Closed toreanderson closed 9 years ago
Yet another argument in favor of #140. Once traffic intended to be translated has to be manually routed towards Jool, it will naturally stop seeing link-local traffic.
As a workaround, 224.0.0.0/24 /4 can be manually blacklisted; correct?
As a workaround, 224.0.0.0/24 can be manually blacklisted; correct?
Correct. So this is just a suggestion, it doesn't cause me any operational problems in any way.
Tore Anderson notifications@github.com wrote:
The last sentence there makes me think that IPv4 multicast (224.0.0.0/4) should probably be implicitly blacklisted. At the very least, IPv4 multicast addresses which the local node has subscribed to should be - there's some application running on the Jool node that's interested in those packets; Jool shouldn't steal them.
I can't see a good reason to translate class D addresses (multicast). Probably should blacklist 127/8, 0/8 and some other invalid things. I don't know what to do with 240/16 (class E).
Probably should blacklist 127/8, 0/8 and some other invalid things.
Don't worry; those are already blacklisted. This is the relevant code.
I don't know what to do with 240/16 (class E).
Well, since they are experimental, I don't think we should completely ban the user from doing whatever with them.
They can always be manually blacklisted if use cases or security implications arise.
The code blacklists 224.0.0.0/4 https://github.com/NICMx/NAT64/blob/master/mod/common/blacklist.c
But this documentation doesn't mention that range can't be used: https://www.jool.mx/usr-flags-blacklist.html
Fixed; thank you!
I've noticed that Jool will happily by default translate traffic destined for multicast addresses. Here's an example showing a VRRP packet being translated:
RFC 6052 is silent on the subject of multicast, while RFC 6145 says:
The last sentence there makes me think that IPv4 multicast (224.0.0.0/4) should probably be implicitly blacklisted. At the very least, IPv4 multicast addresses which the local node has subscribed to should be - there's some application running on the Jool node that's interested in those packets; Jool shouldn't steal them.