On Fri, Dec 4, 2020, at 12:50 AM Hal Murray via devel <devel@ntpsec.org> wrote:
> > That should lead to better (earlier) time stamps. It will bypass the > interrupt response time and the coalesce delays. > > As I understand it, the catch is that the clock out near the network is > not > directly connected to the CPU clock so the time stamps from the hardware > need > to be translated to system time. > > I'm looking for a HOWTO type recipe for setting things up. > I found a question [1] that is tangential and led me to a piece of documentation [2] and a PTP(ish) code sample [3] that looks to be closer. It also looks like it might be used in the ntpd/ntp_packetstamp.c code. I didn't look very hard though. [1] https://networkengineering.stackexchange.com/questions/23369/how-to-check-if-a-nic-supports-hardware-timestamps [2] https://www.kernel.org/doc/Documentation/networking/timestamping.txt [3] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/testing/selftests/net/timestamping.c
_______________________________________________ devel mailing list devel@ntpsec.org http://lists.ntpsec.org/mailman/listinfo/devel
