Re: HPET timer broken using 2.6.23.13 / nanosleep() hangs

[Thomas Gleixner <tglx@linutronix.de>]

On Wed, 16 Jan 2008, Thomas Gleixner wrote:

> On Wed, 16 Jan 2008, Thomas Gleixner wrote:
> 
> > On Sun, 13 Jan 2008, Dhaval Giani wrote:
> > > On Sun, Jan 13, 2008 at 08:10:46AM -0500, Andrew Paprocki wrote:
> > > > I applied the patch to my 2.6.23.13 tree and upon reboot it stopped right after:
> > > > 
> > > > Clocksource tsc unstable (delta = ... ns)
> > > > Time: hpet clocksource has been installed.
> > > > 
> > > > It locked up hard.. cursor stopped blinking and SysRq isn't working either.
> > 
> > The hint is in the original report:
> > 
> > Clocksource tsc unstable (delta = 500013666 ns)
> > 
> > That means, that the TSC was counting, but HPET not. So it looks like
> > the HPET readout is stale. Which explains, why after switching to HPET
> > clocksource the time system gets hosed.
> > 
> > I wonder why the HPET counter works check did not catch this.
> > 
> > Can you please apply the path below and provide the output ?

This time against 2.6.23

Thanks,
	tglx


--- a/arch/i386/kernel/hpet.c
+++ b/arch/i386/kernel/hpet.c
@@ -233,7 +233,7 @@ int __init hpet_enable(void)
 	unsigned long id;
 	uint64_t hpet_freq;
 	u64 tmp, start, now;
-	cycle_t t1;
+	cycle_t t1, t2;
 
 	if (!is_hpet_capable())
 		return 0;
@@ -295,12 +295,17 @@ int __init hpet_enable(void)
 		rdtscll(now);
 	} while ((now - start) < 200000UL);
 
-	if (t1 == read_hpet()) {
+	t2 = read_hpet();
+	if (t1 == t2) {
 		printk(KERN_WARNING
 		       "HPET counter not counting. HPET disabled\n");
 		goto out_nohpet;
 	}
 
+	printk(KERN_INFO "HPET check: t1=%llu t2=%llu s=%llu n=%llu\n",
+	       (unsigned long long) t1, (unsigned long long) t2,
+	       (unsigned long long) start, (unsigned long long) now);
+
 	/* Initialize and register HPET clocksource
 	 *
 	 * hpet period is in femto seconds per cycle