Re: RFC: Implementation of RFC 7217 [A Method for Generating Semantically Opaque Interface Identifiers, with IPv6 Stateless Address Autoconfiguration (SLAAC)]
Date: Wed, 07 May 2025 16:59:49 UTC
On 4/9/25 13:19, Guido Falsi wrote: > On 4/9/25 13:10, Guido Falsi wrote: >> On 4/9/25 12:51, Ronald Klop wrote: >>> Hi, >>> >>> Next to hostuuid you could add a jailname in the mix. >>> >>> That is what ether_gen_addr(9) does to make it easier to prevent >>> collisions while copying jails around or run a jail on a readonly >>> shared base filesystem. >> >> The RFC is very clear on what should be used to derive the address, so >> I'm not very keen on adding things around. >> >> The UUID should be changed when copying jails that run in parallel, >> they ARE different machines. although I am also at fault here. >> >> But the jailname is the correct parameter? This would change the >> address if the name is changed, which could be ok I guess. >> >> I'd also add this parameter only if actually jailed, skipping it for >> the host. >> >> My real issue with this approach is, the RFC is quite detailed on hash >> parameters. Will the implementation still be conforming if adding >> local ones? >> >> > > BTW, this is easy to add and also add conditionally on being jailed or > not, I'd just like some consensus on this before adding, especially the > RFC compliance issue. > A small addition here: I have already taken a look, since the RFC provides for part of the algorithm to be configurable I have some code I'm testing (which I plan to submit after the present code is committed) to add a sysctl to choose what to use for hashing between interface ID, interface name and interface MAC (if that's something the interface has, while ethernet like interfaces are most common today, the world, as you all know, is not all ethernet). I plan to leave the default at the interface name though. -- Guido Falsi <madpilot@FreeBSD.org>