On Tuesday, 9. December 2014 06:26:49 Eric Dumazet wrote:

quoted

If it helps, I'm running the reverted patch on five production boxes
hitherto without a hiccup. As far as I understood the original commit
message, some packet counters might me wrong without it.

@Eric: What could possibly go wrong(tm)? :)

Crashes in TCP stack, because of packet count mismatches.

alright, that sounds like a pretty good argument.

...
I would disable TSO/GSO on xfrm, and problem should disappear.

I guess you can't explicitly disable this with the "ip xfrm" command?
Or do you mean this should be disabled on the ethX device
serving the xfrm connection?

We are about to push out this code to ten more machines,
so the best time (for me) to do any changes  that increases
stability would be now :o)

Cheers,
Thomas