🚀 go-pugleaf

rstevens@noao.edu (W. Richard Stevens) writes:

>> Just wondering if anyone can give any additional information on
>> just what the "dosynctodr" kernel flag does in SunOS, and why
>> it should be set to 0.

>OK, I can answer some of my own question after finding the code in
>ntp_util.c that handles this case.  BUT, the code says that setting
>synctodr to 0 is not what you want (as a reboot will start off with
>a wacky time),

No, that's not really what I meant.

Setting dosynctodr to 0 is what you want to do. The wackiness after
reboot is a consequence of setting dosynctodr to 0, but it is far
more desirable to do it than not.

>but the tickadj(8) man pages says you should set it
>to 0 for SunOS 4.0.  Also, the distributed Config.sun file does not
>set DOSYNCTODR_SUCKS, which goes against the source code comments.

I disagree slightly with the comment in the Config file about
DOSYNCTODR_SUCKS. I recommend doing it on all Suns. I have heard
arguments that it causes a 800 us jitter when it happens. This
too is true, but I know of no other documented way to set the battery
clock than to call settimeofday() (though I have heard that
doresettodr is a possibility, I haven't been able to confirm that
it works as advertised).

>I'm confused as to how I should set it, both in the Config file and
>using tickadj.

Use both tickadj and DOSYNCTODR_SUCKS.

>Lastly, the man page doesn't say so, but I assume that Sun's settimeofday(2)
>system call is what syncs the battery clock to its argument?

Yes. That's the whole point behind DOSYNCTODR_SUCKS.

-- 
Nick Sayer <mrapple@quack.sac.ca.us> | "You see it. You hear it.
N6QQQ @ N0ARY.#NOCAL.CA.USA.NOAM     |  You want to turn away.
+1 408 249 9630, log in as 'guest'   |  BUT YOU CAN'T!!!"
PGP 2.1 public key on request        | 	 -- Ren & Stimpy add