24.7.2 IPv6 woes

[franco]

1. The srandom() call was omitted. The arc4random documentation from OpenBSD says: "The subsystem is re-seeded from the kernel random(4) subsystem using getentropy(2) on a regular basis, and also upon fork(2)." Maybe one should still call srandom() to initialize arc4random alongside?

The call is hidden under !HAVE_ARC4RANDOM so it wouldn't compile. But I actually need to put the shims back in order for that define to be created. -.-


Cheers,
Franco

[franco]

@CruxtheNinth so I think you making clear that the last one "somewhat" worked I reverted the other commit that seems to play a role here and this should be the final thing for 24.7.4:

# pkg add -f https://pkg.opnsense.org/FreeBSD:14:amd64/snapshots/misc/dhcp6c-20240907.pkg

I'm actually glad that neither of the two fixes for operational issues we did caused this, but it's still a bit unfortunate there is something weird going on here with the random value generation.


Cheers,
Franco

[CruxtheNinth]

@CruxtheNinth so I think you making clear that the last one "somewhat" worked I reverted the other commit that seems to play a role here and this should be the final thing for 24.7.4:

# pkg add -f https://pkg.opnsense.org/FreeBSD:14:amd64/snapshots/misc/dhcp6c-20240907.pkg

I'm actually glad that neither of the two fixes for operational issues we did caused this, but it's still a bit unfortunate there is something weird going on here with the random value generation.


Cheers,
Franco

thank you. dhcp6c-20240907.pkg running without any issues so far. still feels like 24.7.1 reconciles a bit faster but i think thats me and just caused by DGs general weird non-standard behaviour.

How come not more people were affected by this?

[franco]

I think that DG is very sensitive, but the problem was probably in dhcp6c for decades, if not forever. I made it worse and DG suddenly stopped working.

I have the issue here sometimes, but not very reliable. It's probably the same for most.


Cheers,
Franco

[FatWightMan]

I had the same problem with my ISP WightFibre.  The dhcp6c-20240907.pkg has fixed things for me too.

[franco]

@FatWightMan thanks, appreciate the feedback!

[meyergru]

FWIW, I sometimes had slow buildup of IPv6 on DG as well, but this was hit rather seldomly.

Franco, did you fix the "strictly greater than 0" case with that dhcpc version or switched back to random()? Will that new version be in 24.7.4 alongside the no-sa kernel?

[franco]

Essentially I've only reverted the bulk of the two commits around the arc4random() rework. I'll be dissecting them a bit more and fix whatever seems to come up during that. It may take more time though.

And the no_sa kernel is just for reference. We're going with the same kernel side as 24.7.3 which is half the patches and disabling ND states. FreeBSD is in too deep on this, so eventually they will fix it. That's my hope at least.


Cheers,
Franco

[Server07]

Good morning,

since I updated to OPNsense 24.7.3_1-amd64 I have the same issue on not running IPV6DHCP.

Reading through did not see clear the solution so far... do I need to return to older version or did any tweak help? I went through all 5 pages but maybe missed the key sentence, maybe also due to lack of coffee this morning....

[Server07]

the IPv6 should not be local with start fe80... or?

[CruxtheNinth]

Good morning,

since I updated to OPNsense 24.7.3_1-amd64 I have the same issue on not running IPV6DHCP.

Reading through did not see clear the solution so far... do I need to return to older version or did any tweak help? I went through all 5 pages but maybe missed the key sentence, maybe also due to lack of coffee this morning....

to install the  new "fixed" version for testing:

pkg add -f https://pkg.opnsense.org/FreeBSD:14:amd64/snapshots/misc/dhcp6c-20240907.pkg

OR

revert to previously known working from 24.7.1

opnsense-revert -r 24.7.1 dhcp6c

reboot afterwards

[CruxtheNinth]

the IPv6 should not be local with start fe80... or?

valid but not what you are looking for in this case.
FE80::/10 is link-local and not routed, you are missing a GUA out of 2000::/3

[franco]

the IPv6 should not be local with start fe80... or?

valid but not what you are looking for in this case.
FE80::/10 is link-local and not routed, you are missing a GUA out of 2000::/3

No, all of this question is completely out of context and irrelevant. There are literally dozens of threads about this topic.


Cheers,
Franco

[meyergru]

And the no_sa kernel is just for reference. We're going with the same kernel side as 24.7.3 which is half the patches and disabling ND states. FreeBSD is in too deep on this, so eventually they will fix it. That's my hope at least.

But there is the new problem with chopped IPv6 packets that doktornotor has reported and to which - again - there was no reaction at all. I am not even talking about the one with the sloppy IPv4 TCP states which seems unrelated as it is not fixed by no-sa.

[franco]

At this point I'm not surprised (mostly because I predicted so in the upstream report for the SA regressions), but there will be no wiggle room for 24.7.4 as we try to stabilise from this and dhcp6c issues being found at the same time. I can't have this go back and forth all the time for each release as it masquerades where actual IPv6 issues are coming from.


Cheers,
Franco