Re: [RFC PATCH v2 0/4] Add support for S3 non-stop TSC support.

[Feng Tang <feng.tang@intel.com>]

On Tue, Mar 05, 2013 at 02:17:59PM +0800, John Stultz wrote:
> On 03/05/2013 12:32 PM, Jason Gunthorpe wrote:
> >On Tue, Mar 05, 2013 at 11:53:02AM +0800, Feng Tang wrote:
> >
> >>>// Drops some small precision along the way but is simple..
> >>>static inline u64 cyclecounter_cyc2ns_128(const struct cyclecounter *cc,
> >>>                                           cycle_t cycles)
> >>>{
> >>>     u64 max = U64_MAX/cc->mult;
> >>>     u64 num = cycles/max;
> >>>     u64 result = num * ((max * cc->mult) >> cc->shift);
> >>>     return result + cyclecounter_cyc2ns(cc, cycles - num*cc->mult);
> >>>}
> >
> Probably want to use clocksource instead of cyclecounter, but I
> think Jason's approach sounds ok. I might suggest that you initially
> make the function static to the timekeeping code, just so we don't
> get unexpected users.

Thought more about it, can we directly make clocksource_cyc2ns() cover
the overflow case? Something like:

diff --git a/include/linux/clocksource.h b/include/linux/clocksource.h
index aa7032c..1ecc872 100644
--- a/include/linux/clocksource.h
+++ b/include/linux/clocksource.h
@@ -274,7 +274,16 @@ static inline u32 clocksource_hz2mult(u32 hz, u32 shift_constant)
  */
 static inline s64 clocksource_cyc2ns(cycle_t cycles, u32 mult, u32 shift)
 {
-	return ((u64) cycles * mult) >> shift;
+	u64 max = ULLONG_MAX / mult;
+	s64 nsec = 0;
+
+	/* The (mult * cycles) may overflow 64 bits, so add a max check */
+	if (cycles > max) {
+		nsec = ((max * mult) >> shift) * (cycles / max);
+		cycles %= max;
+	}
+	nsec += ((u64) cycles * mult) >> shift;
+	return nsec;
 }

Thanks,
Feng