If a port goes down because of loss of link, ptp4l tries to avoid
clearing the fault. However, if you set fault_reset_interval to ASAP,
then ptp4l immediately clears the fault and tries to send and eventually
gets a timeout.
Is this expected behavior? I feel like we might want to be checking the
port link state to ensure that we don't clear the fault if link on the
port is still down (since it will still be unusable).
> int port_state_update(struct port *p, enum fsm_event event, int mdiff)
> {
> enum port_state next = p->state_machine(p->state, event, mdiff);
>
> if (PS_FAULTY == next) {
> struct fault_interval i;
> fault_interval(p, last_fault_type(p), &i);
> if (clear_fault_asap(&i)) {
> pr_notice("%s: clearing fault immediately",
> p->log_name);
> next = p->state_machine(next, EV_FAULT_CLEARED, 0);
> }
For context, I'm looking at this block
> }
>
> if (PS_INITIALIZING == next) {
> /*
> * This is a special case. Since we initialize the
> * port immediately, we can skip right to listening
> * state if all goes well.
> */
> if (port_is_enabled(p)) {
> port_disable(p);
> }
> if (port_initialize(p)) {
> event = EV_FAULT_DETECTED;
> } else {
> event = EV_INIT_COMPLETE;
> }
> next = p->state_machine(next, event, 0);
> }
>
> if (mdiff) {
> unicast_client_state_changed(p);
> }
> if (next != p->state) {
> port_show_transition(p, next, event);
> p->state = next;
> port_notify_event(p, NOTIFY_PORT_STATE);
> unicast_client_state_changed(p);
> return 1;
> }
>
> return 0;
> }
>
the block where we issue a fault cleared normally is:
>
> /*
> * When the fault timer expires we clear the fault,
> * but only if the link is up.
> */
> if (cur[N_POLLFD].revents & (POLLIN|POLLPRI)) {
> clock_fault_timeout(p, 0);
> if (port_link_status_get(p)) {
> port_dispatch(p, EV_FAULT_CLEARED, 0);
> }
> }
>
We check port_link_status_get to determine if we still have link. It
seems like we ought to check this when immediately clearing the fault
too, no?
In the setup I have, link goes down and we were using an immediate fault
clear while attempting to debug issues with Tx timestamps (Yep, i know.
ugh).
Thanks,
Jake
_______________________________________________
Linuxptp-devel mailing list
Linuxptp-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linuxptp-devel
