[SRU][J][PATCH 0/1] Fix udpgro.sh selftest failure on Jammy 5.15

[Aaron Ma]

BugLink: https://bugs.launchpad.net/bugs/2129861

SRU Justification:

[ Impact ]
Running `tools/testing/selftests/net/udpgro.sh` fails on Jammy 5.15 kernels with:
  GRO ./udpgso_bench_rx: recv: bad packet len, got 1472, expected 14720

In commit 886a0a957ce0 ("net: veth: do not manipulate GRO when using XDP")
(upstream commit d7db7775ea2e), veth was changed so that attaching an XDP
program no longer automatically enables the GRO feature bit. However, udpgro.sh
previously relied on attaching xdp_dummy.o to trigger GRO.

Furthermore, upstream commit 7167395a4be7 ("selftests: udpgro: report error
when receive failed") landed in Jammy 5.15 in commit f83d21a592ca (via v5.15.190),
causing udpgro.sh to check the receiver process exit code and actively report the
failure that had previously been silently ignored.

[ Fix ]
Backport upstream commit d7818402b1d8 ("selftests: udpgro: no need to load
xdp for gro"). This replaces the XDP dummy program attachment with explicit
`ethtool -K veth1 gro on` and removes the dependency on xdp_dummy.o.

Context was adjusted due to the absence of commit 6b88ce902f0b ("selftests:
net: name bpf objects consistently and simplify Makefile") in 5.15.

[ Test Plan ]
Build and execute the udpgro test:
  cd tools/testing/selftests/net
  make udpgso_bench_tx udpgso_bench_rx
  sudo ./udpgro.sh

All test cases must report "ok" with exit code 0.
Verified on a target machine running Jammy 5.15 kernel (5.15.0-191-generic).

[ Where problems could occur ]
Changes are strictly isolated to tools/testing/selftests/net/udpgro.sh and
do not touch any kernel code or runtime drivers. Risk of regression in
production environments is none.

Hangbin Liu (1):
  selftests: udpgro: no need to load xdp for gro

 tools/testing/selftests/net/udpgro.sh | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)

-- 
2.53.0