Re: [PATCH bpf-next v1] bpftool: Add support for custom BTF path in prog load/loadall

["Jiayuan Chen" <jiayuan.chen@linux.dev>]

2025/5/15 24:52, "Quentin Monnet" <qmo@kernel.org> wrote:

> 
> 2025-05-14 09:39 UTC-0700 ~ Andrii Nakryiko <andrii.nakryiko@gmail.com>
> 
> > 
> > On Tue, May 13, 2025 at 6:51 PM Jiayuan Chen <jiayuan.chen@linux.dev> wrote:
> > 
> > > 
> > > 2025/5/14 05:19, "Alexei Starovoitov" <alexei.starovoitov@gmail.com> wrote:
> > > 
> > 
> >  On Mon, May 12, 2025 at 8:59 PM Jiayuan Chen <jiayuan.chen@linux.dev> wrote:
> > 
> >  This patch exposes the btf_custom_path feature to bpftool, allowing users
> > 
> >  to specify a custom BTF file when loading BPF programs using prog load or
> > 
> >  prog loadall commands. This feature is already supported by libbpf, and
> > 
> >  this patch makes it accessible through the bpftool command-line interface.
> > 
> >  Signed-off-by: Jiayuan Chen <jiayuan.chen@linux.dev>
> > 
> >  ---
> > 
> >  tools/bpf/bpftool/prog.c | 11 ++++++++++-
> > 
> >  1 file changed, 10 insertions(+), 1 deletion(-)
> > 
> >  diff --git a/tools/bpf/bpftool/prog.c b/tools/bpf/bpftool/prog.c
> > 
> >  index f010295350be..63f84e765b34 100644
> > 
> >  --- a/tools/bpf/bpftool/prog.c
> > 
> >  +++ b/tools/bpf/bpftool/prog.c
> > 
> >  @@ -1681,8 +1681,17 @@ static int load_with_options(int argc, char **argv, bool first_prog_only)
> > 
> >  } else if (is_prefix(*argv, "autoattach")) {
> > 
> >  auto_attach = true;
> > 
> >  NEXT_ARG();
> > 
> >  + } else if (is_prefix(*argv, "custom_btf")) {
> > 
> >  + NEXT_ARG();
> > 
> >  +
> > 
> >  + if (!REQ_ARGS(1))
> > 
> >  + goto err_free_reuse_maps;
> > 
> >  +
> > 
> >  + open_opts.btf_custom_path = GET_ARG();
> > 
> >  I don't see a use case yet.
> > 
> >  What exactly is the scenario where it's useful ?
> > 
> > > 
> > > This patch just exposes the btf_custom_path feature of libbpf to bpftool.
> > > 
> > >  The argument 'btf_custom_path' in libbpf is used for those kernes that
> > > 
> > >  don't have CONFIG_DEBUG_INFO_BTF enabled but still want to perform CO-RE
> > > 
> > >  relocations. Specifically for older kernels, separate BTF files are already
> > > 
> > >  provided: https://github.com/aquasecurity/btfhub-archive/.
> > > 
> > >  If we want load prog using bpftool on those systems, we have to hack
> > > 
> > >  btf__load_vmlinux_btf() before or write custom loader with libbpf and specify
> > > 
> > >  'btf_custom_path'.
> > > 
> > >  I also found a the similar topic:
> > > 
> > >  https://lore.kernel.org/bpf/20220215225856.671072-1-mauricio@kinvolk.io/
> > > 
> > >  Additionally, pwru supports "--kernel-btf" which serves the same purpose as
> > > 
> > >  this patch.
> > > 
> > >  Therefore, using an external BTF file is a common practice.
> > > 
> > 
> >  
> > 
> >  I think it's fine to expose this to bpftool. But maybe call the option
> > 
> >  "kernel_btf" to make it more obvious that this is BTF representing
> > 
> >  kernel types, as opposed to program BTF itself.
> > 
> 

"kernel_btf" is better, thanks.

> Hi Jiayuan, we'll also need to update the documentation (the man page,
> 
> in the summary at the top and in the subcommand description), the
> 
> interactive help message at the bottom of bpftool's prog.c, and the bash
> 
> completion (I can help with it if necessary), please.
> 
> Thanks,
> 
> Quentin
>

Thank you for pointing that out. I did miss those parts, and I also
noticed that others had overlooked similar issues. I will address all
the subcommands of "load prog" in this update.