[ntp:hackers] NTP two-step on-wire protocol

Greg Dowd GDowd at symmetricom.com
Mon Jul 23 02:10:43 UTC 2007


The problem of when the tx int goes off isn't really different that the
issue of the rx.  Nics are quite happy to dump multiple packets in queue
b4 interrupting, so_timestamp is stack so it's triggered by the bottom
half of the driver unless the nic driver itself wants to handle the
timestamp.  I'd be concerned about the use of polled mode in a lot of
the newer highspeed nic drivers (at least in linux).  I'm actually
wondering how those might impact so_timestamp use but there isn't enough
data yet.  There are a few hi frequency interrupt mitigators in nic
designs like a leaky bucket register or on/off transitions that could
cause jitter or wander depending on the network traffic.  But, on most
machines, it should be better than the current tx timestamps.

-----Original Message-----
From: hackers-bounces+gdowd=symmetricom.com at lists.ntp.org
[mailto:hackers-bounces+gdowd=symmetricom.com at lists.ntp.org] On Behalf
Of M. Warner Losh
Sent: Sunday, July 22, 2007 3:08 PM
To: mills at udel.edu
Cc: hackers at ntp.org
Subject: Re: [ntp:hackers] NTP two-step on-wire protocol

In message: <46A3C9EB.7080800 at udel.edu>
            "David L. Mills" <mills at udel.edu> writes:
: Okay, the gauntlet is tossed. Can an output interrupt timestamp be
: implemented in FreeBSD? I am completely incompetent in how best to
: shoehorn such a thing in a politically correct way. Contrary to what I
: said in a previous message, the only new feature required is this
timestamp.

Yes.  I'm sure that it can.  I'd be happy to help you navigate the
political minefield as well, since I know a thing or two about that...
:-)

There are some 'issues' here.  The timestamp will be when the packet was
transmitted.  Depending on the NIC, one can queue up sequences packets
to be sent, and only get an interrupt at the end, or when the transmit
buffers have completely drained.  It varies a lot by NIC chip.

Warner
_______________________________________________
hackers mailing list
hackers at lists.ntp.org
https://lists.ntp.org/mailman/listinfo/hackers


More information about the hackers mailing list