[ntp:questions] Re: sntp: totally spurious NTP packet rejected on socket 0
David L. Mills
mills at udel.edu
Thu Oct 28 21:13:01 UTC 2004
As you will note from the sntp.html page, there is strong advice NOT to
use SNTP as a server unless the configuration includes a reference
clock. There are many reasons for this; the most important is that,
since the system clock is not disciplined in frequency, the client will
find rather serious jitter unless the client polls at an unreasonable
rate. The recent experience with NIST, USNO and UDel flooding attacks
may be due to seriously broken implementations that do try to minimize
jitter by polling at one-second intervals continuously and forever. I've
even found rascals doing that at three times per second.
I have looked at the SNTP code in the current NTPv4 distribution and am
not at all happy with the programming or protocol model. I advised our
release engineering staff to NOT include that program in the
distribution, but if people really want it, make it a separate
distribution and remove the dangerous "feature" of running as a server.
You may need to contact the author of that program.
Wagener, Brian H wrote:
> I am trying to install a SNTP server and client to run on a embedded QNX system to act as a simple time server for other embedded computers. I built the sntp package as part of ntp-4.2.0, but when a standard ntpd 4.2.0 client connects to it, I get a receive error. Below is the output.
> Packet of length 48 received from c009c968/007b
> Incoming packet on socket 0:
> e30006f0 00000000 00000041 00000000 00000000 00000000 00000000 00000000
> 00000000 00000000 c52b9860 791d8a54
> sta=3 ver=4 mod=3 str=0 pol=6 dis=0.000992 ref=0.000000
> ori=0.000000 rec=0.000000
> tra=3307968608.473107 cur=3307968606.360500
> sntp: totally spurious NTP packet rejected on socket 0
> Read Packet
> sntp: after 1 mins 48 secs 0 acc. 6 rej. 0 b'cast
> Brian Wagener
More information about the questions