Re: [PATCH] udp: let rx-udp-gro-forwarding aggregate under encapsulation

[Eva Kurchatova (Virtuozzo)]

On 10/19/26 21:19, Willem de Bruijn wrote:
> On Sun, Oct 18, 2026 at 12:54 PM Eva Kurchatova (Virtuozzo)
> <eva.kurchatova@xxxxxxxxxxxxx> wrote:
> > On Wed, 15 Oct 2026 02:56, Willem de Bruijn wrote:
> >   > Uses encapsulation in the same network namespace as where the test
> >   > runs? If different namespaces, the socket lookup should fail and
> >   > thus the behavior be the same as on a host without a tunnel.
> >
> > Different one: the vxlan sat in the init namespace, the test in the
> > ns it creates. The static key is global, though, so the tunnel turns
> > the lookup on everywhere, and the lookup then runs in the test's
> > receiving namespace, where it finds the test's own receiver: a plain
> > socket bound to any address, without UDP_GRO. That is why the check
> > goes from 1 to 10 the moment a tunnel exists anywhere on the host.
> I also don't have a straightforward fix to the test. Maybe just make
> this one optionally SKIP if it fails.
>
> This is the following test in veth.sh, right?
>
> "
> chk_gro() {
>          local msg="$1"
>          local expected=$2
>
>          ip netns exec $BASE$SRC ping -qc 1 $BM_NET_V4$DST >/dev/null
>          NSTAT_HISTORY=$STATS ip netns exec $NS_DST nstat -n
>
>          printf "%-60s" "$msg"
>          ip netns exec $BASE$DST ./udpgso_bench_rx -C 1000 -R 10 &
>          local spid=$!
>          sleep 0.1
>
>          ip netns exec $NS_SRC ./udpgso_bench_tx -4 -s 13000 -S 1300 -M
> 1 -D $BM_NET_V4$DST
>          local retc=$?
>          wait $spid
>          local rets=$?
> "
>
> which gets called both to verify coalsecing and no coalescing
>
> "
> ip netns exec $NS_SRC ethtool -K veth$SRC tx-udp-segmentation off
> chk_gro "        - aggregation with TSO off" 10
>
> ...
>
> ip netns exec $NS_SRC ethtool -K veth$SRC tx-udp-segmentation off
> ip netns exec $NS_DST ethtool -K veth$DST rx-udp-gro-forwarding on
> chk_gro "        - aggregation with TSO off" 1
> "
>
> The full solution would be to add a third true routing namespace in
> between, which would not have that local socket without -G..


I agree, and withdraw the patch. The lookup behaves as designed, yet the 
test asserts it isn't running.

Will follow-up with a fix to the kselftest via the third namespace, with 
your Suggested-by.