increase MAX_IF to 100

[fichtner]

Several users of OPNsense and pfSense report a persistent error
with their configuration: "There must be at least 1 Vif as upstream."

It seems that there is no real error in igmpproxy, but rather with
the way that network addresses are looked up prior to starting up.
This can break when many addresses are present in the system.

There is no exact science for the increase to 100 and previous custom
patches on both projects have had this raised to 80 back when version
0.1 was used. There could be further work required to dynamically
parse addresses in a way that this raised limit does not cause further
issues in the future.

See also: opnsense/plugins#590
See also: https://redmine.pfsense.org/issues/8935

[fichtner]

As pointed out via #38 (comment) and further patch analysis this seems to solve the issues at hand, see bug references in OPNsense and pfSense.

I want to upstream more patches from FreeBSD so 0.2.2 has no priority. FreeBSD port was updated in parallel: https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=231800

[pali]

I really do not like an idea to change upper limit from one magic constant to another without exact reason. You also wrote There is no exact science for the increase to 100.

Therefore rather completely remove that array limit and allocate array dynamically to size which is needed on running system.

And if you really really think that there should be some upper limit then make it configurable via config file.

[fichtner]

I agree, your project, your rules. I also think this is the best way forward for the time being for approaching a bug that existed for many years. And it’s hard to argue against it from a technical perspective: it’s in line with your code, it doesn’t introduce regressions and it doesn’t require additional vetting.

[pali]

Main problem is that I still do not see "why 100". Why it is enough now and also for future? It does not matter if it is one line change or 100 lines change in code. Basically it is not a fix, just workaround about which nobody knows why is working and if it would work also in future.

[fichtner]

Yes, and I said so initially. We should be beyond this talking point by now, but we are not.