Re: [PATCH net-next v2] selftests: net: Add support for testing SO_RCVMARK and SO_RCVPRIORITY

[Willem de Bruijn <willemdebruijn.kernel@gmail.com>]

Anna Nyiri wrote:
> Paolo Abeni <pabeni@redhat.com> ezt írta (időpont: 2025. febr. 11., K, 12:19):
> >
> > On 2/10/25 8:22 PM, Anna Emese Nyiri wrote:
> > > Introduce tests to verify the correct functionality of the SO_RCVMARK and
> > > SO_RCVPRIORITY socket options.
> > >
> > > Key changes include:
> > >
> > > - so_rcv_listener.c: Implements a receiver application to test the correct
> > > behavior of the SO_RCVMARK and SO_RCVPRIORITY options.
> > > - test_so_rcv.sh: Provides a shell script to automate testing for these options.
> > > - Makefile: Integrates test_so_rcv.sh into the kernel selftests.
> > >
> > > v2:
> > >
> > > - Add the C part to TEST_GEN_PROGS and .gitignore.
> > > - Modify buffer space and add IPv6 testing option
> > > in so_rcv_listener.c.
> > > - Add IPv6 testing, remove unnecessary comment,
> > > add kselftest exit codes, run both binaries in a namespace,
> > > and add sleep in test_so_rcv.sh.
> > > The sleep was added to ensure that the listener process has
> > > enough time to start before the sender attempts to connect.
> > > - Rebased on net-next.
> > >
> > > v1:
> > >
> > > https://lore.kernel.org/netdev/20250129143601.16035-2-annaemesenyiri@gmail.com/
> >
> > Unfortunately the added self-test does not run successfully in the CI:
> 
> I think the test is not running because it is added to TEST_GEN_PROGS.
> However, after reconsidering, I'm not sure it should be there, since
> this test does not run on its own but is executed by the
> test_so_rcv.sh shell script.
> Wouldn't it be more appropriate to add so_rcv_listener to
> TEST_GEN_FILES instead?

Yes,

+TEST_PROGS += test_so_rcv.sh

is correct, as this needs to be called from kselftest.

+TEST_GEN_PROGS += so_rcv_listener

needs to be TEST_GEN_FILES as it is not intended to be called directly.

Documentation/dev-tools/kselftest.rst for details.