Re: NTP dumps Linux, film at 11. [Fwd/FYI]

Vern Hoxie (
Thu, 3 Dec 1998 15:23:42 -0700

On Wed, 02 Dec 1998, Brandon S. Allbery KF8NH wrote:

> 13 seconds a day is easily within normal PC timer lossage, though
> (my old 486 pretty consistently loses that much time, regardless of
> OS). 50 seconds in 12 hours is more worrisome, especially when it
> doesn't happen in OS/2....

I use a personal program to call the NIST (National Institute of
Standards and Technology) at Boulder, Colo. once each day. I use the
data collected to reset the system clock with 'stime(2)' and a copy of
the kernel routine to directly set the RTC. These both reset the time
to 1 sec granularity.

Over extended periods of time, the system clock gains a nominal 1 sec
per day and the RTC loses 1 sec every 5 or 6 days (no reset if time
error equal 0 sec).

The machine is:
AST 484DX2-66
Linux 2.0.29
libc 5.4.23

I tried to use 'adjtime()' and 'timex()' with only 1 sec granularity
and no recalc of the PLL constant (I couldn't grok all the obscure
references). They screwed things up enormously!

FWIW, the same program runs on my 3b1. It gains a nominal 7 sec per

Both machines run continuously (extended power failures excepted) under
very light loads. The Linux machine does news and mail reading with
an occasional compile or web surf with Netscape.


Vernon C. Hoxie                           
3975 W. 29th Ave.                                        uucp: 303-455-2670
Denver, Colo., 80212                                    voice: 303-477-1780
                 Heroes and winners aren't the same thing.

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to Please read the FAQ at