[patch 12/19] SPARC64: Fix serious bug in sched_clock() on sparc64

[Greg KH <gregkh@suse.de>]

[-- Attachment #1: sparc64-fix-serious-bug-in-sched_clock-on-sparc64.patch --]
[-- Type: text/plain, Size: 1160 bytes --]

-stable review patch.  If anyone has any objections, please let us know.

------------------
From: David S. Miller <davem@davemloft.net>

[SPARC64]: Fix sched_clock() wrapping every ~17 seconds.

Unfortunately, sparc64 doesn't have an easy way to do a "64 X 64 -->
128" bit multiply like PowerPC and IA64 do.  We were doing a
"64 X 64 --> 64" bit multiple which causes overflow very quickly with
a 30-bit quotient shift.

So use a quotientshift count of 10 instead of 30, just like x86 and
ARM do.

This also fixes the wrapping of printk timestamp values every ~17
seconds.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

---
 arch/sparc64/kernel/time.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

--- linux-2.6.17.13.orig/arch/sparc64/kernel/time.c
+++ linux-2.6.17.13/arch/sparc64/kernel/time.c
@@ -1105,7 +1105,7 @@ static struct time_interpolator sparc64_
 };
 
 /* The quotient formula is taken from the IA64 port. */
-#define SPARC64_NSEC_PER_CYC_SHIFT	30UL
+#define SPARC64_NSEC_PER_CYC_SHIFT	10UL
 void __init time_init(void)
 {
 	unsigned long clock = sparc64_init_timers();

--