we have found a bug in the ntp-4.2.4p0-sleep.patch.
The problem is that when allowing ntp to use stepping,
and not only slewing, ntp will crash using this patch.
* The "slew only"-option, or option -x, is not recommended
by the ntp-manual in the general case.
* The bug is triggered when disabling the -x option and
starting ntpd with a large offset to the timeserver (-x
is disabled by default, so this is easily tested if
one bothers with doing basic testing...)
In detail the problem arises because timer_gettime() and
timer_settime() is still called in reinit_timer() even
though the call to timer_create() in init_timer() has
been effectively removed. The "get_systime(); return;"
in reinit_timer() is never reached because the #ifdef-statements
before it are in fact true (else statement never reached.)
Another problem is that this patch has, obviously, not
been properly tested, and that the patch is for
an acknowledged complex issue. Just look at the
Based on the these facts we will probably not be
using this patch, even if the specific issue raised
here is fixed. At least not until it has been
properly scrutinized and reviewed by the upstream
ntp-folk, and (I want to repeat this) properly
Have a nice day, and best regards,
Redhat-devel-list mailing list