ntp: Do leapsecond adjustment in adjtimex read path
authorJohn Stultz <john.stultz@linaro.org>
Thu, 11 Jun 2015 22:54:56 +0000 (15:54 -0700)
committerThomas Gleixner <tglx@linutronix.de>
Fri, 12 Jun 2015 09:15:49 +0000 (11:15 +0200)
Since the leapsecond is applied at tick-time, this means there is a
small window of time at the start of a leap-second where we cross into
the next second before applying the leap.

This patch modified adjtimex so that the leap-second is applied on the
second edge. Providing more correct leapsecond behavior.

This does make it so that adjtimex()'s returned time values can be
inconsistent with time values read from gettimeofday() or
clock_gettime(CLOCK_REALTIME,...)  for a brief period of one tick at
the leapsecond.  However, those other interfaces do not provide the
TIME_OOP time_state return that adjtimex() provides, which allows the
leapsecond to be properly represented. They instead only see a time
discontinuity, and cannot tell the first 23:59:59 from the repeated
23:59:59 leap second.

This seems like a reasonable tradeoff given clock_gettime() /
gettimeofday() cannot properly represent a leapsecond, and users
likely care more about performance, while folks who are using
adjtimex() more likely care about leap-second correctness.

Signed-off-by: John Stultz <john.stultz@linaro.org>
Cc: Prarit Bhargava <prarit@redhat.com>
Cc: Daniel Bristot de Oliveira <bristot@redhat.com>
Cc: Richard Cochran <richardcochran@gmail.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jiri Bohac <jbohac@suse.cz>
Cc: Ingo Molnar <mingo@kernel.org>
Link: http://lkml.kernel.org/r/1434063297-28657-5-git-send-email-john.stultz@linaro.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
kernel/time/ntp.c

index 033743e3647a6d626229451eb76aee1aa22edeec..fb4d98c7fd43e715d795d8df2c522c80b2204361 100644 (file)
@@ -740,6 +740,24 @@ int __do_adjtimex(struct timex *txc, struct timespec64 *ts, s32 *time_tai)
        if (!(time_status & STA_NANO))
                txc->time.tv_usec /= NSEC_PER_USEC;
 
+       /* Handle leapsec adjustments */
+       if (unlikely(ts->tv_sec >= ntp_next_leap_sec)) {
+               if ((time_state == TIME_INS) && (time_status & STA_INS)) {
+                       result = TIME_OOP;
+                       txc->tai++;
+                       txc->time.tv_sec--;
+               }
+               if ((time_state == TIME_DEL) && (time_status & STA_DEL)) {
+                       result = TIME_WAIT;
+                       txc->tai--;
+                       txc->time.tv_sec++;
+               }
+               if ((time_state == TIME_OOP) &&
+                                       (ts->tv_sec == ntp_next_leap_sec)) {
+                       result = TIME_WAIT;
+               }
+       }
+
        return result;
 }