On Thu, 18 Jun 2026 18:45:19 +0200 Morten Brørup <[email protected]> wrote:
> > From: Stephen Hemminger [mailto:[email protected]] > > Sent: Thursday, 18 June 2026 17.23 > > > > On Thu, 18 Jun 2026 10:38:15 -0400 > > Dawid Wesierski <[email protected]> wrote: > > > > > + * @param ts > > > + * Packet timestamp in nanoseconds since the Unix epoch. If zero, > > the > > > + * current TSC is captured and converted to epoch ns by > > > + * rte_pcapng_write_packets() when the packet is written. > > > * > > > > It might help users if a helper rte_tsc_to_epoch() was exposed. > > +1 > > Please note that such a helper would need to compensate for rte_rdtsc() drift. > > Simplified: > > int64_t rte_tsc_to_ns(tsc) > { > struct timespec ts = clock_gettime(CLOCK_REALTIME); > int64_t now_ns = ts.tv_sec * NS_PER_S + ts.tv_nsec; > int64_t now_tsc = rte_rdtsc(); > int64_t diff_tsc = tsc - tsc_now; > return now_ns + diff_tsc * NS_PER_S / rte_get_tsc_hz(); > } > > A performance optimized version would take "now_ns" and "now_tsc" as > parameters. > And with "now_ns" passed as a parameter, the function also works with other > clocks, such as CLOCK_MONOTONIC. > > Also see this discussion on the Grout mailing list: > https://inbox.dpdk.org/grout/[email protected]/T/#m465a04ca2e8219612dd9c3efb4198d23d5813422 > > Inside rte_pcapng is already all the necessary state and setup. Just not exposed to user. It does it without divide operation
