I have this small network with 4 hAP ac and 4 hAPac2, so it is still running the old Capsman, and today I observed a strange thing under “Last IP”: sometimes it is OK, showing the IP of the client, but sometimes it is in reversed byte order (see the picture, top vs bottom half).
I took me more than two minutes to understand that those IP’s are “just” reversed, as oposed to something trying to access random (but similar, except for the first byte) addresses on the Internet.
Not a big problem, this whole setup will be soon migrated to wifi-qcom(-ac), just an observation…
Actually I don’t want to lose time on this, it is not important enough for me, I just wanted to report my observation
If they (or anyone else) can replicate that - fine, if not - then it’s probably just something in my config (really simple: one bridge, capsman forwarding to that bridge, DHCP server, MASQUERADE everything)
In other places I don’t have capsman forwarding, so no “Last IP” either, so I cannot compare.
Can you see whether the "wrong" IP addresses are associated with the hAP acs or the hAP ac²s? My guess is that the hAP acs having a MIPSBE processor are big-endian, and the hAP ac²s with ARM Cortex-A53 while supporting both endiannesses use little-endian by default, and MikroTik is sloppy in the part of the code that read or write the last IP values (as binary) when sending them from the APs to CAPSMAN and did not use the correct functions/macros that handle the difference between machine endianness and serialized data endianness.
Besides the (evidently little) severity of this bug, it is so basic that it cannot possibly have passed, besides the developers' own checks, any quality control procedure.
I find it a good (meaning bad) example of general sloppiness, the same that is evident in documentation, products page and what not.