Fix crash upon PSDBU report triggered on a multi-buffer frame

[ivan4th]

Make sure that the session is not deleted multiple times

[RoadRunnr]

This whole problem makes we wonder if the current URR event queuing model should be changed (but not in this bugfix).
It might be worth having a outstanding URR events queue per session and only queue a trigger to process the URRs to the PFCP server.
That way additional events can be added to the URR queue for a session while the trigger is in the queue. Of course, some mutual exclusion on the per session URR event buffer would be needed.

[RoadRunnr]

The continue skips over the vec_free_h at the end. Wouldn't this cause a leak of the ueh entry?

[ivan4th]

Thanks for noticing this problem. Adjusted the code slightly to avoid the leak

[ivan4th]

I agree concerning the per-session URR event queue (to be done later)

[RoadRunnr]

LGTM

[RoadRunnr]

Those lines are fine, still I would have coded this as:

Suggested change

	if (!sx || sx->cp_seid != ueh->cp_seid)
	{
	vec_free_h (uev, sizeof (upf_event_urr_hdr_t));
	continue;
	}
	if (!sx || sx->cp_seid != ueh->cp_seid)
	{
	goto next_ev_urr;
	}

with a the goto label` place right before the free.

That way it would be harder to accidentally use different cleanups for both cases.

[ivan4th]

Updated the code to use goto, as indeed we might introduce a hard-to-catch bug later with different cleanup code paths