So I had assumed this was picked up already, but apparently it somehow
got lost in the early 2.6.26 time-frame.
Anyway, I pinged Roman about it, but he hasn't responded, so I wanted to
send this on for inclusion into -mm (I've not had much time to test with
it, so it probably could use some wider testing) and then hopefully
2.6.27.
thanks
-john
From: Roman Zippel <[email protected]>
Due to a rounding problem during a clock update it's possible for
readers to observe the clock jumping back by 1nsec. The following
simplified example demonstrates the problem:
cycle xtime
0 0
1000 999999.6
2000 1999999.2
3000 2999998.8
...
1500 = 1499999.4
= 0.0 + 1499999.4
= 999999.6 + 499999.8
When reading the clock only the full nanosecond part is used, while
timekeeping internally keeps nanosecond fractions. If the clock is now
updated at cycle 1500 here, a nanosecond is missing due to the
truncation.
The simple fix is to round up the xtime value during the update, this
also changes the distance to the reference time, but the adjustment will
automatically take care that it stays under control.
Signed-off-by: Roman Zippel <[email protected]>
Signed-off-by: John Stultz <[email protected]>
---
kernel/time/timekeeping.c | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
Index: linux-2.6-mm/kernel/time/timekeeping.c
===================================================================
--- linux-2.6-mm.orig/kernel/time/timekeeping.c
+++ linux-2.6-mm/kernel/time/timekeeping.c
@@ -484,7 +484,7 @@ void update_wall_time(void)
#else
offset = clock->cycle_interval;
#endif
- clock->xtime_nsec += (s64)xtime.tv_nsec << clock->shift;
+ clock->xtime_nsec = (s64)xtime.tv_nsec << clock->shift;
/* normally this loop will run just once, however in the
* case of lost or late ticks, it will accumulate correctly.
@@ -515,9 +515,12 @@ void update_wall_time(void)
/* correct the clock when NTP error is too big */
clocksource_adjust(offset);
- /* store full nanoseconds into xtime */
- xtime.tv_nsec = (s64)clock->xtime_nsec >> clock->shift;
+ /* store full nanoseconds into xtime after rounding it up and
+ * add the remainder to the error difference.
+ */
+ xtime.tv_nsec = ((s64)clock->xtime_nsec >> clock->shift) + 1;
clock->xtime_nsec -= (s64)xtime.tv_nsec << clock->shift;
+ clock->error += clock->xtime_nsec << (NTP_SCALE_SHIFT - clock->shift);
update_xtime_cache(cyc2ns(clock, offset));
--
Hi,
On Tue, 15 Jul 2008, john stultz wrote:
> So I had assumed this was picked up already, but apparently it somehow
> got lost in the early 2.6.26 time-frame.
>
> Anyway, I pinged Roman about it, but he hasn't responded, so I wanted to
Sorry about that, I wanted to update and test the patches over the
weekend, but other things kept me busy. I'll sent the other patches out
this week.
bye, Roman