https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=268246
--- Comment #70 from Kristof Provost <k...@freebsd.org> --- (In reply to jjasen from comment #68) Okay, excellent. I think we've figured out what's going on here. I also think I see why I can't reproduce it yet. It's interpreting the IPv6 header as an IPv4 header (and we didn't notice so far, because the ip_fragment() code overwrites the version field in the header), and it's reading the flow label as length. We're defaulting to a flow label of 0, and that's putting us in a code path that avoids ip_fragment(). (The "If small enough for interface" case in ip_output()) I'm having an argument with some of the test tools to try to get a flow label set so I can make it panic locally. It should be relatively straightforward to fix this, but I'm going to be on poor connectivity for a few days, so it won't be for tomorrow. The pfsync_defer_tmo fix is in stable/13 and will be part of 13.2, but it's not in releng/13.1 (and never will be). -- You are receiving this mail because: You are the assignee for the bug.
