Re: [PATCH bpf-next 3/5] selftests/bpf: add selftest for BPF_PROG_TYPE_USER

From: Andrii Nakryiko <andrii.nakryiko@gmail.com>
To: Song Liu <songliubraving@fb.com>
Cc: open list <linux-kernel@vger.kernel.org>,
	bpf <bpf@vger.kernel.org>, Networking <netdev@vger.kernel.org>,
	Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Kernel Team <Kernel-team@fb.com>,
	john fastabend <john.fastabend@gmail.com>,
	KP Singh <kpsingh@chromium.org>,
	Jesper Dangaard Brouer <brouer@redhat.com>,
	Daniel Xu <dlxu@fb.com>
Subject: Re: [PATCH bpf-next 3/5] selftests/bpf: add selftest for BPF_PROG_TYPE_USER
Date: Sun, 2 Aug 2020 22:07:04 -0700	[thread overview]
Message-ID: <CAEf4BzZuZt52M1Ta42D=z3m5jC4gi0K=_dzWhDB8DVV5Dkxwmg@mail.gmail.com> (raw)
In-Reply-To: <C5606E9B-D3EC-4425-82F5-DA5865836D3E@fb.com>

On Sun, Aug 2, 2020 at 9:33 PM Song Liu <songliubraving@fb.com> wrote:
>
>
>
> > On Aug 2, 2020, at 6:43 PM, Andrii Nakryiko <andrii.nakryiko@gmail.com> wrote:
> >
> > On Sat, Aug 1, 2020 at 1:50 AM Song Liu <songliubraving@fb.com> wrote:
> >>
> >> This test checks the correctness of BPF_PROG_TYPE_USER program, including:
> >> running on the right cpu, passing in correct args, returning retval, and
> >> being able to call bpf_get_stack|stackid.
> >>
> >> Signed-off-by: Song Liu <songliubraving@fb.com>
> >> ---
> >> .../selftests/bpf/prog_tests/user_prog.c      | 52 +++++++++++++++++
> >> tools/testing/selftests/bpf/progs/user_prog.c | 56 +++++++++++++++++++
> >> 2 files changed, 108 insertions(+)
> >> create mode 100644 tools/testing/selftests/bpf/prog_tests/user_prog.c
> >> create mode 100644 tools/testing/selftests/bpf/progs/user_prog.c
> >>
> >> diff --git a/tools/testing/selftests/bpf/prog_tests/user_prog.c b/tools/testing/selftests/bpf/prog_tests/user_prog.c
> >> new file mode 100644
> >> index 0000000000000..416707b3bff01
> >> --- /dev/null
> >> +++ b/tools/testing/selftests/bpf/prog_tests/user_prog.c
> >> @@ -0,0 +1,52 @@
> >> +// SPDX-License-Identifier: GPL-2.0
> >> +/* Copyright (c) 2020 Facebook */
> >> +#include <test_progs.h>
> >> +#include "user_prog.skel.h"
> >> +
> >> +static int duration;
> >> +
> >> +void test_user_prog(void)
> >> +{
> >> +       struct bpf_user_prog_args args = {{0, 1, 2, 3, 4}};
> >> +       struct bpf_prog_test_run_attr attr = {};
> >> +       struct user_prog *skel;
> >> +       int i, numcpu, ret;
> >> +
> >> +       skel = user_prog__open_and_load();
> >> +
> >> +       if (CHECK(!skel, "user_prog__open_and_load",
> >> +                 "skeleton open_and_laod failed\n"))
> >> +               return;
> >> +
> >> +       numcpu = libbpf_num_possible_cpus();
> >
> > nit: possible doesn't mean online right now, so it will fail on
> > offline or non-present CPUs
>
> Just found parse_cpu_mask_file(), will use it to fix this.
>
> [...]
>
> >> +
> >> +volatile int cpu_match = 1;
> >> +volatile __u64 sum = 1;
> >> +volatile int get_stack_success = 0;
> >> +volatile int get_stackid_success = 0;
> >> +volatile __u64 stacktrace[PERF_MAX_STACK_DEPTH];
> >
> > nit: no need for volatile for non-static variables
> >
> >> +
> >> +SEC("user")
> >> +int user_func(struct bpf_user_prog_ctx *ctx)
> >
> > If you put args in bpf_user_prog_ctx as a first field, you should be
> > able to re-use the BPF_PROG macro to access those arguments in a more
> > user-friendly way.
>
> I am not sure I am following here. Do you mean something like:
>
> struct bpf_user_prog_ctx {
>         __u64 args[BPF_USER_PROG_MAX_ARGS];
>         struct pt_regs *regs;
> };
>
> (swap args and regs)?
>

Yes, BPF_PROG assumes that context is a plain u64[5] array, so if you
put args at the beginning, it will work nicely with BPF_PROG.

> Thanks,
> Song
>
>