Re: [PATCH 5/5] bpf: sockmap, sockhash: test looking up fds

From: Lorenz Bauer <lmb@cloudflare.com>
To: Jakub Sitnicki <jakub@cloudflare.com>
Cc: Shuah Khan <shuah@kernel.org>,
	Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	kernel-team <kernel-team@cloudflare.com>,
	linux-kselftest@vger.kernel.org,
	Networking <netdev@vger.kernel.org>, bpf <bpf@vger.kernel.org>,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 5/5] bpf: sockmap, sockhash: test looking up fds
Date: Wed, 11 Mar 2020 17:24:59 +0000	[thread overview]
Message-ID: <CACAyw9_DC5ewgLX=Qrvmbs3i3YvtuLBFbe1Hr6SJtbE+L1P2eQ@mail.gmail.com> (raw)
In-Reply-To: <87wo7rxal4.fsf@cloudflare.com>

On Wed, 11 Mar 2020 at 13:52, Jakub Sitnicki <jakub@cloudflare.com> wrote:
>
> On Tue, Mar 10, 2020 at 06:47 PM CET, Lorenz Bauer wrote:
> > Make sure that looking up an element from the map succeeds,
> > and that the fd is valid by using it an fcntl call.
> >
> > Signed-off-by: Lorenz Bauer <lmb@cloudflare.com>
> > ---
> >  .../selftests/bpf/prog_tests/sockmap_listen.c | 26 ++++++++++++++-----
> >  1 file changed, 20 insertions(+), 6 deletions(-)
> >
> > diff --git a/tools/testing/selftests/bpf/prog_tests/sockmap_listen.c b/tools/testing/selftests/bpf/prog_tests/sockmap_listen.c
> > index 52aa468bdccd..929e1e77ecc6 100644
> > --- a/tools/testing/selftests/bpf/prog_tests/sockmap_listen.c
> > +++ b/tools/testing/selftests/bpf/prog_tests/sockmap_listen.c
> > @@ -453,7 +453,7 @@ static void test_lookup_after_delete(int family, int sotype, int mapfd)
> >       xclose(s);
> >  }
> >
> > -static void test_lookup_32_bit_value(int family, int sotype, int mapfd)
> > +static void test_lookup_fd(int family, int sotype, int mapfd)
> >  {
> >       u32 key, value32;
> >       int err, s;
> > @@ -466,7 +466,7 @@ static void test_lookup_32_bit_value(int family, int sotype, int mapfd)
> >                              sizeof(value32), 1, 0);
> >       if (mapfd < 0) {
> >               FAIL_ERRNO("map_create");
> > -             goto close;
> > +             goto close_sock;
> >       }
> >
> >       key = 0;
> > @@ -475,11 +475,25 @@ static void test_lookup_32_bit_value(int family, int sotype, int mapfd)
> >
> >       errno = 0;
> >       err = bpf_map_lookup_elem(mapfd, &key, &value32);
> > -     if (!err || errno != ENOSPC)
> > -             FAIL_ERRNO("map_lookup: expected ENOSPC");
> > +     if (err) {
> > +             FAIL_ERRNO("map_lookup");
> > +             goto close_map;
> > +     }
> >
> > +     if ((int)value32 == s) {
> > +             FAIL("return value is identical");
> > +             goto close;
> > +     }
> > +
> > +     err = fcntl(value32, F_GETFD);
> > +     if (err == -1)
> > +             FAIL_ERRNO("fcntl");
>
> I would call getsockopt()/getsockname() to assert that the FD lookup
> succeeded.  We want to know not only that it's an FD (-EBADFD case), but
> also that it's associated with a socket (-ENOTSOCK).
>
> We can go even further, and compare socket cookies to ensure we got an
> FD for the expected socket.

Good idea, thanks!

> Also, I'm wondering if we could keep the -ENOSPC case test-covered by
> temporarily dropping NET_ADMIN capability.

You mean EPERM? ENOSPC isn't reachable, since the map can only be created
with a map_value of 4 or 8.

>
> > +
> > +close:
> > +     xclose(value32);
> > +close_map:
> >       xclose(mapfd);
> > -close:
> > +close_sock:
> >       xclose(s);
> >  }
> >
> > @@ -1456,7 +1470,7 @@ static void test_ops(struct test_sockmap_listen *skel, struct bpf_map *map,
> >               /* lookup */
> >               TEST(test_lookup_after_insert),
> >               TEST(test_lookup_after_delete),
> > -             TEST(test_lookup_32_bit_value),
> > +             TEST(test_lookup_fd),
> >               /* update */
> >               TEST(test_update_existing),
> >               /* races with insert/delete */

-- 
Lorenz Bauer  |  Systems Engineer
6th Floor, County Hall/The Riverside Building, SE1 7PB, UK

www.cloudflare.com