@Guy: Once again thanks for your detailed answer. I'll try to modify pcap then.
@James and Rixed:
I just ran a test on my device. The test takes 150 timestamps and calculates
the difference between of every timestamp and the timestamp taken after it.
My device is a little older. Its cpu only runs at 800MHz and is a single core
cpu. That output looks pretty good to me, even though there is a zero diff here
and there.
I also ran the test in an infinity loop while putting the cpu under a stress
test. The results are the same. The values are either 30517,30518 or 0ns.
This is my loop code (res is a timespec struct):
for(i = 0; i < 150; i++){
clock_gettime(CLOCK_REALTIME, &res);
secDiff = res.tv_sec - oldSec;
nsecDiff = 1000000000 * secDiff + res.tv_nsec - oldNSec;
oldSec = res.tv_sec;
oldNSec = res.tv_nsec;
printf("Diff: %ldns\n", nsecDiff);
}
This is the output:
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 0ns
Diff: 30517ns
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 0ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 0ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
Diff: 30517ns
Diff: 30518ns
----- Ursprüngliche Mail -----
Von: ri...@happyleptic.org
An: tcpdump-workers@lists.tcpdump.org
CC: "Maik Jäkel" <em...@maikjaekel.de>
Gesendet: Samstag, 29. Dezember 2012 10:15:51
Betreff: Re: [tcpdump-workers] Fwd: Re: Printing nanosecond timestamp
information in raw output
> I didn't know a better way, though and hoped that the execution time between
> the reception of the packet and taking the timestamp is relatively constant
> (with an uncertainty of 4ns or so).
On a 1.2GHz cpu clock, as is typical of android devices these days, that
would be 5 clock ticks, ie 5 ARM instructions at best. I'd be very
surprised if the uncertainty for such a long periple from reception to
timestamping were that low. Please keep the list informed when you have
measured the actual variation between timestamps of outgoing packets.
_______________________________________________
tcpdump-workers mailing list
tcpdump-workers@lists.tcpdump.org
https://lists.sandelman.ca/mailman/listinfo/tcpdump-workers
