Re: [PATCH bpf-next 5/7] selftests: bpf: add test for sk_assign

From: Martin KaFai Lau <kafai@fb.com>
To: Joe Stringer <joe@wand.net.nz>
Cc: <bpf@vger.kernel.org>, Lorenz Bauer <lmb@cloudflare.com>,
	netdev <netdev@vger.kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Alexei Starovoitov <ast@kernel.org>,
	Eric Dumazet <eric.dumazet@gmail.com>,
	Andrii Nakryiko <andriin@fb.com>
Subject: Re: [PATCH bpf-next 5/7] selftests: bpf: add test for sk_assign
Date: Wed, 18 Mar 2020 10:27:35 -0700	[thread overview]
Message-ID: <20200318172735.kxwuvccegquupkwh@kafai-mbp> (raw)
In-Reply-To: <CAOftzPjBo6r2nymjUn4qr=N4Zd7rF=03=n45HDvyXfSXfDnBtg@mail.gmail.com>

On Tue, Mar 17, 2020 at 01:56:12PM -0700, Joe Stringer wrote:
> On Tue, Mar 17, 2020 at 12:31 AM Martin KaFai Lau <kafai@fb.com> wrote:
> >
> > On Thu, Mar 12, 2020 at 04:36:46PM -0700, Joe Stringer wrote:
> > > From: Lorenz Bauer <lmb@cloudflare.com>
> > >
> > > Attach a tc direct-action classifier to lo in a fresh network
> > > namespace, and rewrite all connection attempts to localhost:4321
> > > to localhost:1234.
> > >
> > > Keep in mind that both client to server and server to client traffic
> > > passes the classifier.
> > >
> > > Signed-off-by: Lorenz Bauer <lmb@cloudflare.com>
> > > Signed-off-by: Joe Stringer <joe@wand.net.nz>
> > > ---
> > >  tools/testing/selftests/bpf/.gitignore        |   1 +
> > >  tools/testing/selftests/bpf/Makefile          |   3 +-
> > >  .../selftests/bpf/progs/test_sk_assign.c      | 127 +++++++++++++
> > >  tools/testing/selftests/bpf/test_sk_assign.c  | 176 ++++++++++++++++++
> > Can this test be put under the test_progs.c framework?
> 
> I'm not sure, how does the test_progs.c framework handle the logic in
> "tools/testing/selftests/bpf/test_sk_assign.sh"?
> 
> Specifically I'm looking for:
> * Unique netns to avoid messing with host networking stack configuration
> * Control over routes
> * Attaching loaded bpf programs to ingress qdisc of a device
> 
> These are each trivial one-liners in the supplied shell script
> (admittedly building on existing shell infrastructure in the tests dir
> and iproute2 package). Seems like maybe the netns parts aren't so bad
> looking at flow_dissector_reattach.c but anything involving netlink
> configuration would either require pulling in a netlink library
> dependency somewhere or shelling out to the existing binaries. At that
> point I wonder if we're trying to achieve integration of this test
> into some automated prog runner, is there a simpler way like a place I
> can just add a one-liner to run the test_sk_assign.sh script?
I think running a system(cmd) in test_progs is fine, as long as it cleans
up everything when it is done.  There is some pieces of netlink
in tools/lib/bpf/netlink.c that may be reuseable also.

Other than test_progs.c, I am not aware there is a script to run
all *.sh.  I usually only run test_progs.

Cc: Andrii who has fixed many selftest issues recently.