Flow_offloading=1 is broken on latest snapshot (4.19 issue)

For now I have to disable flow offloading completely.

I just hope some developers will notice it and be willing to try fixing this mess.

So the fix is in, have not tried it myself.

2 Likes

It totally breaks the Internet..

1 Like

Same here...
//Edit
Works fine now. Thanks :smiley:
Let's see if ipv6 is stable now.

Yep, just flashed an image, she be dead Jim.

There has been an additional change a few minutes ago.
https://git.openwrt.org/?p=openwrt/openwrt.git;a=commitdiff;h=a438eac9ac98d98f594ec56ec023979be41035ca;hp=b2c55d50f8aeb21ef4cd3948034a158f7dd743c2

1 Like

Ya, still a no go for me.

"she be dead Jim" -- is this in general, or just if flow-offloading is enabled?

Edit: Resolved with subsequent pushes.

I did not try turning flowoffload off, just booted back to the other partition; so unsure. Gimme a sec and I will try.

Edit: so the latest patch push, with offload off is working.

2 Likes

Still Broken

Edit: after the recent change
https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=c8933ce53365671de98a303c8d94e089ea76754e
There is already routing and it seems to work very well (tested on a Linksys Wrt32x)

gratitude

1 Like

OK, that seems to be working, thanks much.

1 Like

Works perfectly, and does not break policy routing :slight_smile:

1 Like

ipv6 still breaks after some time.

What is the problem for ipv6?

After some time (~24h) clients lose ipv6 connectivity.
I have to double check if the router also loses ipv6 connectivity.

Now with flow_offloading set to '1', the connection between the ed2k client and the server got disconnected every like 30 minutes. Any idea why software offloading can cause this?

Logs would be a good, first step in diagnosing what @shm0 and @Enig123 are reporting.

Could be some conntrack timeout somewhere, as the connection bypasses the normal routing/firewall/whatever etc. after the initial connection. Or something similar.

No idea if this has any relevance for your problem, but the discussion in the following thread gives two alternative explanations: losing ipv6 connectivity due to ipv6 address timeout at 30 minutes or some multicast problem (no idea what that actually means...).

It is quite possible that as offloading bypasses normal routing to some extent, some trigger that would keep the connection alive, temporary ipv6 address renewed, or something gets also bypassed.

The IPv6 disconnect issue has changed in latest trunk.
It still breaks after some time, but it is possible to reintiate the connection.
In the versions before IPv6 connectivity was completely broken after some time.