We should be able to independently flip either XDP or GRO states and toggling
one should not affect the other.

Signed-off-by: Ignat Korchagin <ignat@cloudflare.com>
---
 tools/testing/selftests/net/veth.sh | 24 +++++++++++++++++++++---
 1 file changed, 21 insertions(+), 3 deletions(-)

diff --git a/tools/testing/selftests/net/veth.sh b/tools/testing/selftests/net/veth.sh
index 5ae85def0739..3a394b43e274 100755
--- a/tools/testing/selftests/net/veth.sh
+++ b/tools/testing/selftests/net/veth.sh
@@ -249,9 +249,9 @@ cleanup
 create_ns
 ip -n $NS_DST link set dev veth$DST up
 ip -n $NS_DST link set dev veth$DST xdp object ${BPF_FILE} section xdp
-chk_gro_flag "gro vs xdp while down - gro flag on" $DST on
+chk_gro_flag "gro vs xdp while down - gro flag off" $DST off
 ip -n $NS_DST link set dev veth$DST down
-chk_gro_flag "                      - after down" $DST on
+chk_gro_flag "                      - after down" $DST off
 ip -n $NS_DST link set dev veth$DST xdp off
 chk_gro_flag "                      - after xdp off" $DST off
 ip -n $NS_DST link set dev veth$DST up
@@ -260,6 +260,21 @@ ip -n $NS_SRC link set dev veth$SRC xdp object ${BPF_FILE} section xdp
 chk_gro_flag "                      - after peer xdp" $DST off
 cleanup
 
+create_ns
+ip -n $NS_DST link set dev veth$DST up
+ip -n $NS_DST link set dev veth$DST xdp object ${BPF_FILE} section xdp
+ip netns exec $NS_DST ethtool -K veth$DST generic-receive-offload on
+chk_gro_flag "gro vs xdp while down - gro flag on" $DST on
+ip -n $NS_DST link set dev veth$DST down
+chk_gro_flag "                      - after down" $DST on
+ip -n $NS_DST link set dev veth$DST xdp off
+chk_gro_flag "                      - after xdp off" $DST on
+ip -n $NS_DST link set dev veth$DST up
+chk_gro_flag "                      - after up" $DST on
+ip -n $NS_SRC link set dev veth$SRC xdp object ${BPF_FILE} section xdp
+chk_gro_flag "                      - after peer xdp" $DST on
+cleanup
+
 create_ns
 chk_channels "default channels" $DST 1 1
 
@@ -327,11 +342,14 @@ if [ $CPUS -gt 2 ]; then
 fi
 
 ip -n $NS_DST link set dev veth$DST xdp object ${BPF_FILE} section xdp 2>/dev/null
-chk_gro_flag "with xdp attached - gro flag" $DST on
+chk_gro_flag "with xdp attached - gro flag" $DST off
 chk_gro_flag "        - peer gro flag" $SRC off
 chk_tso_flag "        - tso flag" $SRC off
 chk_tso_flag "        - peer tso flag" $DST on
 ip netns exec $NS_DST ethtool -K veth$DST rx-udp-gro-forwarding on
+chk_gro "        - no aggregation" 10
+ip netns exec $NS_DST ethtool -K veth$DST generic-receive-offload on
+chk_gro_flag "        - gro flag with GRO on" $DST on
 chk_gro "        - aggregation" 1
 
 
-- 
2.39.2

On Tue, Mar 12, 2024 at 04:05:52PM +0000, Ignat Korchagin wrote:
> We should be able to independently flip either XDP or GRO states and toggling
> one should not affect the other.
> 
> Signed-off-by: Ignat Korchagin <ignat@cloudflare.com>

Missing "Fixes" tag for the patch targeted to the "net" tree.

Thanks,
Michal

On Wed, 13 Mar 2024 12:28:41 +0100 Michal Kubiak wrote:
> On Tue, Mar 12, 2024 at 04:05:52PM +0000, Ignat Korchagin wrote:
> > We should be able to independently flip either XDP or GRO states and toggling
> > one should not affect the other.
> > 
> > Signed-off-by: Ignat Korchagin <ignat@cloudflare.com>  
> 
> Missing "Fixes" tag for the patch targeted to the "net" tree.

it's adjusting a selftest, I don't think we need a Fixes tag for that

On Wed, Mar 13, 2024 at 06:57:25AM -0700, Jakub Kicinski wrote:
> On Wed, 13 Mar 2024 12:28:41 +0100 Michal Kubiak wrote:
> > On Tue, Mar 12, 2024 at 04:05:52PM +0000, Ignat Korchagin wrote:
> > > We should be able to independently flip either XDP or GRO states and toggling
> > > one should not affect the other.
> > > 
> > > Signed-off-by: Ignat Korchagin <ignat@cloudflare.com>  
> > 
> > Missing "Fixes" tag for the patch targeted to the "net" tree.
> 
> it's adjusting a selftest, I don't think we need a Fixes tag for that

OK, sorry! My mistake, then.

Michal

On Wed, 13 Mar 2024 15:03:16 +0100 Michal Kubiak wrote:
> > > Missing "Fixes" tag for the patch targeted to the "net" tree.  
> > 
> > it's adjusting a selftest, I don't think we need a Fixes tag for that  
> 
> OK, sorry! My mistake, then.

No worries, it's pretty subjective. And the patchwork check for
the Fixes tag doesn't do a great job on the corner cases either :(