Re: [PATCH bpf-next 08/10] selftests/bpf: demonstrate use of custom .rodata/.data sections

[Andrii Nakryiko <andrii.nakryiko@gmail.com>]

On Mon, Oct 11, 2021 at 3:57 PM Daniel Borkmann <daniel@iogearbox.net> wrote:
>
> On 10/8/21 2:03 AM, andrii.nakryiko@gmail.com wrote:
> > From: Andrii Nakryiko <andrii@kernel.org>
> >
> > Enhance existing selftests to demonstrate the use of custom
> > .data/.rodata sections.
> >
> > Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
>
> Just a thought, but wouldn't the actual demo / use case be better to show that we can
> now have a __read_mostly attribute which implies SEC(".data.read_mostly") section?
>
> Would be nice to add a ...
>
>    #define __read_mostly    SEC(".data.read_mostly")
>
> ... into tools/lib/bpf/bpf_helpers.h along with the series for use out of BPF programs
> as I think this should be a rather common use case. Thoughts?

But what's so special about the ".data.read_mostly" ELF section for
BPF programs? It's just another data section with no extra semantics.
So unclear why we need to have a dedicated #define for that?..

>
> > ---
> >   .../selftests/bpf/prog_tests/skeleton.c       | 25 +++++++++++++++++++
> >   .../selftests/bpf/progs/test_skeleton.c       | 10 ++++++++
> >   2 files changed, 35 insertions(+)
> [...]
> > diff --git a/tools/testing/selftests/bpf/progs/test_skeleton.c b/tools/testing/selftests/bpf/progs/test_skeleton.c
> > index 441fa1c552c8..47a7e76866c4 100644
> > --- a/tools/testing/selftests/bpf/progs/test_skeleton.c
> > +++ b/tools/testing/selftests/bpf/progs/test_skeleton.c
> > @@ -40,9 +40,16 @@ int kern_ver = 0;
> >
> >   struct s out5 = {};
> >
> > +const volatile int in_dynarr_sz SEC(".rodata.dyn");
> > +const volatile int in_dynarr[4] SEC(".rodata.dyn") = { -1, -2, -3, -4 };
> > +
> > +int out_dynarr[4] SEC(".data.dyn") = { 1, 2, 3, 4 };
> > +
> >   SEC("raw_tp/sys_enter")
> >   int handler(const void *ctx)
> >   {
> > +     int i;
> > +
> >       out1 = in1;
> >       out2 = in2;
> >       out3 = in3;
> > @@ -53,6 +60,9 @@ int handler(const void *ctx)
> >       bpf_syscall = CONFIG_BPF_SYSCALL;
> >       kern_ver = LINUX_KERNEL_VERSION;
> >
> > +     for (i = 0; i < in_dynarr_sz; i++)
> > +             out_dynarr[i] = in_dynarr[i];
> > +
> >       return 0;
> >   }
> >
> >
>