
Systems Administrator
Linode, LLC
+1 (609) 593-7103 x1209
jed@xxxxxxxxxx
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|
Home