tcpsso: fix when used without -i option

[tuexen]

Summary

It looks like that since fdb987bebddf05e15a5af840379c7715a94aec1c it is not possible anymore to use the inp_next iterator
for bound but unconnected sockets. In the TCP case, these are listening sockets.
This breaks tcpsso when used without the -i option.
Fallback to iterate through all endpoints, which is not as performant as it was, but un-breaks tcpsso.

[markj]

Could you please add a regression test which uses tcpsso (or at least this sysctl interface), so that this kind of bug can be caught quickly next time?

[tuexen]

In D43353#988054, @markj wrote:

Could you please add a regression test which uses tcpsso (or at least this sysctl interface), so that this kind of bug can be caught quickly next time?

Are there examples for tests of programs in /usr/sbin/?

[markj]

In D43353#988055, @tuexen wrote:

In D43353#988054, @markj wrote:

Could you please add a regression test which uses tcpsso (or at least this sysctl interface), so that this kind of bug can be caught quickly next time?

Are there examples for tests of programs in /usr/sbin/?

find usr.sbin/ -name tests turns up some examples. There are some in usr/share/examples/tests as well.

[tuexen]

In D43353#988056, @markj wrote:

In D43353#988055, @tuexen wrote:

In D43353#988054, @markj wrote:

Could you please add a regression test which uses tcpsso (or at least this sysctl interface), so that this kind of bug can be caught quickly next time?

Are there examples for tests of programs in /usr/sbin/?

find usr.sbin/ -name tests turns up some examples. There are some in usr/share/examples/tests as well.

Thanks for the pointers. Will figure out something an bring it up as a separate review/commit.