Last update: 10-Mar-2014 05:24 UTC
The tickadj program reads, and optionally modifies, several timekeeping-related variables in older kernels that do not have support for precision timekeeping, including HP-UX, SunOS, Ultrix, SGI and probably others. Those machines provide means to patch the kernel /dev/kmem. Newer machines with kernel time support, including Solaris, Tru64, FreeBSD and Linux, should NOT use the program, even if it appears to work, as it will destabilize the kernel time support. Use the ntptime program instead.
The particular variables that can be changed with tickadj include tick, which is the number of microseconds added to the system time for a clock interrupt, tickadj, which sets the slew rate and resolution used by the adjtime system call, and dosynctodr, which indicates to the kernels on some machines whether they should internally adjust the system clock to keep it in line with time-of-day clock or not.
By default, with no arguments, tickadj reads the variables of interest in the kernel and displays them. At the same time, it determines an "optimal" value for the value of the tickadj variable if the intent is to run the ntpd Network Time Protocol (NTP) daemon, and prints this as well. Since the operation of tickadj when reading the kernel mimics the operation of similar parts of the ntpd program fairly closely, this can be useful when debugging problems with ntpd.
Note that tickadj should be run with some caution when being used for the first time on different types of machines. The operations which tickadj tries to perform are not guaranteed to work on all Unix machines and may in rare cases cause the kernel to crash.
Home Page | Site Map | Contacts |