Re: [PATCH bpf-next v2 1/9] bpf: selftests: Use libbpf 1.0 API mode in bpf constructor

From: Andrii Nakryiko <andrii.nakryiko@gmail.com>
To: Yafang Shao <laoar.shao@gmail.com>
Cc: Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Andrii Nakryiko <andrii@kernel.org>, Martin Lau <kafai@fb.com>,
	Song Liu <songliubraving@fb.com>, Yonghong Song <yhs@fb.com>,
	john fastabend <john.fastabend@gmail.com>,
	KP Singh <kpsingh@kernel.org>,
	Networking <netdev@vger.kernel.org>, bpf <bpf@vger.kernel.org>
Subject: Re: [PATCH bpf-next v2 1/9] bpf: selftests: Use libbpf 1.0 API mode in bpf constructor
Date: Sun, 3 Apr 2022 18:26:54 -0700	[thread overview]
Message-ID: <CAEf4BzYGOgrbvobqPBW+1Zdb5W7Cj0WUvQNitnrxJNgSOCnzQQ@mail.gmail.com> (raw)
In-Reply-To: <CAEf4BzZ2U=H-FEft3twSV7RCgTHHVJ8Dt6_RuYMdHdtC17WM1A@mail.gmail.com>

On Sun, Apr 3, 2022 at 6:24 PM Andrii Nakryiko
<andrii.nakryiko@gmail.com> wrote:
>
> On Sun, Apr 3, 2022 at 7:43 AM Yafang Shao <laoar.shao@gmail.com> wrote:
> >
> > In libbpf 1.0 API mode, it will bump rlimit automatically if there's no
> > memcg-basaed accounting, so we can use libbpf 1.0 API mode instead in case

also very eye catching typo: basaed -> based

> > we want to run it in an old kernel.
> >
> > The constructor is renamed to bpf_strict_all_ctor().
> >
> > Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
> > ---
> >  tools/testing/selftests/bpf/bpf_rlimit.h | 26 +++---------------------
> >  1 file changed, 3 insertions(+), 23 deletions(-)
> >
> > diff --git a/tools/testing/selftests/bpf/bpf_rlimit.h b/tools/testing/selftests/bpf/bpf_rlimit.h
> > index 9dac9b30f8ef..d050f7d0bb5c 100644
> > --- a/tools/testing/selftests/bpf/bpf_rlimit.h
> > +++ b/tools/testing/selftests/bpf/bpf_rlimit.h
> > @@ -1,28 +1,8 @@
> >  #include <sys/resource.h>
> >  #include <stdio.h>
> >
> > -static  __attribute__((constructor)) void bpf_rlimit_ctor(void)
> > +static  __attribute__((constructor)) void bpf_strict_all_ctor(void)
>
> well, no, let's get rid of bpf_rlimit.h altogether. There is no need
> for constructor magic when you can have an explicit
> libbpf_set_strict_mode(LIBBPF_STRICT_ALL).
>
> >  {
> > -       struct rlimit rlim_old, rlim_new = {
> > -               .rlim_cur       = RLIM_INFINITY,
> > -               .rlim_max       = RLIM_INFINITY,
> > -       };
> > -
> > -       getrlimit(RLIMIT_MEMLOCK, &rlim_old);
> > -       /* For the sake of running the test cases, we temporarily
> > -        * set rlimit to infinity in order for kernel to focus on
> > -        * errors from actual test cases and not getting noise
> > -        * from hitting memlock limits. The limit is on per-process
> > -        * basis and not a global one, hence destructor not really
> > -        * needed here.
> > -        */
> > -       if (setrlimit(RLIMIT_MEMLOCK, &rlim_new) < 0) {
> > -               perror("Unable to lift memlock rlimit");
> > -               /* Trying out lower limit, but expect potential test
> > -                * case failures from this!
> > -                */
> > -               rlim_new.rlim_cur = rlim_old.rlim_cur + (1UL << 20);
> > -               rlim_new.rlim_max = rlim_old.rlim_max + (1UL << 20);
> > -               setrlimit(RLIMIT_MEMLOCK, &rlim_new);
> > -       }
> > +       /* Use libbpf 1.0 API mode */
> > +       libbpf_set_strict_mode(LIBBPF_STRICT_ALL);
> >  }
> > --
> > 2.17.1
> >