ACL-Rule Behaviour for TCP/UDP-Protocol without start- and endport

[mwaag]

Hi,

we noticed cloudstack let you successfully define ACL-Ingress-Rules for TCP (and UDP) without setting a start- and endport.
Many of our users (even we) assumed, that it stands for 'all ports'. But instead the router keeps on blocking traffic.
(We didn't test this on UDP explicitly)

We know, we can workaround this with just setting start- and endports or use protocol: All

Is this expected behaviour or should this be handled as a bug?
(We probably would suggest to either restrict defining rules without setting start- and endports at all or treat this kind of rules as "all ports" - rule)

Tested Versions are:
4.18.2.4
4.20.3.0

[weizhouapache]

people may have different opinions whether it is expected or a bug

to avoid confusion, we could mark start port and end port as required on UI, whart do you think ?

[mwaag]

Yeah, i fully agree to different possible opinions about it.

Making these fields required would help, i think 👍

[weizhouapache]

created issue #13122