From: Richard Cochran <richardcoch...@gmail.com> Sent: 05 December 2022 04:16
>On Fri, Nov 11, 2022 at 07:53:46AM -0800, Vadim Fedorenko via Linuxptp-devel >wrote: >> The latest proposal for DC PTP Profile[1] provides logic to make load >> balancing of unicast clients when running multiple GMs in active-active >> mode. This patch adds this logic to ptp4l using already existing >> localPriority field for telecom profile. Together with telecom BMCA >> algorithm it implements sharding based on provided GM priority. >> >> https://www.opencompute.org/documents/ocp-dc-ptp-profile-v1r1-pdf-2 > > I'll need some time to digest that... > > but in the mean time, > >> @@ -677,18 +678,27 @@ static int parse_unicast_mtab_line(struct config *cfg, >> char *line, int line_num) >> if (cnt == 1) { >> return config_unicast_mtab_peer(address, line_num); >> } >> - cnt = sscanf(line, " %16s %64s", transport, address); >> - if (cnt != 2) { >> + cnt = sscanf(line, " %16s %64s %d", transport, address, &prio); >> + if (cnt < 2) { >> fprintf(stderr, "bad master table at line %d\n", line_num); >> return -1; >> } >> + if (cnt == 3 && (prio < 0 || prio > 255)) { >> + fprintf(stderr, "bad address priority(%d) at line %d\n", prio, >> line_num); >> + prio = 0; >> + } >> + if (cnt == 2) { >> + fprintf(stderr, "address priority is not parsed at line %d\n", >> line_num); >> + prio = 0; >> + } > >You are adding a third, optional element. In order not to alarm >existing users, it is better not to print warning when only two values >are present, hm? It's a good point, but it will hide an issue when third element is provided in a form that cannot be parsed as a digit. I do agree that there is no need to alarm existing users and will try to find a better way to check for 3rd element. Or will send v2 with simply removing this print if there is no better way. Thanks, Vadim _______________________________________________ Linuxptp-devel mailing list Linuxptp-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linuxptp-devel
