Daniel LakelandMar 12

Ok, my deeply #linux involved #ipv6 nerds. Here's my problem.

When I ping from machine A to a router via ipv6 let's call it fd01::9 the neighbor discovery packet goes through the switch to the host machine, and hits the linux bridge... but after a few minutes, it doesn't get forwarded over the tap108i0 device to the VM where the router is.

Shortly after I restart the VM networking it works... then it times out and stops forwarding. Why?

Show threadDaniel LakelandMar 12

After vm networking restart, the bridge reports:

dev vmbr0 port tap108i0 grp ff02::1:ff00:9 temp

I'm watching it for several minutes to see if it disappears.

Show threadDaniel LakelandMar 12

Yes, it disappears and then I can't ND to the router, and ping won't work.

Why does the #linux bridge lose track of the fact the router wants that #multicast group?

#networking

Show threadMalte StretzMar 12
@dlakelan
There is an ancient bug in the network stack which causes this issue (at least I assume that it is a bug). I tend to forget about this and run into this issue every few years. There is a sysctl knob you have to set, IIRC you have to disable multicast snooping. I'd have to look it up but you might find a related rant from last year in my timeline.
Show threadMalte Stretz

@dlakelan Found it, it was already two years since I ran into that issue again https://norden.social/@mss/112087733854384530

Looks like shorty after I re-fixed it on my side (via `bridge-mcsnoop no`) some discussion started on the bug again. Maybe @T_X recent work will fix it as a side effect?

Mar 18 at 10:24amWeb
Show threadDaniel LakelandMar 18
@mss
thanks for reminding me. i was supposed to test some things related to this, but my wife was staying home to do zoom meetings the last two days so I need to revisit this afternoon.
@T_X